Police Violence and Citizen Crime Reporting in the Black

Police Violence and Citizen Crime Reporting in the Black Community

Matthew Desmond

Harvard University

Andrew V. Papachristos

Yale University

David S. Kirk

University of Oxford

Corresponding Author: Matthew Desmond, Department of Sociology, Harvard University,

William James Hall, 33 Kirkland Street, Cambridge, MA 02138

E-mail: mdesmond@fas.harvard.edu

Abstract

High-profile cases of police violence—disproportionately experienced by black men—may present a serious

threat to public safety if they lower citizen crime reporting. Using an interrupted time series design, this study

analyzes how one of Milwaukee’s most publicized cases of police violence against an unarmed black man,

the beating of Frank Jude, affected police-related 911 calls. Controlling for crime, prior call patterns, and

several neighborhood characteristics, we find that residents of Milwaukee’s neighborhoods, especially

residents of black neighborhoods, were far less likely to report crime after Jude’s beating was broadcast. The

effect lasted for over a year and resulted in a total net loss of approximately 22,200 calls for service. Other

local and national cases of police violence against unarmed black men also had a significant impact on citizen

crime reporting in Milwaukee. Police misconduct can powerfully suppress one of the most basic forms of

civic engagement: calling 911 for matters of personal and public safety.

Keywords

crime reporting, police, police violence, race, crime, inner city

High-profile cases of police violence against unarmed citizens can undermine the legitimacy of

legal authority. Numerous studies document stark racial disparities in police maltreatment,

finding that black boys and men are disproportionately subject to excessive and sometimes

deadly police force, even after accounting for situational factors of the encounter (e.g., resisting

arrest) and officer characteristics (e.g., age, training) (Terrill and Reisig 2003). Studies show

that police disproportionately use improper and deadly force on black men (Matulia 1982;

Worden 1996).

With the proliferation of portable video recording technology, police misconduct that was once

witnessed only by immediate bystanders can now be viewed around the world; and the rise of

Internet-based news sources and 24-hour news networks provide a powerful medium for

broadcasting police misconduct. In recent years, the American public has been captivated and

outraged by such events. The fatal 2012 shooting of 17-year-old Trayvon Martin by a

neighborhood watch coordinator, for example, was the most followed news story in the United

States for several straight weeks (Pew Research Center 2012). Other instances of law-

enforcement officers killing unarmed black men—such as Eric Garner in New York City (July

2014); Michael Brown in Ferguson, Missouri (August 2014); and Freddie Gray in Baltimore,

Maryland (April 2015)—became national news stories and sparked widespread protest.

Previous research shows that negative encounters with law enforcement, as well as high-profile

cases of police misconduct, contribute to the spread of legal cynicism within black communities

(Kirk and Papachristos 2011; Sampson 2012; Weitzer 2002). Less well understood, however, is

the extent to which legal cynicism affects citizens’ reliance on police. In the aftermath of

publicized police violence against unarmed black men, do residents of black neighborhoods

disengage with the police, especially when it comes to reporting crime via 911? Police work of

every kind relies on citizen participation, especially reports of law breaking (Banton 1964;

Black 1970). If police misconduct lowers crime reporting throughout black communities, it

directly threatens public safety within those communities, many of which already face high

levels of crime.

This study examines how incidents of excessive police violence influence citizen crime

reporting. As our main case study, we examine one of Milwaukee’s most publicized moments of

police brutality: the 2004 beating of an unarmed black man, Frank Jude, by white police officers.

Using an interrupted time series design, we analyze patterns of police-related 911 calls before

and after the Jude beating was made public in the Milwaukee press. Controlling for several

neighborhood characteristics, along with fixed effects for neighborhoods and months, we find

that residents of Milwaukee’s neighborhoods were far less likely to report crime after Jude’s

beating was broadcast. Over half of the total loss in 911 calls occurred in black neighborhoods.

Supplemental analyses of other local and national cases of police violence against unarmed

black men also had an impact on crime reporting in Milwaukee.

LEGAL CYNICISM AND CITIZEN CRIME REPORTING

A spirit of legal cynicism—the deep-seated belief in the incompetence, illegitimacy, and

unresponsiveness of the criminal justice system (Kirk and Papachristos 2011)—is thought to

pervade many poor, minority communities. Legal cynicism is widely believed to result in

citizens withdrawing from the police, particularly by refusing to report crime to the authorities

(Baumer 2002). What emerges, in Anderson’s (1999:66) classic formulation, is the code of the

street, “a kind of people’s law” based on respect and payback. Residents who refuse to seek help

from the police are thought to rely on the code to dole out sanctions through violence or other

informal means (Kirk and Papachristos 2011; Sampson 2012).

The vast majority of research on citizens’ views toward the police and law breaking is based on

interviews (e.g., Brunson and Miller 2006; Carr, Napolitano, and Keating 2007) or surveys (e.g.,

Avakame, Fyfe, and McCoy 1999; Baumer 2002). Interviews and surveys can serve as effective

methods to capture individuals’ attitudes and cultural scripts, but studies show surveys are less

reliable when it comes to measuring interactions with the police (Rosenfeld, Jacobs, and Wright

2003). Moreover, a close reading of ethnographic reports reveals that the code’s manifestations

may not be as uniform as is often believed. At some points, Anderson (1999:72–73, 112) refers

to the code as a belief system or cultural frame, but at others he presents it as a dissemblance

designed to prevent, not encourage, violence. Cold looks and bold boasts aside, the people

Anderson observed often relied on the police, a pattern observed by other urban ethnographers

as well (e.g., Sánchez Jankowski 1991; Venkatesh 2000).

When it comes to relying on and cooperating with the police, what one does might not resemble

what one says one will do. The legal cynic may report crime just as regularly as the citizen who

views the criminal justice system with reverence. Studies that measure attitudes toward the

police and hypothesized citizen crime reporting find racial differences in the former but not the

latter. Bobo and Thompson (2006) asked respondents (1) if they would call the police and file a

report if their house was burglarized and (2) if they believed their complaint would be taken

seriously. The vast majority of both white and black respondents (90 and 93 percent,

respectively) said they would call the police; however, only 35 percent of blacks expected to be

taken seriously, compared with 60 percent of whites. Although sizeable racial differences were

observed in peoples’ attitudes toward the police, respondents were virtually identical in their

surmised actions with law enforcement. People’s attitudes toward the criminal justice system

might be poor, even misleading, indicators of their real-life dealings with the police.

STUDY DESIGN AND RESEARCH QUESTIONS

This study moves beyond the limitations of previous research by analyzing the effects of

publicized police violence cases on crime reporting via 911 calls. The evidence for this study

does not rely on accounts of past behavior or projections of future action but on the complete

universe of every crime reported in Milwaukee over the course of seven years. Calling 911 to

report a crime is the most basic act of reliance on (and engagement with) the legal system in

contemporary society (Manning 1988; Reiss 1992). It is also foundational to public safety,

being the first and among the most important acts of cooperation with the police, who in many

cases can respond to criminal activity only if it is reported.

In studying the effects of highly publicized cases of police misconduct on citizen crime

reporting, our study also investigates a possible new mechanism for a reluctance to cooperate

with formal law enforcement agencies. Recent studies find that one’s attitude toward the police

can be influenced by vicarious experiences of one’s social ties (Brunson 2007; Rosenbaum et al.

2005) and by repeated exposure to news reports of misconduct (Weitzer and Tuch 2005). Some

studies document a drop in public approval of the police after publicized cases of police

brutality, especially among African Americans (Jefferis et al. 1997; Weitzer 2002). These

studies are primarily descriptive, and they do not control for crime rates. But they do suggest

that police violence may have reverberations that extend well beyond the trauma and tragedy of

an individual act—or even the neighborhood and city in which it took place.

To examine the relationship between publicized reports of police violence and citizen crime

reporting, we examine 911 calls with an interrupted time series design, testing for changes in the

volume of 911 calls before and after the Milwaukee Journal Sentinel broke the story of the 2004

beating of an unarmed black man, Frank Jude, by white police officers. We are interested in

four aspects of calling patterns. First, we examine the direction of 911 calls reporting crime,

observing if they increased, decreased, or experienced no change. Second, we examine patterns

in different types of calls: namely through analyses of all calls reporting crime and then

separately for calls reporting violence. Third, we focus on the location of calls, observing if

calling patterns in white and black neighborhoods differed after the Sentinel reported Jude’s

assault. Fourth, should there be a “Jude effect” on police-related 911 calls, we investigate its

duration. Additionally, we examine if citizen crime reporting in Milwaukee was affected by

three other cases of police violence against unarmed black men: the 2007 assault of Danyall

Simpson in Milwaukee; the 2006 killing of Sean Bell in Queens, New York; and the 2009

killing of Oscar Grant in Oakland, California. The Simpson case was covered in the local press;

the other two cases were major stories in the national press.

In each case we analyze, we assume that breaking the story in mainstream media outlets was

what brought the incidents to public light, even if community members did not themselves learn

of the incidents through such outlets. Milwaukee residents could have learned of these incidents

through public protest, word of mouth, and alternative media, including the black press and

underground rap songs, which circulated after the deaths of Oscar Grant and Sean Bell.

Nevertheless, news of such incidents spread through various channels only after major media

outlets broke the story. This was especially apparent in Frank Jude’s case. More than three

months elapsed between Jude’s beating and it being reported in the local press. During that time,

there were no public protests; unrest did follow the news story. Additionally, supplemental

analyses find that in black neighborhoods, police-related 911 calls were unaffected in the weeks

immediately following Jude’s beating. This suggests that although knowledge of an incident of

police violence may spread through personal networks close to the victim, those networks are

not expansive enough to affect the crime-reporting behavior of an entire community in a

compressed timespan (see Andrews and Biggs 2006). Even if they were, media reporting on

such episodes draws on medical and police reports, interviews with eyewitnesses, and other

sources, providing resonance and legitimacy to events that otherwise could be dismissed as

hearsay or overlaid with multiple interpretations rooted in assumptions (Koopmans 2004).

THE BEATING OF FRANK JUDE

What follows is an account of the events surrounding the beating of Frank Jude based on news

reports and court documents (listed at the end of the article). To understand how these events

could have affected citizen crime reporting in Milwaukee, it is necessary to provide a somewhat

detailed account of the severity of Jude’s beating and its aftermath.

On October 23, 2004, Milwaukee police officer Andrew Spengler hosted a house warming party

in a middle-class white neighborhood. Several police officers attended. After 2:30 a.m., Frank

Jude and Lovell Harris, both black men, arrived at the party as guests of two female college

students. The group of four began to feel uncomfortable and left the party, returning to the truck

in which they had arrived. But before they could pull away, at least 10 men surrounded the truck

and accused Jude and Harris of stealing Spengler’s police badge. Jude, Harris, and their hosts

were pulled from the truck. After having his face slit with a knife, Harris would escape. Jude

would not (Diedrich 2005b).

Spengler and several other off-duty police officers began beating Jude about the face and torso

while another pinned his arms behind his back. When Jude collapsed to the ground, the officers

kicked his head. A pair of on-duty officers arrived around 3:00 a.m. (Jude’s female companions

had called 911 before partygoers seized their phones.) One uniformed officer, Joseph Schabel,

stomped on Jude’s face until he heard bones breaking. The other on-duty officer watched. An

off-duty officer picked Jude up and kicked him in the crotch with such force his feet left the

ground. Another took one of Schabel’s pens and pressed it deep into Jude’s ear canals. Another

bent Jude’s fingers back until they snapped. Spengler put a gun to Jude’s head. Jude was left

naked from the waist down lying on the street in a pool of his own blood. At 3:09 a.m. other on-

duty officers arrived at the scene. Jude was taken by police wagon to the emergency room,

where the admitting physician took photographs of him because his injuries were too extensive

to document in writing. Spengler’s badge was never recovered (Diedrich 2005b; USA v. Bartlett,

Spengler, and Masarik 2009).

The police officers returned to work, and Jude began a slow recovery. But on February 6, 2005,

Sunday readers of the Milwaukee Journal Sentinel were met with the following front-page

headline: “Police Suspected in Man’s Beating” (Diedrich 2005b). Next to the headline was one

of the photographs the attending physician had taken of Jude’s swollen face. The story told of

the beating and the subsequent obstructionism and noncooperation by the police: “Authorities

knew right away who was involved, but more than three months later no one has been charged

or even arrested.” It also listed Jude’s injuries: a concussion; broken noise and fractured sinus

cavity; cuts in both ears; cuts and swelling to the head, eyes, neck, face, legs, and back; a left

eye swollen shut that bled for 10 days; possible permanent disability in one hand; compromised

vision; night terrors; and panic attacks.

Citizens, many of them black, took to the streets in protest. Four days after the story broke,

more than 100 people marched in front of the District Attorney’s office (Nunnally 2005). A

month later, nine officers were fired and four others disciplined. All but one appealed and drew

their salaries. A year later, Spengler and two fellow officers were tried and acquitted of all

charges by an all-white jury (Stingl 2006). More protests followed, prompting a federal

investigation. Later that year, eight officers were federally charged. Seven were convicted

(Borowski 2007; Diedrich 2007a).

DATA AND METHODS

Calls for Service Data

We obtained every 911 call placed in Milwaukee between March 1, 2004 and December 31,

2010 (N = 1,104,369). These data, provided by the Milwaukee Police Department (MPD),

include the date and address of each call, a city code and description of each call, and a

disposition code reporting how the call was resolved. This information allowed us to limit our

analysis to police-related calls by dropping those having to do with traffic and vehicular

accidents; injuries, fires, and medical emergencies; internal police calls (recovered property,

assisting an officer); and calls triggered by security alarms.

Our analyses focus on 911 calls

placed the year before and after a given event: for example, we examine police-related calls in

2004 and 2005 when estimating the impact of the Frank Jude beating on citizen crime reporting.

Our analyses draw on all 596 city block groups, the average population of which is roughly

1,000 people. We exclude from our analyses any 911 calls made from block groups beyond

Milwaukee city limits, bringing the total number of calls reporting crime to 883,146. We use

“neighborhood” and “block group” interchangeably and designate a neighborhood as black or

white if these respective groups account for at least two-thirds of neighborhood residents. Last,

all crimes recorded by the MPD between 2004 and 2010 were geocoded, aggregated to the

block group, and merged with our dataset.

Measures

Dependent variables. We examine the weekly number of 911 calls at the block-group level.

Calls are divided into two categories: total and violent calls. Total calls include all police-related

911 crime reports, not only those for violent and property crime but also reports of public

disturbances and nuisances. Calls for violence include those for battery, robbery,

weapons/shootings, sexual assault, kidnapping, and threats to person. When attempting to

isolate the effect of the beating of Frank Jude on citizen crime reporting, the dependent variable

is the number of weekly calls per block group during the observation period: March 2004 to

December 2005.

This observation period centers on the date the Milwaukee Journal Sentinel

first reported Jude’s beating (February 6, 2005), allowing us to analyze roughly the same

number of weeks before and after the story broke.

Pre-intervention time trend. Two parameters define the pattern of 911 calls prior to the

Sentinel’s report of Jude’s beating: the initial level of calls and the slope indicating the time

trend. The intercept value in our models indicates the estimated number of 911 calls per block

group at the beginning of the observation period. The weekly pre-Jude time measure indicates

the pre-intervention change in the volume of 911 calls over time (Singer and Willett 2003).

Modeling the time trend allows us to establish a counterfactual for what the calling pattern

would have been like in the absence of Frank Jude’s story. We estimated the pre-intervention

time trend with a linear specification as well as a quadratic specification, comparing model fit

between specifications using the Bayesian Information Criterion. A linear pre-Jude specification

provided a better fit to the data (see Table A1 in the Appendix).

Explanatory variables. To assess the impact of the Jude beating on 911 calling behavior, we test

for possible shifts in the level of 911 calls in a block group as well as changes in the slope of the

time trend. We do so by including a binary measure equal to 0 for observations prior to the news

story of Jude’s beating and equal to 1 for the period after the story broke. To measure an

interruption in the slope of the time series, we include linear and quadratic time variables,

measured as the number of weeks since the story broke. For observations prior to the story,

these time variables are set to 0. As with the pre-intervention time trend, we use model fit

statistics to assess whether the post-intervention trend is more appropriately modeled with linear

or quadratic time specifications (see Table A1). The quadratic post-Jude specification fits best.

By including a quadratic term for the post-Jude time period, we are effectively testing for a

discontinuous effect of the Jude beating on 911 calls as well as modeling the possibility that 911

calls eventually drift back to the pre-Jude trend.

Fixed effects. We include block-group and month fixed effects in our models. Block-group fixed

effects control for all time-invariant block-group-specific factors potentially correlated with 911

calls. Month fixed effects control for seasonal variation in crime and crime reporting, as calls

peak in the summer and slump in the winter.

Control variables. To account for the structural characteristics commonly associated with

neighborhood-level crime (Land, McCall, and Cohen 1990), we draw on U.S. Census data and

calculate the following measures at the block-group level: percent of the population that is non-

Hispanic black; percent of the population that is Latino; percent of households below the

poverty line; and percent of households that rent.

Our models also include weekly crime counts to rule out the possibility that the observed levels

of 911 calls are nothing more than an indicator of neighborhood crime levels (Kirk and Matsuda

2011; Klinger 1997). In our models, the control variable reported crime corresponds to the total

number of weekly crimes in each block group during the observation period. Derived from

aggregate data generated by the Milwaukee Police Department for the FBI’s Uniform Crime

Report, this variable is organized into two categories: total and violent crimes. When our

outcome is all 911 calls, we control for all reported crime during the same week; when it is calls

for violence, we control for violent crime. If legal cynicism increased after the media broke the

story of Frank Jude’s beating, then it is likely that crime also increased during this time period.

For the same reason, it is also likely that some crime went unreported to the police (Kirk et al.

2012). If neighborhoods experienced a concomitant increase in crime and decrease in citizen

crime reporting, then any negative estimate of the effect of Jude’s beating on citizen crime

reporting is conservative. In other words, we model a counterfactual 911 calling trend based on

pre-Jude calling and crime patterns, but the gap between estimated and counterfactual calling

trends is probably larger than we observed because, post-Jude, unreported crime likely increased.

Descriptive statistics are presented in Table 1, which reveals a picture of the city’s social,

demographic, and crime patterns. In the average Milwaukee neighborhood, there are

approximately three 911 calls a week. However, there is wide variation in call levels, ranging

from 0 to 41 weekly calls per neighborhood.

<<< Table 1 about here >>>

Methods

To assess the effect of the Frank Jude beating on citizen crime reporting, we use an interrupted

time series design with block-group and month fixed effects. With this design, we assess

whether the extensive media coverage of Jude’s beating that followed the Sentinel’s breaking

story induced an interruption in the level and slope of the time series of 911 calls. We use a

negative binomial specification to account for the skewness of the 911 calling pattern, with

many block groups reporting few or no 911 calls in a given week. We measure exposure as the

number of residents in a block group, which we determine from Geolytics annual population

estimates.

Previous research suggests that African American communities would be more affected than

white communities by the Jude beating (Kirk and Papachristos 2011; Sampson 2012). Hence, in

addition to analyzing all block groups, we also estimate models separately for black and white

neighborhoods. We hypothesize that the effect will be more pronounced in black block groups.

Additionally, we replicate our analyses on the number of weekly calls for automobile accidents,

a “nonequivalent dependent variable” that should have been unaffected by the Jude beating

(Shadish, Cook, and Campbell 2002:184). Because calls related to car accidents are typically

made to report injuries or for insurance purposes, having little to do with trust in law

enforcement, we expect them to be unaffected by the incident.

Additional Cases

To examine if the Frank Jude case was unique, we also investigate if three other cases of police

violence against unarmed black men affected citizen crime reporting in Milwaukee. The news

reports and court documents on which this section draws are listed at the end of the article.

Danyall Simpson, Milwaukee, Wisconsin; story broke in the local press on February 18, 2007.

On May 21, 2006, Danyall Simpson, a 19-year-old black man, was confronted in the driveway

of his mother’s home by police officer James Langer, a 32-year-old white man. Simpson was

suspected of throwing a brick through a neighbor’s window. The interaction between Simpson,

Langer, and Langer’s partner grew confrontational and resulted in Simpson being taken to the

ground. While Simpson was lying on his back with his arms pinned, Langer hit him in the face

with his flashlight. The blow broke Simpson’s eye socket and cheekbone. The Milwaukee

Journal Sentinel reported the story on February 18, 2007 (Diedrich 2007b). Readers learned that

Langer had a history of using excessive force but was not reprimanded; that it took the MPD

three months to deliver the case to the District Attorney’s office; and that according to a witness,

Langer’s hit “sounded as if someone threw a melon in the air and smashed it with a baseball bat.”

The Sentinel printed a photograph of Simpson’s bloodied face.

Sean Bell, Queens, New York; story broke in the national press on November 27, 2006. In the

morning hours of November 25, 2006, Sean Bell, a 23-year-old black man due to be married

later that day, was shot and killed by New York City police officers. Five plainclothes officers

conducting an undercover investigation of a prostitution ring fired over 50 rounds at the car that

held Bell and two friends. Bell had been celebrating his last night of bachelorhood at a Queens

strip club and was unarmed when killed. The story made national headlines the following day,

with coverage in the New York Times, CNN, and other major outlets, eliciting large protests in

New York City and beyond (Buckley and Rashbaum 2006; Fransisco, Frieden, and Rose 2006).

Oscar Grant, Oakland, California; story broke in the national press on January 8, 2009. In the

first hours of 2009, Oscar Grant III, a 22-year-old black man, was detained on the platform of

the Fruitvale Station by Bay Area Rapid Transit police officers, who had received word of a

possible fight underway in a crowded train returning from San Francisco. Two officers were

attending to Grant, who was lying face down with his hands behind his back, “restrained and

unarmed” according to court documents. One of the officers, Johannes Mehserle, a 26-year-old

white man, then pulled out his .40 caliber sidearm and shot Grant in the back, puncturing his

lung. Grant died the following morning. The incident was recorded by multiple bystanders using

portable cameras and disseminated to Internet and media outlets. Protests immediately followed

in Oakland, but it took a week for the story to receive widespread coverage in the national press.

Mehserle was later found guilty of involuntary manslaughter (Chideya 2009; McKinley 2009).

We selected the Simpson case because it enables us to observe if a local incident of police

violence less extreme than Jude’s beating also affected citizen crime reporting in Milwaukee.

We selected the killings of Sean Bell and Oscar Grant because they were the most nationally

publicized cases of police violence that took place during our study period. These two cases

allow us to investigate if national reports of deadly force used against unarmed black men in

other locales had any observable impact on citizen crime reporting in Milwaukee. For each case,

we estimate an interrupted time series model with block-group and month fixed effects, as with

the Frank Jude case. In contrast to the Frank Jude analyses, model fit statistics revealed that a

specification with quadratic time trends for both pre- and post-intervention periods resulted in

better model fit.

RESULTS

We begin by mapping 911 calls against the racial composition of Milwaukee neighborhoods.

Figure 1 shows two maps of the city: panel A displays the percent of neighborhood residents

who are black; panel B depicts the total number of police-related 911 calls between March 2004

and December 2005. Side-by-side, the maps reveal that predominantly black neighborhoods

have higher call volumes. In an average week, residents in black neighborhoods made 5.64 calls

to 911 (min. = 0, max. 41), more than triple the calling rate of residents in white neighborhoods

(1.7 calls per week; min. = 0, max. = 17).

<<< Figure 1 about here >>>

Did the Beating of Frank Jude Affect Citizen Crime Reporting?

Table 2 displays the results of negative binomial regression models estimating the relationship

between neighborhood characteristics and citizen crime reporting, first for all block groups in

the city, then for subsamples of predominately black and white block groups. Model 1 takes as

its outcome all 911 calls reporting crime recorded in Milwaukee between March 1, 2004 and

December 31, 2005. Model 2 focuses on a subset of calls reporting violent crime. All three

models include block-group and month fixed effects, sociodemographic controls, and the

number of reported crimes during each observation week.

<<< Table 2 about here >>>

Looking first at all neighborhoods, the overall trend for total calls (as seen in the weeks pre-Jude

story coefficient) is positive and significant, indicating that police-related 911 calls trended

slightly upward during the observation period.

The non-significant coefficient for Jude story

indicates that the weekly number of 911 calls did not shift abruptly in either direction

immediately following the publication of Jude’s beating. However, the negative linear post-

intervention slope trend (weeks post-Jude story) finds that 911 calls for service declined

significantly in the period after the story broke. This observed decline is net of controls for

reported crime, neighborhood sociodemographic characteristics, and fixed effects for time-

invariant block-group characteristics and month fixed effects for the seasonality of 911 calls.

The small, positive quadratic time term reveals that the “Jude effect” diminished over time, with

911 calls eventually returning to expected levels. Calls for violent crime declined significantly

following Jude’s story before eventually returning to pre-intervention levels.

Figure 2 displays the estimated pattern of 911 calls before and after the press reported Frank

Jude’s story. We estimated counts using Stata’s margins command following the execution of

our negative binomial regressions. The dotted line represents a counterfactual calling pattern

that extends the pre-Jude time trend into the post-intervention period. It estimates the expected

number of 911 calls in the absence of Jude’s story. As Figure 2 shows, the number of calls

reporting crime declined substantially in the wake of Jude’s beating, relative to the

counterfactual, and did not return to expected levels for more than a year.

<<< Figure 2 about here >>>

Table 2 also contrasts the estimated effect of the Jude beating in black and white neighborhoods.

In tabular form, the results appear similar across neighborhoods. In both subsamples, the change

in the intercept (denoted by the coefficient Jude story) is not significant, but the linear post-story

trend is significant and negative, with the effect appearing larger in white neighborhoods. The

quadratic term is positive but non-significant in the black-neighborhoods subsample but positive

and significant in the white-neighborhoods subsample, indicating that the initial decline in calls

diminished with time in white but not black neighborhoods.

Figure 3 reveals the benefit of visually displaying these results. In black neighborhoods, the

decline in 911 calls is large and durable, with citizen crime reporting dropping and staying

lower than expected for over a year after the story broke. In white neighborhoods, by contrast, a

small decline in 911 calls followed Jude’s story, but the effect dissipated rapidly as 911 calls

eventually surpassed the counterfactual line within a year’s time. The beating of Frank Jude had

a much more substantial effect on citizen crime reporting in black neighborhoods than in white

neighborhoods. With respect to reporting violent crime in particular, calls declined temporarily

in white but not black neighborhoods after the story broke.

<<< Figure 3 about here >>>

Drawing on the results presented in Figures 2 and 3, we calculate a “bottom line” estimate of the

total number of police-related 911 calls that were not made in the wake of Milwaukeeans

learning of Jude’s beating. We do so by summing the difference between the counterfactual

time trend and the estimated post-Jude trend across different call types. We estimate that the

police beating of Frank Jude resulted in a net loss of approximately 22,200 911 calls reporting

crime the year after Jude’s story broke. Over half (56 percent) of the total loss in calls occurred

in black neighborhoods. For comparison purposes, there were approximately 110,000 police-

related 911 calls from all Milwaukee neighborhoods during this time period.

Calls for Automobile Accidents

To test whether calls for incidents that should not have been affected by Frank Jude’s beating

also declined during the observation period, we analyze calls reporting automobile accidents. If

we find that calls for accidents declined following Jude’s story, we would have reason to be

concerned that some other concurrent event besides the beating—for example, reductions in

staff capacity at 911 call centers or a local sports team winning the championship—had

influenced crime reporting.

As with our previous analyses, we compared model fit to different specifications of the pre- and

post-Jude time period, finding that a model with a linear specification for both time periods

provided the best fit. As Table 3 shows, 911 calls related to traffic accidents increased during

the weeks following Jude’s story, net of sociodemographic controls and weather conditions

(weekly rainfall and snowfall).

These findings show that the observed decline in police-related

911 calls that followed Jude’s story was not a product of a general decline in all 911 calls; rather,

the decline was specific to crime reporting.

<<< Table 3 about here >>>

Danyall Simpson, Sean Bell, and Oscar Grant

Having found evidence indicating that the story of Frank Jude’s beating depressed citizen crime

reporting, particularly in Milwaukee’s black neighborhoods, we now investigate whether 911

calls were affected by other cases of police violence against unarmed black men documented in

the local and national press. We replicate our fixed-effects regression analyses for Danyall

Simpson’s beating and the fatal shootings of Sean Bell and Oscar Grant. Table 4 displays the

results for all neighborhoods and a subsample of black neighborhoods. We focus on the weekly

number of 911 calls at the neighborhood level, following the reporting of each incident (for

Simpson: February 18, 2007; for Bell: November 27, 2006; for Grant: January 8, 2009).

<<< Table 4 about here >>>

Consistent with our findings from the Frank Jude case, 911 calls reporting crime declined

significantly following the breaking of Danyall Simpson’s beating. Both the level and slope of

crime reporting declined. This finding holds when examining calls reporting violent crime.

Likewise, citizen crime reporting fell significantly in black neighborhoods after publication of

the Simpson beating.

Local incidents of police violence, when publicized in the press, dampen crime reporting in

Milwaukee. But could non-local incidents reported in the national press also affect citizen crime

reporting? Our findings suggest they do, although not to the extent of local events. After Sean

Bell’s killing made national headlines, the level of 911 calls declined significantly—in all

Milwaukee neighborhoods and in black neighborhoods, particularly—however, we do not

observe a significant change in the slope of the post-story time trend. By contrast, the killing of

Oscar Grant in Oakland, California, did not result in a decline in police-related 911 calls in

Milwaukee. In fact, 911 calls increased in the weeks following Grant’s death, a trend that may

reflect changes within Milwaukee neighborhoods in the wake of the Great Recession.

In two of our three additional cases, we observed an association between police officers’ use of

excessive or deadly force against an unarmed black man and fewer crimes reported to the police.

The effect of Frank Jude’s beating on citizen crime reporting matched the severity of the

incident, but instances of police violence less extreme by comparison, as well as events that

took place thousands of miles away, still appear to have had an impact on crime reporting in

Milwaukee.

DISCUSSION

Cities across the United States have experienced high-profile cases of excessive use of police

force (Holmes and Smith 2008; Ross 2000). Yet little work has been dedicated to understanding

their effects on police-community relations. Applying stringent statistical analyses to hundreds

of thousands of 911 calls, we sought to identify a relationship between publicized cases of

police violence against unarmed black men and citizen crime reporting.

This study shows that publicized cases of police violence against unarmed black men have a

clear and significant impact on citizen crime reporting. Once the story of Frank Jude’s beating

appeared in the press, Milwaukee residents, especially people in black neighborhoods, were less

likely to call the police, including to report violent crime. Both white and black citizens’

opinions of the police may drop after a high-profile brutality case against unarmed black men

(Lasley 1994; Weitzer 2002), but our study found that only people living in black

neighborhoods altered their crime-reporting behavior for a significant period of time after the

publication of such events. The dampening effect of the Jude beating on citizen crime reporting

persisted for months, resulting in a net loss of more than 20,000 911 calls. It is one thing to

disparage law enforcement in your thoughts and speech after an instance of police violence or

corruption makes the news. It is quite another to witness a crime, or even to be victimized, and

refuse to report it.

The events of that October morning affected not only Frank Jude and his family but Black

Milwaukee as a whole. To date, research has mainly focused on how individuals’ personal

interactions with law-enforcement officers (Skogan 2006; Tyler and Waslak 2004) or the

vicarious experiences of one’s social ties (Brunson 2007; Rosenbaum et al. 2005) influence

attitudes toward the police. This study, however, shows that in predominantly black

neighborhoods, publicized cases of police violence can have a community-wide impact on crime

reporting that transcends individual encounters. An important implication of this finding is that

publicized cases of police violence not only threaten the legitimacy and reputation of law

enforcement; they also—by driving down 911 calls—thwart the suppression of law breaking,

obstruct the application of justice, and ultimately make cities as a whole, and the black

community in particular, less safe.

To gather empirical support for this line of thought, we examined homicides following the

publication of Jude’s beating. Unlike with other crime, murder records should not be affected by

fluctuations in citizen crime reporting, owing to the gravity and nature of the crime. During the

six months that followed Jude’s story (March to August 2005), 87 homicides occurred in

Milwaukee, representing a 32 percent increase in murders relative to the same six-month period

in previous and subsequent years. In fact, as Figure 4 shows, the spring and summer that

followed Jude’s story were the deadliest in the seven years observed in our study. This uptick in

homicides occurred during months in which citizen crime reporting, especially in the black

community, fell by significant margins.

<<< Figure 4 about here >>>

We also found that it does not take a major or local event to reduce crime reporting. Milwaukee

residents issued fewer calls for service after the local press reported the assault of Danyall

Simpson by a white police officer, an assault less severe than Jude’s, and they were less likely

to call the police after the killing of Sean Bell (in Queens, New York) made national headlines.

We found no evidence, however, that the killing of Oscar Grant (in Oakland, California)

lowered crime-reporting behavior in Milwaukee. We found some evidence, then, that high-

profile cases of unarmed black men being assaulted or killed by law-enforcement officers may

affect citizen crime reporting in cities far removed from the site of the incident.

Our analyses of the Frank Jude beating found that crime reporting eventually returned to typical

levels, but it took roughly one year before they did—and even longer in black neighborhoods.

Other studies that examine the impact of major, divisive events on attitudes or behavior have

observed profound but short-lived effects, characterized by a sharp escalation and an equally

sharp de-escalation (King and Sutton 2013; Legewie 2013). The effect of police violence on

citizen crime reporting documented in this study is considerably lengthy by comparison. This

indicates that high-profile cases of excessive police force constitute a severe breach in the social

contract that exists between citizens and the criminal justice system. That breach is so sudden

and violent when unarmed black men are beaten or killed that virtually no institutional response,

from public apologies to sanctioning offending officers, can swiftly repair it. Crime reporting

remained significantly lower in black neighborhoods weeks after officers involved in Jude’s

beating were charged and fired.

With time, however, citizen crime reporting returned to normal levels. Police have a virtual

monopoly on crime control, and alternative forms of law enforcement emerge only under

extreme circumstances (see Venkatesh 2000). Accordingly, residents who witness or are

victimized by crime have little recourse other than to call 911, which is perhaps why people

negatively disposed toward the police often favor increased law enforcement to promote public

safety (Carr et al. 2007; Sampson and Bartusch 1998).

Taken together, our findings present a dynamic, eventful theoretical model that illuminates the

relationship between economic and racial inequality, legal cynicism, and citizen cooperation

with the police. Major episodes of police violence against unarmed black men covered by the

press can interface with a general sentiment of legal cynicism within the black community to

drive behavior, drawing out action from attitudes.

Our findings demonstrate that the media plays a crucial role in this process, not only by raising

awareness of episodes of police violence but also by addressing basic questions about such

episodes. What exactly happened? What was the race of the victim and police officer? Was the

victim armed? When these questions are addressed, either by professional reporters (as in Frank

Jude’s case) or “citizen journalists” who video record and broadcast incidents, that narrows the

number of reasonable interpretations (Koopmans 2004; Scheufele 1999), allowing certain acts

of police force to be seen as unreasonable and disturbing. This can produce widespread effects

on citizen crime reporting when large numbers of people learn of an episode around the same

time (see Andrews and Biggs 2006).

In the short run, publicized episodes of excessive police violence against unarmed black men

can activate legal cynicism within the black community to diminish residents’ cooperation with

law enforcement. In the long run, such episodes likely contribute to that very cynicism by being

incorporated into the community’s collective memory.

We believe collective memory explains

why we found “the Jude effect” to be particularly strong in black neighborhoods. Such incidents

belong to a centuries-old tradition of state-sanctioned assaults on the black body—from slavery

and lynching campaigns to urban riots and mass incarceration (Muller 2012; Patterson 1998)—

an ugly fact that African Americans are more likely to recognize and dread. Against this

backdrop, a violent episode carried out by the police is registered as proof and product of a

violent heritage, rendering victim and perpetrator actors in a larger historical drama. A survey

conducted after a white police officer shot and killed Michael Brown, an 18-year-old black man,

in Ferguson, Missouri, found that 80 percent of African Americans believed the case “raises

important issues about race,” compared to 37 percent of whites (Pew Research Center 2014).

Owing to the legacy of racism in the United States, many members of the black community

perceive an assault of a black man by white police officers as an assault on the black community

writ large, activating a collective response. Within black neighborhoods, the impact of police

violence on citizen crime reporting results not only from African Americans identifying with the

victim but also interpreting the moment of victimization in terms that articulate the nation’s

racial heritage. By extension, we would not expect an equivalent response in predominantly

white or Latino neighborhoods after publicized episodes of police violence against unarmed

white or Latino citizens.

Future research can assess this hypothesis. Because this study relied on administrative records,

its findings can be replicated in other locales. In racial composition, population size, and

unemployment rate, Milwaukee is similar to many midsize U.S. cities (Desmond, Gershenson,

and Kiviat 2015). Like many other urban areas—including cities in which recent well-known

cases of police violence have occurred, such as Baltimore, Cleveland, and New York City—

Milwaukee has high levels of racial segregation. We might not expect the effect of police

violence on citizen crime reporting to be as acute or durable in more integrated cities, a testable

expectation. Our findings can also be extended by investigations of how police maltreatment of

Hispanic victims may affect citizen crime reporting in Hispanic communities, where legal

cynicism is less pronounced than in black neighborhoods (Kirk et al. 2012).

Police departments and city politicians often frame a publicized case of police violence against

an unarmed black man as an “isolated incident.” If it is determined that the officer abused his

authority, the officer is typically presented as a “bad apple” and may be suspended, fired, or

arraigned (Kish-Gephart, Harrison, and Treviño 2010; Lersch and Mieczkowski 2005). These

steps may be an appropriate way of responding to individual actors, but they may not be

sufficient to repair the damage done to the black community as a whole. The findings of this

study promote a more sociological view of the issue by suggesting that no act of police violence

is an isolated incident, in both cause and consequence. Episodes of police wrongdoing are one

ramification of a broader trend within the criminal justice system that in recent decades has

promoted aggressive law enforcement and severe sentencing. The effect of this trend on crime

reduction has been underwhelming (Travis, Western, and Redburn 2014), and some policies and

practices have brought about opposite-than-intended effects by making crime worse. In

undermining public safety efforts by dampening citizen crime reporting and inviting legal

cynicism, police violence is a clear example of this.

Individual instances of institutional failure can have aggregate and lasting consequences that

vitiate police-community relations and thwart crime-reduction efforts. If acts of excessive police

force result in community-level consequences—including the non-reporting of crime, which by

undermining public safety efforts likely contributes to pessimism about law enforcement’s

effectiveness—then police departments may consider implementing community-level

interventions in the aftermath of such acts.

In recent years, city politicians and law enforcement agencies have devoted significant attention

and resources to what has come to be known as the “stop snitching” culture. In disadvantaged

neighborhoods, noncooperation with the police is often attributed to street culture or fear of

retaliation (Brown 2007; Natapoff 2009). In some cases that fear is justified, but one implication

of this study is that should we wish to understand why some crime goes unreported in the black

community, we should try to understand residents’ collective memory regarding police violence

in their city and others.

APPENDIX

Table A1. Model Fit Statistics for Frank Jude Analyses: All Neighborhoods

Bayesian Information Criterion

Model

Total Calls

Violent Calls

Linear Pre- and Post-Jude

208,328.2

117,140.4

Linear Pre-Jude and Linear + Quadratic Post-Jude

208,325.2

117,138.9

Linear + Quadratic Pre-Jude and Linear Post-Jude

208,337.9

117,144.8

Linear + Quadratic Pre-Jude and Linear + Quadratic Post-Jude

208,331.3

117,137.0

Acknowledgments

We thank Javier Auyero, Monica Bell, Lawrence Bobo, John Diedrich, Felix Elwert, Peter Hart-

Brinson, Ryan King, Nikki Jones, Christopher Muller, Devah Pager, and Christopher Wildeman

for helpful comments; Jeffrey Blossom for geocoding expertise; Deborah De Laurell and

Elizabeth Roberto for research assistance; and the Milwaukee Police Department for supplying

data. Earlier drafts of this paper were presented at the Workshop on Violence at the Urban

Margins, University of Texas at Austin, April 4, 2013, and the American Sociological

Association Annual Meeting, New York City, August 12, 2013.

Funding

This research was supported by Harvard University.

Notes

1. The MPD redacted addresses from 2.7 percent of 911 calls having to do with crimes against

children and suicides. We dropped these calls, along with calls having to do with commerce-

related crime (e.g., shoplifting, forgery).

2. The City of Milwaukee implemented a new 911 reporting system in early 2004. Because

recorded calls during the implementation period are potentially undercounted, we excluded calls

placed during the first two months of 2004.

3. Supplemental models (available upon request) examined 911 calling patterns in Jude’s

neighborhood and adjacent neighborhoods directly after the beating, finding no significant

effects on call patterns. At the time of his beating, Jude lived in a majority-white neighborhood

on the far south side of the city.

4. We obtained weekly rainfall and snowfall data from the National Oceanic and Atmospheric

Administration.

5. Following the procedure used to estimate the effect of Frank Jude’s beating on citizen crime

reporting, the observation periods for our analyses of these additional cases were centered on

the date the news reported on them. Thus the observation period in which we observe crime and

911 calls runs from January 1, 2006 to December 31, 2008 for the Simpson case; from January

1, 2005 to December 31, 2007 for the Bell case; and from January 1, 2008 to December 31,

2010 for the Grant case.

6. For an extended discussion about the interaction between major events and entrenched beliefs

and prejudices, see Legewie (2015).

Newspaper and Court Documents

Associated Press. 2009. “Police Crack Down after Train Shooting Protests, Businesses Smashed,

Cars Set Blaze in Oakland after Transit Cop Killed Man.” January 8.

Barton, Gina. 2006. “2 Plead Guilty in Federal Jude Case.” Milwaukee Journal Sentinel,

November 4.

Board of Fire and Police Commissioners of the City of Milwaukee. 2008. In the Matter of the

Appeal of Police Officer Zoe Jackson from MPD Personnel Order Number 2007-184, March 5.

Borowski, Greg. 2007. “2nd Jude Trail Stands as a Rarity.” Milwaukee Journal Sentinel, July 28.

Buckley, Cara, and William Rashbaum. 2006. “A Day after a Fatal Shooting, Questions,

Mourning, and Protest.” New York Times, November 27.

Chideya, Farai. 2009. “Hundreds Protest in Oakland over BART Shooting.” National Public

Radio, January 8.

CNN. 2007. “Police Fire 50 Rounds, Kill Groom on Day of Wedding.” November 26.

Diedrich, John. 2005a. “8 More Fired in Jude Case.” Milwaukee Journal Sentinel, May 24.

Diedrich, John. 2005b. “Police Suspected in Man’s Beating.” Milwaukee Journal Sentinel,

February 6.

Diedrich, John. 2007a. “3 Ex-Officers Guilty.” Milwaukee Journal Sentinel, July 27.

Diedrich, John. 2007b. “Force Blurs the Line.” Milwaukee Journal Sentinel, February 18.

Diedrich, John. 2012. “City Reaches Tentative Settlement in Jude Lawsuit.” Milwaukee Journal

Sentinel, January 20.

Fransisco, Stacey, Terry Frieden, and Ellen Rose. 2006. “N.Y. Mayor Meets with Dead

Groom’s Family.” CNN, November 29.

McKinley, Jesse. 2009. “Oakland Turns Violent Over Shooting.” New York Times, January 7.

McLaughlin, Eliott, Augie Martin, Dan Simon, and Mallory Simon. 2009. “Hundreds Demand

Answers, Action in Subway Shooting.” CNN, January 8.

Milwaukee Journal Sentinel. 2007. “Timeline of Events in Frank Jude Beating Case.” July 27.

Milwaukee Journal Sentinel. 2012. “Timeline: Deaths in Milwaukee Police Custody, Proposed

Reforms.” October 27.

Nunnally, Derrick. 2005. “100 Protest Progress of Police Beating Case.” Milwaukee Journal

Sentinel, February 11.

Stingl, Jim. 2006. “Jude Case Plays out Like a Tragedy.” Milwaukee Journal Sentinel, April 16.

USA v. Bartlett, Spengler, and Masarik. 2009. Appeals from the United States District Court for

the Eastern District of Wisconsin. No. 06-CR-273, June 8.

Wilson, Michael. 2006. “50 Shots Fired, and the Experts Offer a Theory.” New York Times,

November 27.

References

Anderson, Elijah. 1999. Code of the Street: Decency, Violence, and the Moral Life of the Inner

City. New York City: Norton.

Andrews, Kenneth, and Michael Biggs. 2006. “The Dynamics of Protest Diffusion: Movement

Organizations, Social Networks, and News Media in the 1960 Sit-Ins.” American Sociological

Review 71(5):752–77.

Avakame, Edem, James Fyfe, and Candace McCoy. 1999. “‘Did You Call the Police? What Did

They Do?’ An Empirical Assessment of Black’s Theory of Mobilization of Law.” Justice

Quarterly 16(4):765–92.

Banton, Michael. 1964. The Policeman in the Community. New York: Basic Books.

Baumer, Eric. 2002. “Neighborhood Disadvantage and Police Notification by Victims of

Violence.” Criminology 40(3):579–616.

Black, Donald. 1970. “The Production of Crime Rates.” American Sociological Review

35(4):733–48.

Bobo, Lawrence, and Victor Thompson. 2006. “Unfair by Design: The War on Drugs, Race,

and the Legitimacy of the Criminal Justice System.” Social Research 73(2):445–72.

Brown, Ethan. 2007. Snitch: Informants, Cooperators, and the Corruption of Justice. New

York: PublicAffairs.

Brunson, Rod. 2007. “‘Police Don’t Like Black People’: African-American Young Men’s

Accumulated Police Experiences.” Criminology and Public Policy 6(1):71–102

Brunson, Rod, and Jody Miller. 2006. “Young Black Men and Urban Policing in the United

States.” British Journal of Criminology 46(4):613–40.

Carr, Patrick, Laura Napolitano, and Jessica Keating. 2007. “We Never Call the Cops and Here

Is Why: A Qualitative Examination of Legal Cynicism in Three Philadelphia Neighborhoods.”

Criminology 45(2):445–80.

Desmond, Matthew, Carl Gershenson, and Barbara Kiviat. 2015. “Forced Relocation and

Residential Instability among Urban Renters.” Social Service Review 89(2):227–62.

Holmes, Malcolm D., and Brad W. Smith. 2008. Race and Police Brutality: Roots of an Urban

Dilemma. Albany: SUNY Press.

Jefferis, Eric, Robert Kaminski, Stephen Holmes, and Dena Hanley. 1997. “The Effect of a

Videotaped Arrest on Public Perceptions of Police Use of Force.” Journal of Criminal Justice

25(5):381–95.

King, Ryan D., and Gretchen M. Sutton. 2013. “High Times for Hate Crimes: Explaining the

Temporal Clustering of Hate-Motivated Offending.” Criminology 51(4):871–94.

Kirk, David, and Mauri Matsuda. 2011. “Legal Cynicism, Collective Efficacy, and the Ecology

of Arrest.” Criminology 49(2):443–72.

Kirk, David, and Andrew Papachristos. 2011. “Cultural Mechanisms and the Persistence of

Neighborhood Violence.” American Journal of Sociology 116(4):1190–1233.

Kirk, David, Andrew Papachristos, Jeffrey Fagan, and Tom R. Tyler. 2012. “The Paradox of

Law Enforcement in Immigrant Communities: Does Tough Immigration Enforcement

Undermine Public Safety?” Annals of the American Academy of Political and Social Science

641(1):79–98.

Kish-Gephart, Jennifer, David Harrison, and Linda Klebe Treviño. 2010. “Bad Apples, Bad

Cases, and Bad Barrels: Meta-Analytic Evidence about Sources of Unethical Decisions at Work.”

Journal of Applied Psychology 95(1):1–31.

Klinger, David. 1997. “Negotiating Order in Patrol Work: An Ecological Theory of Police

Response to Deviance.” Criminology 35(2):277–306.

Koopmans, Ruud. 2004. “Movements and Media: Selection Processes and Evolutionary

Dynamics in the Public Sphere.” Theory and Society 33(3):367–91.

Land, Kenneth C., Patricia L. McCall, and Lawrence E. Cohen. 1990. “Structural Covariates of

Homicide Rates: Are There Any Invariances across Time and Social Space?” American Journal

of Sociology 95(4):922–63.

Lasley, James. 1994. “The Impact of the Rodney King Incident on Citizen Attitudes toward

Police.” Policing and Society 3(4):245–55.

Legewie, Joscha. 2013. “Terrorist Events and Attitudes toward Immigrants: A Natural

Experiment.” American Journal of Sociology 118(5):1199–1245.

Legewie, Joscha. 2015. “Racial Profiling and Use of Force in Police Stops: How Local Events

Trigger Periods of Increased Discrimination.” Working paper, New York University.

Lersch, Kim Michelle, and Tom Mieczkowski. 2005. “Violent Police Behavior: Past, Present,

and Future Research Directions.” Aggression and Violent Behavior 10(5):552–68.

Manning, Peter. 1988. Symbolic Communication: Signifying Calls and the Police Response.

Cambridge, MA: MIT Press.

Matulia, Kenneth. 1982. A Balance of Forces: National Survey of Police Deadly Force.

Gaithersburg, MD: International Association of Chiefs of Police.

Muller, Christopher. 2012. “Northward Migration and the Rise of Racial Disparity in American

Incarceration, 1880–1950.” American Journal of Sociology 118(2):281–326.

Natapoff, Alexandra. 2009. Snitching: Criminal Informants and the Erosion of American Justice.

New York: New York University Press.

Patterson, Orlando. 1998. Rituals of Blood: Consequences of Slavery in Two American

Centuries. New York: Basic Civitas Books.

Pew Research Center. 2012. “News about Trayvon Martin Cast Still Top Story.” Washington,

DC: Pew Research Center for People and the Press.

Pew Research Center. 2014. “Stark Racial Divisions in Reactions to Ferguson Police Shooting.”

Washington, DC: Pew Research Center for People and the Press.

Reiss, Albert, Jr. 1992. “Police Organization in the Twentieth Century.” Crime and Justice

15:51–97.

Rosenbaum, Dennis, Amie Schuck, Sandra Costello, Darnell Hawkins, and Marianne Ring.

2005. “Attitudes toward the Police: The Effects of Direct and Vicarious Experience.” Police

Quarterly 8(3):343–65.

Rosenfeld, Richard, Bruce Jacobs, and Richard Wright. 2003. “Snitching and the Code of the

Street.” British Journal of Criminology 43(2):291–309.

Ross, Jeffrey Ian. 2000. Making News of Police Violence: A Comparative Study of Toronto and

New York. Westport, CT: Praeger.

Sampson, Robert. 2012. Great American City: Chicago and the Enduring Neighborhood Effect.

Chicago: University of Chicago Press.

Sampson, Robert, and Jeglum Bartusch. 1998. “Legal Cynicism and (Subcultural?) Tolerance of

Deviance: The Neighborhood Context of Racial Differences.” Law and Society Review

32(4):777–804.

Sánchez Jankowski, Martín. 1991. Islands in the Street: Gangs and American Urban Society.

Berkeley and Los Angeles: University of California Press.

Scheufele, Dietram. 1999. “Framing as a Theory of Media Effects.” Journal of Communication

49(1):103–122.

Shadish, William, Thomas Cook, and Donald Campbell. 2002. Experimental and Quasi-

Experimental Designs for Generalized Causal Inference. Independence, KY: Wadsworth

Cengage Learning.

Singer, Judith, and John Willett. 2003. Applied Longitudinal Data Analysis: Modeling Change

and Event Occurrence. New York: Oxford University Press.

Skogan, Wesley. 2006. “Asymmetry in the Impact of Encounters with the Police.” Policing and

Society 16(2):99–126.

Terrill, William, and Michael Reisig. 2003. “Neighborhood Context and Police Use of Force.”

Journal of Research on Crime and Delinquency 40(3):291–321.

Travis, Jeremy, Bruce Western, and Steve Redburn, eds. 2014. The Growth of Incarceration in

the United States: Exploring Causes and Consequences. Washington, DC: National Academies

Press.

Tyler, Tom R., and Cheryl Waslak. 2004. “Profiling and Police Legitimacy: Procedural Justice,

Attributions of Motive, and Acceptance of Police Authority.” Criminology 42(2):253–81.

Venkatesh, Sudhir. 2000. American Project: The Rise and Fall of a Modern Ghetto. Cambridge,

MA: Harvard University Press.

Weitzer, Ronald. 2002. “Incidents of Police Misconduct and Public Opinion.” Journal of

Criminal Justice 30(5):397–408.

Weitzer, Ronald, and Steven Tuch. 2005. “Racially Biased Policing: Determinants of Citizen

Perceptions.” Social Forces 83(3):1009–30.

Worden, Robert. 1996. “The Causes of Police Brutality: Theory and Evidence on Police Use of

Force.” Pp. 23–51 in Police Violence: Understanding and Controlling Police Abuse of Force,

edited by W. Geller and H. Toch. New Haven, CT: Yale University Press.

Matthew Desmond is the John L. Loeb Associate Professor of the Social Sciences and the Co-

Director of the Justice and Poverty Project at Harvard University. A sociologist of urban

poverty, racial inequality, housing insecurity, and public policy, his latest book is Evicted:

Poverty and Profit in the American City.

Andrew V. Papachristos is an Associate Professor in the Department of Sociology at Yale

University. His research examines neighborhood social organization, street gangs, interpersonal

violence, illegal gun markets, and social networks. He is currently completing a multicity study

on the diffusion of gun violence within high-risk social networks.

David S. Kirk is Associate Professor in the Department of Sociology and a Professorial Fellow

of Nuffield College at the University of Oxford. For more than a decade he has been examining

the consequences of the forced migration from Hurricane Katrina on the lives of former

Louisiana prisoners. One extension to this work involves an experimental housing mobility

program for ex-prisoners. Kirk’s recent research has appeared in the Proceedings of the

National Academy of Sciences, Criminology, Social Problems, and American Journal of

Epidemiology.

Table1. Summary Statistics for Population, Call, and Crime Variables

Mean

Min.

Max.

Neighborhood Structural Variables

Percent Black

.392

.370

.990

Percent Latino

.113

.183

.828

Percent Poverty

.203

.154

.817

Percent Renter

.533

.225

Weekly Calls for Service

Total

3.367

3.565

Violence

.898

1.437

Weekly Reported Crimes

Total Crimes

1.823

1.950

Violent Crimes

.403

.753

Note: Summary data for 596 Milwaukee block groups. Here, crime and 911 call statistics pertain to the

observation period before and after the newspaper broke the story of Frank Jude’s beating: March 1, 2004 to

December 31, 2005.

Table 2. Negative Binomial Regressions with Neighborhood and Month Fixed Effects Estimating the Effect of Frank Jude’s Beating on Citizen Crime Reporting: All, Black,

and White Neighborhoods

All Neighborhoods

Black Neighborhoods

White Neighborhoods

Total Calls

Violent Calls

Total Calls

Violent Calls

Total Calls

Violent Calls

Intercept

–6.378***

–7.876***

–6.436***

–7.594***

–6.679***

–8.574***

(.087)

(.140)

(.418)

(.538)

(.173)

(.348)

Weeks Pre-Jude Story

.036***

.019

.031**

–.007

.044*

.037

(.008)

(.015)

(.011)

(.020)

(.022)

(.051)

Jude Story

–.009

–.021

–.005

.036

.114

.005

(.034)

(.065)

(.048)

(.089)

(.095)

(.224)

Weeks Post-Jude Story

–.088***

–.117**

–.069*

–.019

–.192**

–.281*

(.021)

(.040)

(.030)

(.053)

(.061)

(.137)

Weeks Post-Jude Story (squared)

.002***

.003***

.001

.000

.004***

.008**

(.000)

(.001)

(.003)

Reported Crime

.039***

.202***

.037***

.171***

.036***

.465***

(.002)

(.005)

(.002)

(.006)

(.004)

(.023)

Percent Poverty

.999***

1.338***

1.087**

2.000***

1.733

2.736

(.224)

(.299)

(.330)

(.422)

(1.113)

(1.706)

Percent Renter

.314*

.547**

–.166

–.304

.550

1.107**

(.150)

(.209)

(.294)

(.367)

(.295)

(.424)

Percent Black

1.030***

1.353***

1.330**

1.511**

2.369**

1.797

(.092)

(.130)

(.437)

(.549)

(.857)

(1.122)

Percent Latino

.634***

.919***

2.775

1.159

.907

.685

(.158)

(.219)

(1.502)

(1.694)

(1.050)

(1.520)

Note: These models focus on the observation period of March 1, 2004 to December 31, 2005. The reported crime variable represents all reported weekly crimes when the

outcome is total police-related 911 calls and violent crimes when the outcome is calls reporting violence. Black/white neighborhoods are defined as those in which at least

65 percent of residents are black/white. Coefficients and standard errors (in parentheses) for the week measures of time were multiplied by 10 for ease of interpretation. N

= 596 neighborhoods.

*p < .05; **p < .01; ***p < .001 (two-tailed test).

Table 3. Nonequivalent Dependent Variable Models Estimating the Effect of

Frank Jude’s Beating on Calls for Automobile Accidents

Accident Calls

Intercept

–8.371***

(.287)

Weeks Pre-Jude Story

–.004

(.003)

Jude Story

–.128

(.165)

Weeks Post-Jude Story

.016***

(.003)

Percent Poverty

.315

(.597)

Percent Renter

1.116**

(.387)

Percent Black

–.303

(.229)

Percent Latino

–.663

(.358)

Weekly Rainfall

.031

(.023)

Weekly Snowfall

.051***

(.012)

Note: N = 596 neighborhoods. Standard errors are in parentheses.

*p < .05; **p < .01; ***p < .001 (two-tailed test).

Table 4. Negative Binomial Regressions with Neighborhood and Month Fixed Effects Estimating the Effect of the Danyall Simpson, Sean Bell, and Oscar Grant Stories on Citizen Crime

Reporting: All and Black Neighborhoods

Danyall Simpson

Sean Bell

Oscar Grant

Total Calls

Violent Calls

Total Calls

Violent Calls

Total Calls

Violent Calls

All Neighborhoods

Intercept

–4.116***

–5.333***

–6.386***

–7.737***

–12.532***

–11.451***

(,311)

(.579)

(.070)

(.115)

(1.141)

(2.106)

Weeks Pre-Story

–.221***

–.293**

.093***

.061***

.629***

.344

(.047)

(.089)

(.008)

(.015)

(.098)

(.181)

Weeks Pre-Story (squared)

.001***

.001**

–.000***

–.000

–.001***

–.001

(.000)

Story

–.032*

–.041

–.090***

–.152***

.167***

.022

(.015)

(.028)

(.016)

(.030)

(.015)

(.028)

Weeks Post-Story

–.014

–.062**

.014

.009

.061***

.047*

(.011)

(.020)

(.011)

(.021)

(.012)

(.022)

Weeks Post-Story (squared)

–.001***

.000

–.001

.002***

.001

(.000)

Reported Crime

.053***

.231***

.049***

.229***

.056***

.244***

(.001)

(.003)

(.001)

(.003)

(.001)

(.003)

Percent Poverty

.294

.208

.596**

.443

–.003

.041

(.176)

(.214)

(.179)

(.239)

(.161)

(.197)

Percent Renter

.348**

.835***

.348**

.745***

.150

.853***

(.116)

(.154)

(.118)

(.172)

(.103)

(.135)

Percent Black

.892***

1.192***

.965***

1.236***

.939***

1.328***

(.075)

(.095)

(.076)

(.106)

(.069)

(.087)

Percent Latino

.402**

.690***

.478***

.718***

.391**

.796***

(.127)

(.163)

(.128)

(.180)

(.113)

(.148)

Black Neighborhoods

Intercept

–5.980***

–5.685***

–7.012***

–7.587***

–12.874***

–17.424***

(.535)

(.867)

(.327)

(.440)

(1.700)

(2.932)

Weeks Pre-Story

–.098

–.247*

.080***

.070***

.634*

.863**

(.068)

(.121)

(.011)

(.020)

(.144)

(.025)

Weeks Pre-Story (squared)

.000

.001

–.000***

–.000

–.001***

–.002***

(.000)

(.001)

Story

–.029

–.034

–.056*

–.108**

.201

.074

(.022)

(.039)

(.023)

(.041)

(.023)

(.039)

Weeks Post-Story

.031*

–.032

–.029

–.022

.070**

.108***

(.016)

(.028)

(.017)

(.029)

(.018)

(.031)

Weeks Post-Story (squared)

–.001**

–.001*

.001**

.000

.002***

(.000)

(.001)

Reported Crime

.049***

.185***

.045***

.186***

.055***

.196***

(.001)

(.004)

(.001)

(.004)

(.002)

(.004)

Percent Poverty

.346

.111

.798**

.687*

.122

.417

(.266)

(.315)

(.273)

(.346)

(.242)

(.287)

Percent Renter

–.176

.467

–.036

.351

.023

.309

(.229)

(.293)

(.239)

(.323)

(.214)

(.260)

Percent Black

2.477***

1.717***

1.895***

1.276**

1.374***

1.511***

(.335)

(.421)

(.353)

(.468)

(.320)

(.379)

Percent Latino

2.323*

.049

2.840*

.390

1.411

1.729

(1.045)

(1.220)

(1.425)

(.945)

(1.148)

Note: The reported crime variable represents all reported weekly crimes when the outcome is total police-related 911 calls and violent crimes when the outcome is calls reporting violence.

Black neighborhoods are defined as those in which at least 65 percent of residents are black. Coefficients and standard errors (in parentheses) for the week measures of time were

multiplied by 10 for ease of interpretation. N = 596 neighborhoods.

*p < .05; **p < .01; ***p < .001 (two-tailed test).

(A) (B)

Figure 1. Milwaukee Maps Showing (A) the Percent of Neighborhood Residents Who Are Black and (B) the Total Number of

Police-Related 911 Calls: March 2004 to December 2005

Note: Darker shading indicates a larger percentage of black residents or a higher 911 calling rate.

Figure 2. Estimated Number of Weekly 911 Calls Reporting Crime Before and After the Frank Jude Story

-52 -48 -44 -40 -36 -32 -28 -24 -20 -16 -12 -8 -4 0 4 8 12 16 20 24 28 32 36 40 44 48 52

Estimated 911 Calls per Week per Block Group

Weeks since Frank Jude Story Broke

Estimated Calls Counterfactual

Figure 3. Estimated Number of Weekly 911 Calls Reporting Crime Before and After the Frank Jude Story in White

and Black Neighborhoods

-52 -48 -44 -40 -36 -32 -28 -24 -20 -16 -12 -8 -4 0 4 8 12 16 20 24 28 32 36 40 44 48 52

Estimated 911 Calls per Week per Block Group

Weeks since Frank Jude Story Broke

Estimated Calls, Black Nhoods Counterfactual, Black Nhoods

Estimated Calls, White Nhoods Counterfactual, White Nhoods

Figure 4. Homicides in Milwaukee, March through August, 2004 to 2010

100

2004 2005 2006 2007 2008 2009 2010