Date of Degree


Document Type


Degree Name



Criminal Justice


Michael G. Maxfield

Committee Members

Brian Lawton

Jon M. Shane

Bryce Peterson

Subject Categories

Criminology | Criminology and Criminal Justice


Police shooting; Fatality; Open source; Problem-solving; Life-saving


Purpose and Significance: Police shootings do not always result in death. This fact raises a question of what distinguishes fatal and non-fatal police shootings (FNFPS). However, no existing database is available to address how often and under what circumstances civilians have died from or survived police shootings in the United States. To fill the gaps, this dissertation research uses open sources to create a crowdsourced national database on FNFPS in the United States in 2015. The creation of this database provides researchers insights into the suitability and sustainability of open-source research applied for studying police shootings and offers practitioners a guideline to develop a use of deadly force data reporting and collection system. Using this database, the study aims to identify incident-, context-, and agency-level factors that distinguish FNFPS. Findings can help us better understand how selected features of police agencies, situations, and contexts interact with key parts of police shootings: officers, civilians, and places.

Methods: A police shooting problem-solving framework was adopted to guide the coding and collection of incident-, context-, and agency-related information. This study identified eligible police shootings (N=1,907) from the A research team was built to code incident attributes described mainly by local media reports. After examining intercoder reliability and missing values of incident attributes, this study appended additional situational, contextual, and organizational information obtained from multiple publicly available databases. After descriptive analysis of the relative incidence of FNFPS by selected incident-, context-, and agency-level variables, this study performed a series of logistic regression models to identify dangerousness-related and life-saving-related factors that impacted the relative odds of fatal and non-fatal police shootings. When modeling with organizational covariates, a random intercept was specified to control for the observed significant heterogeneity among agencies. Multiple imputation by chained equations was used to handle missing data that largely existed on civilian race and age. Multiple-imputation estimates were used to complement the interpretations of the results from complete case analyses.

Findings: The observations and results through this open-source data collection confirm that open sources, particularly local media outlets, can constitute a more complete universe of police shootings nationwide. But the newsworthiness-driven media coverage restricts the content and amount of information that can be used for theory testing. The results, based on the data complied, support the notion that police shooting does not always kill, and add evidence to the ongoing debates on whether or not police shootings are just urban problem. More importantly, the results demonstrate that conventional scholarly accounts of police decision to shoot may not further explain police shooting fatality. There may exist divergent mechanisms underlying the occurrence and outcome of police shootings.

Conclusion: Reducing fatality differs from reducing police shootings. A more complete understanding of police shooting fatality requires considering factors facilitating or precluding life-saving interventions that could be implemented in the post-shooting stage. This also requires a research agenda on systematically integrating data from law enforcement, public health, academia, and the media.