This comment reports a preliminary examination of variations over time in reported killings of and by the police. Most of the data we examine was collected from police departments by the Uniform Crime Reporting Program as part of a statistical compilation of lethal violence that has reported yearly since 1976. The data on killings of police officers includes all deaths of police reported by agencies as a result of attacks by other persons, and is a high priority for the Uniform Crime Report [UCR] system. Data related to police officer deaths is carefully compiled separately from data relating to other homicides in a yearly Law Enforcement Officers Killed and Assaulted [LEOKA] subset of the UCR. The data on killings by police officers is incomplete in theory because the deaths reported must be classified as "justifiable" by the reporting agency and the UCR, but the vast majority of killings by police officers are included in this classification.

The data set we use in this analysis of killings by police is problematic in three further respects that significantly hamper its utility. The first problem is that the supplemental homicide reports are always incomplete and also vary over time in the number of agencies that report killings by police. The second problem is that very little information about the circumstances that led to the killings by police is reported to the FBI. The third problem is that there is no auditing process to assure the accuracy of what individual agencies choose to report. Even though the data from this program may be the best information currently available in comparison to the alternatives, it must be upgraded to permit effective policy analysis.

This report is divided into four short sections. Part I compares the trends over time in killings of police with time trends in the number of persons killed by police in the almost four decades that both rates have been reported in the United States. A second section discusses the instruments and circumstances that are associated with killings of and by police. A third section examines the changes over time in the ratio of killings of and by police since the mid-1970s. One byproduct of the improvements in protecting police against violence is that while the ratio of citizens killed by police to police deaths was 3.4 to 1 in the 1970s, it has averaged 7.8 to 1 in the most recent reports.

The final section of our note discusses two asymmetrical patterns in the data we analyzed. We have good information on killings of police. There is not a similar emphasis on the careful collection and analysis of killings by police. There is a clear need for careful documentation and analysis of the 400 to 500 civilians killed each year in encounters with police. The current lack of complete and reliable data on killings by police is a scandal. And there are, in the data we examine, reasons to suggest that police use of deadly force is not a necessity of police safety when citizens brandish knives, blunt instruments, or use personal force.