The effect of highly publicized police killings on policing: Evidence from large U.S. cities☆
Introduction
Recent technological developments have revolutionized the dissemination of information on newsworthy events: internet-based media reporting and social network sharing can result in news spreading instantaneously and extensively. As a result, controversial police use-of-force against civilians easily can be widely publicized. A prime example of this is in the United States, where police killings of African-American civilians frequently have been highly publicized since 2014. For instance, the deaths of Michael Brown (Ferguson, Missouri, August 2014), Freddie Gray (Baltimore, Maryland, April 2015), and more recently George Floyd (Minneapolis, Minnesota, May 2020) soon made national news headlines and drew international attention.
These high-profile killings generate massive ripple effects, which could subsequently reduce policing activities. Specifically, they often spark public outrage toward the police; draw extensive scrutiny of policing practices; and cause probations, firings, indictments, and charges of the involved officers. Such high-profile repercussions could substantially increase police officers’ real and perceived costs of policing, leading many to speculate that police officers have responded by exercising “de-policing” (proactively withdrawing from law enforcement activities). For example, former Federal Bureau of Investigation (FBI) Director James Comey suggested that there was “a chill wind blowing through American law enforcement” in the wake of highly publicized police killings that was “surely changing behavior” (Comey 2015). In addition, the controversial killings of civilians could greatly hurt police legitimacy and damage community trust in the police, potentially alienating civilians from the police. If lack of trust makes civilians less cooperative in crime reporting and investigations, law enforcement activities could drop as well.
Existing studies have offered credible evidence that salient police killings lead to fewer policing activities locally.1 However, two important issues that are of particular interest to a wider range of policy makers and the public remain not well understood. One is whether recent highly publicized police killings can have a sweeping and wide-ranging effect on policing across a broad range of locations.2 The other is that, despite many studies documenting reductions in law enforcement activities following high-profile police killings, there is little direct evidence of de-policing. This paper fills those gaps in two ways. First, we examine the high-profile police killings in the United States during 2014–2016 and study their combined effect on policing in large U.S. cities. Second, the novel data on police self-initiated activities – those initiated by and not assigned to police officers – allow us to investigate de-policing by focusing on policing activities driven by officers’ own discretion.
Our empirical investigation first presents a single-city analysis of the City of St. Louis, where no highly publicized police killing occurred during our examination period. The city’s administrative data provide rich information, including data on self-initiated activities, allowing us to analyze policing activities in great detail. We begin by estimating the immediate effect of the wildly publicized Ferguson shooting of Michael Brown. To do so, we adopt a regression discontinuity (RD) design, comparing policing outcomes at the Census tract level right before and right after the shooting. We estimate a 62 percent reduction in police self-initiated activities, which provides direct evidence of de-policing. Similarly, we document a 42 percent reduction when examining arrest, which is arguably the most studied routine policing activity. We further show that the estimated reduction in misdemeanor arrests, on which police officers can exercise more discretion, is larger than that in felony arrests, presenting indirect evidence of de-policing.
Next, we estimate the longer-run effect of the high-profile police killings occurring in 2014 and 2015. We turn to a different empirical strategy, examining changes in policing outcomes in St. Louis’s Census tracts with the highest black population shares before and after 2014Q3, relative to changes in tracts with relatively low black population shares. Compared to the RD design, this difference-in-differences (DD) strategy focuses on identifying differences in policing outcome trends in a longer time horizon; uses a different set of treatment and control groups; and exploits the potentially larger effect of the high-profile deaths of African-Americans within black-concentrated areas. Our DD estimates continue to provide evidence of de-policing and of reduced law enforcement activities in general, showing large relative reductions in self-initiated activities (17.7 percent) and arrests (23.1 percent), respectively.
The rich policing data from St. Louis allow us to perform additional exercises, shedding more light on the main results of our single-city analysis. For example, our officer-level RD analysis finds a similar reduction in self-initiated inspections (65.5 percent). We also explore the possible reassignment of police officers and find that it is not likely to drive our results. Further, we find evidence of damaged community trust in the police: civilians in tracts with the highest black population shares made fewer 911 emergency calls to seek help with the least urgent issues following high-profile police killings.
In order to better examine whether high-profile police killings have had a widespread effect on policing since Ferguson, we conduct a multi-city analysis of 60 large U.S. cities. Using a DD strategy similar to that in our single-city analysis, we compare the trends of policing outcomes between cities with the highest and relatively low black population shares before and after 2014. Once again we find evidence of reduced policing activities, estimating a 26.4 percent reduction in arrests. While systematic data on self-initiated activities at the city level are not available, we still find suggestive evidence of de-policing, showing a larger reduction in misdemeanor arrests than in felony arrests.
We also examine the potential effect on crime using similar identification strategies. In the single-city analysis, our DD estimates indicate that the Ferguson shooting led to more violent crimes in St. Louis’s Census tracts with the highest black population shares during 2014–2015: a 10.3 percent increase in homicides and a 19.7 percent increase in aggravated assaults. By comparison, our multi-city analysis suggests that such effects might not be present across large cities.
Mainly, this study makes two important contributions. First, it shows that recent highly publicized police killings can have a wide-ranging effect on policing, resulting in fewer law enforcement activities. Our single-city and multi-city analyses demonstrate that these incidents during 2014–2016 led to fewer arrests in large U.S. cities. Second, this study presents the most credible evidence of de-policing to date. Examining police self-initiated activities in St. Louis, we directly show that police officers responded to high-profile police killings by retreating from law enforcement activities. Some caveats should be noted though. Our multi-city analysis is not able to investigate de-policing directly due to data limitations, although it does so indirectly by examining misdemeanor arrests. In addition, this study does not attempt to quantify to what extent the reduction in policing activities, such as arrests, is attributed to de-policing, which is interesting per se. Overall, our finding helps inform the ongoing vigorous debate on the nationwide impact of high-profile police killings on policing and validates a growing concern over de-policing in the post-Ferguson era.
Our work contributes to an extensive literature on policing by providing a detailed analysis of the effects on police self-initiated activity and arrest.3 It joins a growing literature on how exogenous shocks, such as excessive police use of force, scandals, and budget cuts, affect policing (Ba and Rivera, 2019, DeAngelo and Hansen, 2014, Heaton, 2010, Shi, 2009). It is also related to studies that have examined proactive policing (Grogger, 2002, Ridgeway et al., 2019), racial profiling (Antonovics and Knight, 2009, Anwar and Fang, 2006), and the racial or gender composition of the police force (McCrary, 2007, Miller and Segal, 2019). Particularly, our findings illustrate that law enforcement officers respond to perceived changes in incentives (Baicker and Jacobson, 2007, Cheng and Long, 2018, Garrett and Wagner, 2009, Makowsky and Stratmann, 2009, Mas, 2006).
This study also contributes to the broad economics of crime literature (Becker 1968). Particularly, it is closely related to studies that investigate the effects of high-profile incidents (e.g., terrorist attacks) on policing and crime (di Tella and Ernesto, 2004, Draca et al., 2011).
Section snippets
Highly publicized police killings during 2014–2016
Police killings of civilians are not uncommon in the United States. According to the data compiled by FatalEncounters (https://fatalencounters.org), there were at least 25,000 civilians who died in encounters with the police between 2000 and 2018, or approximately 1300 deaths per year. However, it was not until 2014 that such incidents reached a tipping point and started to attract heightened public attention. The landmark incident was the police shooting of Michael Brown in Ferguson, Missouri,
Single-city analysis: St. Louis
We begin by asking whether high-profile police killings could affect policing in St. Louis, Missouri, a large city with population over 300,000 and located about 10 miles southeast of Ferguson.19 The purpose of our single-city analysis is twofold. First, we are able to
Multi-city analysis
The single-city analysis illustrates that highly publicized police killings reduced policing activities in St. Louis, even though no such high-profile incidents occurred there. Are those effects likely to occur in other large U.S. cities? To answer this question, we undertake a multi-city analysis of 60 large cities. We focus on estimating the combined effect during 2014–2016 using the FBI’s Uniform Crime Reporting (UCR) data on arrest.
Discussion
It has been widely speculated that recent highly publicized police killings caused de-policing. Our analysis of St. Louis strongly supports this hypothesis by providing direct evidence of de-policing. Specifically, we find statistically significant evidence that police officers in St. Louis responded to high-profile police killings by substantially reducing self-initiated activities. In addition, by showing that high-profile police killings led to more reductions in misdemeanor arrests than in
The effect on crime
After demonstrating that highly publicized police killings have reduced policing activities in large U.S. cities, we further explore their potential impact on crime.
First, those high-profile killings had the potential to lead to more crimes through the channel of policing. Economics studies have well documented an inverse policing-crime relationship: more policing activities are expected to reduce crimes through incapacitation (e.g., arresting criminals) (Berthelon and Kruger, 2011, Jacob and
Conclusion
With today’s technology capable of virally disseminating news and information, controversial use-of-force by police can become publicized quickly and widely. In recent years, high-profile police killings of African-American civilians have drastically changed the social landscape in the United States. In this paper, we show that these high-profile incidents can have a wide-ranging effect on policing.
Our single-city analysis of St. Louis and multi-city analysis of 60 large U.S. cities, combined
References (68)
- et al.
Finders keepers: forfeiture laws, policing incentives, and local budgets
J. Publ. Econ.
(2007) - et al.
Risky behavior among youth: incapacitation effects of school on adolescent motherhood and crime in Chile
J. Public Econ.
(2011) - et al.
Improving police services: evidence from the French quarter task force
J. Public Econ.
(2018) The effect of insurance on emergency room visits: an analysis of the 2006 Massachusetts health reform
J. Public Econ.
(2012)- et al.
Was there a Ferguson effect on crime rates in large us cities?
J. Crim. Just.
(2016) The limit of oversight in policing: evidence from the 2001 Cincinnati riot
J. Publ. Econ.
(2009)- et al.
De-Policing and crime in the wake of Ferguson: racialized changes in the quantity and quality of policing among Missouri police departments
J. Crim. Just.
(2017) Subways, strikes, and slowdowns: the impacts of public transit on traffic congestion
Am. Econ. Rev.
(2014)- et al.
A new look at racial profiling: evidence from the Boston police department
Rev. Econ. Statist.
(2009) - et al.
An alternative test of racial prejudice in motor vehicle searches: theory and evidence
Am. Econ. Rev.
(2006)
Clearing the air? The effects of gasoline content regulation on air quality
Am. Econ. Rev.
Crime and punishment: an economic approach
J. Polit. Econ.
# Ferguson: digital protest, hashtag ethnography, and the racial politics of social media in the United States
Am. Ethnol.
“Police don't like black people”: African-American young men's accumulated police experiences
Criminol. Public Pol.
Gender, race, and urban policing: the experience of African American youths
Gender Soc.
Green infrastructure: the effects of urban rail transit on air quality
Am. Econ. J.: Econ. Pol.
In search of attention
J. Finance
The effect of driving restrictions on air quality in Mexico City
J. Polit. Econ.
Life and death in the fast lane: police enforcement and traffic fatalities
Am. Econ. J.: Econ. Pol.
Panic on the streets of London: police, crime, and the July 2005 terror attacks
Am. Econ. Rev.
Red Ink in the Rearview mirror: local fiscal conditions and the issuance of traffic tickets
J. Law Econ.
Cited by (12)
The disproportionate impact of post-George Floyd violence increases on minority neighborhoods in Philadelphia
2023, Journal of Criminal JusticeA dynamic factor model to predict homicides with firearm in the United States
2023, Journal of Criminal JusticeBlack Lives Matter's effect on police lethal use of force
2023, Journal of Urban EconomicsThe effects of government labour intervention on firm performance and innovation
2024, Applied Economics
- ☆
We thank Joshua Gottlieb (co-editor), two anonymous referees, Steve Billings, Kitt Carpenter, Mark Cohen, Gregory DeAngelo, Jennifer Doleac, John Edwards, Anna Harvey, Mark Hoekstra, Ilyana Kuziemko, Anita Mukherjee, Jens Ludwig, Justin McCrary, Emily Owens, Jaevin Park, Jamie Bologna Pavlik, Jonathan Pritchett, Stephen Ross, Megan Stevenson, Matthew Zaragoza-Watkins, and seminar participants at Louisiana State University, Tulane University, University of Alabama, University of Mississippi, Vanderbilt University, Xiamen University, the Econometric Society Asian Meeting and European Meeting, the NBER Summer Institute, and the Southern Economic Association Annual Meeting for helpful discussions and suggestions. All errors are our own. Long gratefully acknowledges financial support from the Carol Lavin Bernick Faculty Grant and the Kurzius Family Early Career Professorship at Tulane University.