Background
For decades, conventional wisdom has held that domestic incidents are inherently more dangerous to responding officers than other calls for service. That is, officers face an increased risk of being assaulted, injured, or killed on these calls. This was echoed as recently as 2016, when the COPS Office and NLEOMF released a report that alleged domestic disputes are “the most dangerous type [of call] for responding officers.” This conclusion was based on the fact that from 2010 to 2014, 22% of the 91 calls for service resulting in a police officer fatality were domestic disputes.
But that sort of analysis suffers from the same “denominator” problem that has historically plagued studies of racial disparities in traffic stops and officer-involved shootings. In order to make sense of how often officers were assaulted during domestic and non-domsetic calls, respectively, we need to know how often they were not assaulted. We can’t simply work backward from the number of fatalities that occur each year. The overwhelming majority of police-citizen interactions end without an officer being assaulted or killed. So, a better method is to consider an officer’s odds of being assaulted on a domestic call relative to an officer’s odds of being assaulted on other types of calls.
Our Forthcoming Study
In a paper recently accepted for publication at Justice Quarterly, my colleagues (Tara Richards, Gillian Pinchevsky, and Emily Wright) and I took this approach with 2016 National Incident Based Reporting System (NIBRS) data. NIBRS data is ideal for this research question because it includes information about victim-offender relationships as well as whether officers were assaulted and/or injured, among other things. Among the 1 million+ incidents with a documented victim-offender relationship in 2016, 0.23% of domestic incidents involved an officer being assaulted, versus 2.11% of non-domestic incidents. Similarly, just 0.15% of domestic incidents involved an officer getting injured or killed, versus 0.81% of non-domestic incidents.
In subsequent logistic regression models, we estimated the odds of officers being assaulted or injured/killed while controlling for other relevant factors, including whether a violent offense had occurred, weapons or drugs/alcohol were involved, location, time of day, the number of parties involved, offender sex, and offender age. We arrived at the same conclusion: in 2016, domestic incidents were considerably less likely to result in officers being assaulted or injured/killed.
We can’t be sure why the numbers look this way. It could be that domestics are in fact far less dangerous than non-domestics. Or, it could be that officers have been trained and socialized to take extra precautions on these calls, which offsets any increased danger. The aformentioned COPS/NLEOMF report recommended having at least three officers respond to domestic calls “to adequately separate parties, monitor family members and, if necessary, physically restrain and arrest a suspect” (p. 16). If in fact officers are more likely to have backup on domestic calls, that could also explain the lower assault rates we observed. Unfortunately, NIBRS only includes such information when incidents resulted in an officer being assaulted (but not the hundreds of thousands of incidents wherein an officer was not assaulted).
The FBI aims to transition completely from UCR to NIBRS by 2021. This is exciting, given how much richer NIBRS data are. Limitations remain, but my hope is that they will be addressed moving forward. For example, NIBRS currently instructs agencies to complete data on officer activity type (e.g., responding to a traffic incident, responding to a disturbance) only when an officer is assaulted. A simple change in instruction - to complete this field for all incidents - would greatly improve research on officer wellness in the field. Again, to better understand trends in assaults and injuries to police officers, we need a more solid understanding of trends in non-assaults and non-injuries to police officers.
As always, feel free to contact me if you would like a copy of the paper.
Acknowledgements
Many thanks to Matt Ashby, Tim Dooley, Kyle McLean, Marie and Rob Tillyer, Sam Walker, Scott Wolfe, and the anonymous reviewers for their helpful feedback on earlier versions of the manuscript.
