Life in a Crime Scene: Stop, Question, and Frisk Activity in New York City Neighborhoods in the Aftermath of Homicides

An incident of extreme violence, such as a homicide, disrupts daily life not only through the incident itself but also through the chaos and disruption that emerge in the aftermath of violence. This article presents descriptive evidence about how communities are affected by increased police activity—specifically, stop, question, and frisk (SQF) activity—following an incident of extreme violence. Our results show that SQF activity in a block group increases in the week following a homicide in New York City, with the largest increases in neighborhoods with high crime rates. Furthermore, neighborhoods with different racial and ethnic compositions have differential levels of average SQF activity and also experience differential responses from the police in the aftermath of a homicide. African American residents have a higher probability of being stopped following a homicide than do nonblack residents across neighborhoods of all types.

A Large body of evidence demonstrates how incidents of extreme violence rever- berate through neighborhoods, affecting not only the victims and perpetrators but entire communities.The shock of violence has an acute effect on children's functioning and academic performance, and the accumulation of violence generates widespread fear and dissatisfaction among residents, disrupting collective life in public spaces and undermining the stability of communities (Sharkey and Sampson  2015; Skogan 1990).
An implicit assumption in the literature on community violence is that the fear, stress, and trauma associated with direct or indirect exposure to incidents of violence account for the relationship between community violence and children's development.This assumption ignores the fact that incidents of extreme violent crime, such as homicides, do not end with the conclusion of the violent interaction between perpetrator and victim.Instead, an incident of extreme violence involves the act of violence, the chaos that may arise in its aftermath, and the ensuing response from authorities, most notably the police.Living in an intensely violent environment means spending a substantial portion of one's time in or near a crime scene.
In this study, we provide a descriptive analysis of how incidents of extreme violence bring about a change in the residential environment through elevated interactions between police and residents.Instead of focusing exclusively on the prevalence and impact of violence, this article shifts attention toward the aftermath of extreme violence by focusing on stop, question, and frisk (SQF) activity that occurs after homicides in New York City (NYC).
Our analysis is designed to shed light on one additional way in which residential environments are transformed by an act of extreme violence, but it also contributes to existing empirical literature on the use of SQF by police in NYC.SQF activity has garnered increasing attention, particularly in NYC, because of its controversial role in policing strategy and the growing prevalence of this tactic in the 2000s.Official data from the New York Police Department (NYPD) show that from 2002 to 2011, the number of reported stops by NYPD officers rose from fewer than 100,000 to a peak of more than 685,000 (New York Civil Liberties Union 2015).From 2011 to 2014, conversely, this trend reversed, and the number of stops has declined rapidly, dropping to just more than 46,000 stops in 2014.
The remarkable fluctuations in SQF activity take on added importance given the unbalanced use of this tactic in communities across the city.Research focusing on the peak years of SQF activity in NYC demonstrates that police stops of residents were disproportionately focused on nonwhite populations and communities, leading to high-profile court cases and politicized debates surrounding the question of whether the strategy has been used as a form of racial profiling (Gelman, Fagan, and Kiss  2007).One of the core arguments made by former representatives of the city and its police department was that the disproportionate prevalence of SQF activity in nonwhite communities could be explained by the disproportionate amount of crime occurring in these communities.This argument has been assessed empirically with analyses of SQF activity in different geographical areas after conditioning on levels of crime (Gelman et al. 2007).In this article, we go one step further and isolate specific incidents of extreme violence-homicides-to provide evidence on how SQF activity changes, for different communities, in the aftermath of violence.
The central empirical goal of this article is to describe the prevalence of SQF activity in different types of communities, characterized by racial and ethnic composition and crime prevalence, in the period before and after a homicide.Exploiting the timing of specific incidents of extreme violence allows us to focus attention on police activity in response to a specific event and to use that event to understand how an act of violence can alter social dynamics within a neighborhood, disrupting everyday life and raising the probability of interactions between residents and law enforcement.Our results show that the impact of violence on police-resident interactions varies across communities and that variation is structured by the racial and ethnic composition of the neighborhood.Just as the burden of violence is unevenly distributed across the neighborhoods of NYC, the change in the activity of law enforcement that arises in the aftermath of violence is acute in some communities and undetectable in others.

Violence, Policing, and Urban Communities
The collateral consequences of violence are visible in an extensive literature on the impact of exposure to violence on children's functioning, behavior, and developmental outcomes.Direct exposure to violence is associated with posttraumatic stress disorder, depression, anxiety, symptoms of asthma, and reduced physical activity (Buka et al. 2001; Buckner, Beardslee, and Bassuk 2004; Cooley-Quille et al.  2001; Fowler et al. 2009; Molnar et al. 2004; White et al. 1998; Wright et al. 2004).
sociological science | www.sociologicalscience.comMore recent literature exploiting the timing of incidents of violence has found that in the aftermath of local violence, children perform worse on assessments of academic skills, cognitive skills, and measures of executive function (McCoy,  Raver, and Sharkey 2015; Sharkey 2010; Sharkey et al. 2012; Sharkey et al. 2014).Longer-term exposure to violence and fear of violence affect children's academic performance (Lacoe 2013), their attitudes toward school (Farrell and Sullivan 2004), and their likelihood of high school graduation (Harding 2009).Beyond the school day, neighborhood violence is also associated with a greater likelihood of teen pregnancy (Harding 2009), drug use (Kilpatrick et al. 2000), and aggressive or criminal behavior (Farrell and Sullivan 2004; Patchin et al. 2006; Preski and Shelton 2001).
Although this empirical literature provides strong evidence that exposure to more violent residential areas has detrimental consequences for young people, very little focus has been given to the mechanisms underlying this relationship.The implicit assumption is that violence itself influences children.Yet there is less focus in the literature on how violence alters the residential environment by generating a response from law enforcement and how interactions with police affect young people.
On this issue, suggestive evidence from a recent study finds that young people who are stopped by police or arrested report higher levels of future delinquency (Wiley, Slocum, and Esbensen 2013).Contact with law enforcement is also associated with residents' perceptions of the legitimacy of the police and the safety of their neighborhoods.Research has found some evidence that perceived racial and ethnic disparities in stops compromise residents' views of the legitimacy of the police (Fagan and Tyler 2005; Gau and Brunson 2010), whereas other work finds no effect of stop-and-frisk-type policing on negative perceptions of the police (Weisburd, Telep,  and Lawton 2011).No matter how the police are perceived within a community, the presence of police activity provides a signal to residents about safety and the problem of crime.As an example, Drakulich (2013) finds that perceptions of disorder and police activity may contribute more to residents' views of neighborhood danger than actual crime rates, suggesting that police activity may have an impact on community life that is distinct from the role of violence itself.

Stop, Question, and Frisk Activity in New York City
Beyond the more abstract focus on the multidimensional consequences of violence is the more straightforward question of which communities bear the burden of intensive policing.In NYC, the debate about SQF and similar policing strategies often centers on the differences in the rate at which black, Hispanic, and white residents are stopped.Existing research provides some evidence of racial disparities in police stops in NYC (Gelman et al. 2007).Higher rates of SQF of people of color do not necessarily reflect higher rates of offending: one study finds that despite large overall differences in stops by race and ethnicity in NYC, the contraband recovery rate for frisked or searched individuals is higher for whites than it is for blacks after adjusting for age, location, crime type, and other factors (Ridgeway  2007).Former mayor Bloomberg and former representatives from the NYPD argue that SQF activity is an essential tool to reducing crime and that patterns of stops are driven by spatial and temporal crime patterns and the probability of individuals being involved in criminal activity, not by race or ethnicity alone (Editorial Board 2013).In fact, Weisburd et al. (2013) find that SQF activity is concentrated in crime "hot spots" on street segments and intersections throughout the city.Using this argument as a starting point, our focus in this article is the escalation of SQF activity in the aftermath of extreme violence, as measured by data on homicides.Regardless of one's perspective on the controversies surrounding the practice, there is little dispute that SQF activity is at least partially a response to criminal activity, and the evidence produced on the topic demonstrates that stops are more common in areas with high levels of crime.However, the question remains as to how much of the volume of SQF activity following a crime is an appropriate response to the crime and how much is driven by other factors, such as neighborhood demographics.

Data and Sample
We estimate the relationship between local homicides and SQF activity using detailed data from the NYPD on all individual police stops from 2006 through 2009.All NYPD offers are trained to conduct SQFs and to fill out a report documenting the interaction (Ridgeway 2007). 1 Data compiled from the NYPD documentation of  1). 2 During this time, more than three-quarters of documented stops have a reason given for the stop.The most common reason is "furtive movements," followed by "other" and "suspect acting as a lookout."However, more than onethird of the reasons given for stops do not relate to characteristics of individuals or their behavior.For instance, almost 20 percent of stops are conducted because the time of day "fits high crime incidence," and over 14 percent of stops occur because the individual is in a "high-crime area." Of the total number of individuals stopped between 2006 and 2009, more than half were frisked.Very few stops resulted in further action: only 6.4 percent of all stopped individuals received a summons, 5.5 percent were arrested, and only 1.7 percent were found to be carrying contraband (such as weapons).More than half of the people stopped during this time period were black, almost one-quarter were Hispanic, and 10 percent were white (Table 1).When they were stopped, more than half of blacks, Hispanics, and individuals coded as black and Hispanic were frisked, whereas less than 40 percent of whites and 48 percent of Asians were frisked.
The point-specific SQF data are geocoded to the census block group. 3The homicide data, also from the NYPD, include information about the location and timing of the homicide and the race and ethnicity of the victim(s). 4Previous research has shown that policing outcomes can be correlated with demographic characteristics of neighborhoods (i.e., Stewart et al. 2009).Using block group level data from the 2000 U.S. Census, we include a measure of the neighborhood poverty rate and stratify the sample by neighborhood demographic composition.The constructed data set includes daily counts of police stops for all block groups where a homicide occurred between 2006 and 2009.
To capture underlying differences in crime rates across neighborhoods, we categorize each block group by whether it is in a "high-crime area."This categorization is based on the share of total stops throughout the study period that occur for the stated reason that the person stopped was in a "high-crime area."The binary measure of "high-crime area" takes the value of 1 if more than half of all stops conducted in that neighborhood were due to this reason, and 0 otherwise.It follows that "high-crime areas" using this measure have more stops per day, on average, than neighborhoods that are not designated as high crime (0.69 stops and 0.44 stops, respectively; Table 2).

Research Design
We estimate the change in the number of police stops in block group, b, in the week after a homicide relative to the week before the homicide occurs, using two indicators.The first variable designates whether the observation is in the two-week period before or after the homicide (BefAfter), and the second identifies observations in the week after the homicide alone (AfterHom).In Equation ( 1 (1) we control for the racial and ethnic composition of the neighborhood using a series of indicator variables for majority black, majority Hispanic, majority Asian, and "mixed" (or no group representing over 50 percent of the block group population) and for the block group poverty rate.The reference category is majority white.Because there is secular variation in crime and stops by year, month, and day of the week, all specifications include controls for month (γ m ), day of the week (δ d ), and year (θ y ).Block groups also may differ in the level or intensity of stop-and-frisk activity because they are located in police precincts with different approaches to controlling local crime.Therefore we add precinct fixed effects (ϑ p ) to compare changes in stop-and-frisk activity before and after a homicide among block groups that are within the same precinct. 5Neighborhoods may differ in the level of stops The difference in stops before and after a homicide varies by the demographic characteristics of the block group.On average, majority black neighborhoods have the highest level of police stops before and after a homicide, followed by majority Hispanic neighborhoods and mixed race neighborhoods (Figure 2).The unadjusted differences show that, on average, stops increase in the week following a homicide in majority black, majority Hispanic, and mixed neighborhoods but not in Asian or white neighborhoods. 6To test for differential changes in stops by neighborhood composition, we estimate a model that includes interactions with the measures of block group racial and ethnic composition. 7The coefficients of interest are on the interactions between neighborhood composition and the "after homicide" indicator.We extend the models to test for differential patterns of stops by race and ethnicity of the person stopped.

Results
Table 3 presents the baseline specification comparing the number of police stops in the week following a homicide to the number of stops in the same block group in the prior week, controlling for neighborhood composition, day of the week, month, and year.On average, there is an increase in number of 0.33 police stops per day in the week following a homicide (column 1).The second specification controls  for the block group poverty rate and shows no change in the magnitude or sign of the coefficient on the main variable of interest.Block groups may differ in the level or intensity of police stops because they are located in police precincts with different strategies, policies, or approaches to controlling local crime.Therefore we add precinct fixed effects to compare changes in SQF activity before and after a homicide among block groups that are in the same precinct.The addition of the precinct fixed effects does not change the magnitude of the estimate (column 3).In all three of these specifications, the results suggest that there are roughly 0.33 more stops in the week after a homicide than in the week prior, a highly significant and substantively large change in the prevalence of police stops.The estimate represents a 70 percent increase in stops following a homicide, relative to the mean number of stops per day across all block groups in the sample (see Table 2 for means).These initial results indicate that police SQF activity intensifies in communities after a homicide occurs, as would be expected with such a serious form of violent crime. 8he second set of analyses considers how the police response to homicides varies across different types of neighborhoods.The models presented in Table 4 stratify the sample of block groups using the "high-crime area" designation. 9Results in columns 1 and 2 show that in the aftermath of a homicide, stops increase by a larger margin in high-crime areas (0.68 stops, or a 99 percent increase from the mean) than in other areas (0.28 stops, or a 63 percent increase).Columns 3 and 4 examine SQF activity in block groups with different racial and ethnic compositions.Focusing first on high-crime areas, we find that in block groups that are majority white (the reference group) and in mixed block groups, there is no significant difference between the number of stops in the week before or after a homicide.In majority black and majority Hispanic block groups, there is a substantially higher number of stops in the period before a homicide, and the increase in stops in the week after a homicide is much greater than in all other block groups.In other words, the level of SQF activity is always higher in high-crime neighborhoods where black Americans or Hispanics comprise the majority group, and the intensification of SQF activity in the aftermath of a homicide is amplified as well.In low-crime areas, the level of SQF activity rises in a similar way in all block groups, except those that are majority white.Homicide does not lead to elevated levels of SQF activity in white block groups but leads to roughly 0.23 to 0.31 more stops in block groups where nonwhites are the majority group. 10ncreased police activity following a homicide may have implications for all neighborhood residents.The main results show that SQF activity increases following a homicide and that increases are larger in areas already labeled as "high crime" as well as in areas that are majority black or Hispanic.To further understand how violence affects neighborhood residents, we explore whether a local homicide affects residents differently based on race or ethnicity (Figure 3).Specifically, we ask, Are black residents more likely to be stopped than nonblack residents following a homicide in the neighborhood?When we investigate the results by the race of the person stopped, we find that following a homicide, the number of stops of black residents increases by a larger margin than the number of stops of residents who are not black. 11The size of the change varies by neighborhood race and ethnicity.The number of stops of black residents increases in almost all neighborhoods following a homicide. 12In majority black neighborhoods, the number of stops of black residents increases by 60 percent following a homicide, but there is no change in the number of stops of nonblack residents.In majority Hispanic neighborhoods, stops of all residents increase, but stops of black residents increase more, on average.Increases in the number of stops of nonblack residents only occur following homicides in majority white and majority Hispanic neighborhoods.
When the underlying crime rate in a neighborhood is higher, do we also see a differential relationship for black residents?Following a homicide, the largest increase in the number of stops is of black residents in high-crime areas (Figure 3). 13he change in the number of stops of black residents is equivalent to a 77 percent increase in the mean number of stops in high-crime areas, compared to stops of nonblack residents (22 percent).In lower-crime areas, the increase in the number of black stops is 44 percent above the mean for low-crime areas, whereas nonblack stops only increase by 19 percent.
Increases in the number of stops of black residents are concentrated in majority black and majority Hispanic neighborhoods. 14In fact, the number of stops of black residents increases everywhere, with the exception of in high-crime mixed race/ethnicity neighborhoods.In high-crime majority black neighborhoods, the number of stops of black residents increases, whereas the number of stops of nonblack residents does not change.In high-crime majority Hispanic neighborhoods,  See Tables 3 and 4, and Online Supplemental Tables B and C for full results.

Discussion
Research on the consequences of living within violent residential settings has focused entirely on the stress, fear, and trauma generated by exposure to actual incidents of violence.This focus is appropriate, considering the substantial evidence that has been generated on how young people interpret, cope with, and respond to acts of extreme violence (Anderson 1999; Harding 2011; Kotlowitz 1991;  Pynoos et al. 1987; Osofsky 1995).In this article, we expand the focus on community violence to consider institutional responses to extreme violence, as measured by police interactions with local residents.We argue that incidents of extreme violence become salient in the lives of residents, and young people in particular, not only because they generate fear, sadness, shock, and anger but also because they transform the individual's residential environment into a crime scene as law enforcement descends on the community where the incident occurred.In the aftermath of violent crime, the residential setting is transformed by the chaos of flashing lights, sirens, police tape, and representatives from law enforcement.Young people within the environment become potential suspects, witnesses, and informants, no matter their relationship to the victim or perpetrator of the crime.Our goal in this article is to describe one dimension of the transformation of residential space that occurs in the aftermath of extreme violence by analyzing data on SQF activity in NYC neighborhoods before and after homicides occur.As expected, we find that the number of police stops rises sharply in the aftermath of a homicide.Model results show that there are 0.33 more stops made in the week after a homicide occurs than in the week prior, a highly significant and substantively large increase.Although this increase is meaningful, the magnitude of the parameter estimate suggests that stopping individuals on the street should not be thought of as a primary strategy used by police in the aftermath of extreme violence.Homicides bring about greater police activity and more stops, but there is no indication in the data that the police begin a systematic, large-scale escalation of SQF activity when a homicide occurs.The increase in the volume of police stops following local violence is, however, much larger in neighborhoods that have higher levels of crime.In high-crime block groups, the increase in stops made in the week after a homicide than in the week before is three times the increase in lower-crime block groups.Although an increase in stops may aim to prevent retaliatory violence after a gang-involved homicide, and that may be smart policing policy, it does not change the fact that young people in these neighborhoods are exposed to greater contact with the police. 15he second goal of the article is to assess whether the change in police activity differs in communities with different racial/ethnic compositions.Consistent with prior research, we find that majority black block groups have the highest levels of overall SQF activity after conditioning on the average level of crime within the community and the level of poverty within the block group.Majority Hispanic block groups have the second highest level of SQF activity, followed by racially and ethnically mixed block groups that do not have a single dominant ethnic group, followed by majority Asian block and majority white block groups.
Neighborhoods with different racial and ethnic compositions have differential levels of average homicide and SQF activity, but they also experience differential responses from the police in the aftermath of a homicide. 16In high-crime block groups that are majority Hispanic or majority black, we estimate that there are more stops in the week after a homicide than in the week before.In all other highcrime communities with different racial and ethnic profiles, there is no significant difference in the number of stops that occur in the week before and after a homicide.When we consider only the city's high-crime neighborhoods, the escalation of police stops in the aftermath of homicide is felt only in predominantly Hispanic and black neighborhoods.Among lower-crime neighborhoods, those that are racially mixed experience an increase in stops after a homicide occurs that is similar to the increase in majority black and majority Hispanic neighborhoods, but majority white neighborhoods do not.Importantly, the absence of a change in SQF activity in white neighborhoods is not a function of sparse data or imprecise estimates.Even in the raw data, there is virtually no difference in the number of stops that take place in white neighborhoods before and after a homicide occurs.Furthermore, black and nonblack residents have different probabilities of being stopped following a homicide in their neighborhoods.Across the city, a homicide induces a larger increase in stops of black residents than nonblack residents.These increases are largest in majority black and Hispanic neighborhoods and in neighborhoods that are already designated by the police department to have "high crime."For nonblack city residents, the increase in police activity following a homicide is fairly uniform across high-and lower-crime areas, and nonblack stops only increase in number significantly in majority white and majority Hispanic neighborhoods.

Limitations
This study has some important limitations.First, it is a descriptive account of how patterns of SQF change following a homicide.Owing to the endogenous relationship between policing and crime, we are not able to make any causal claims about the impact of homicides on SQF activity.Furthermore, given the evidence of disproportionate minority contact with the justice system, there may be unobserved characteristics of neighborhoods that are correlated with the racial and ethnic compositions of the neighborhoods.These unobserved characteristics may play a part in explaining the relationship between homicides and SQF activity, yet they are not included in our models.Instead, this study aims to describe how patterns of SQF activity change across different neighborhoods, and how neighborhood residents are affected.
Second, although characteristics of a neighborhood, such as the racial and ethnic composition, are related to exposure to violence, this relationship is dynamic (i.e., Sharkey and Sampson 2010).For instance, black youths are more likely to report discrimination by police in predominately white neighborhoods, particularly if the neighborhood has experienced recent increases in the black population (Stewart  et al. 2009).How a neighborhood is changing over time will likely affect the probability of a violent crime occurring, and also the policing strategy for the area.In this article, we account for the dynamic nature of policing-through daily measures of SQF activity-but we do not explore whether these policing patterns shift as the neighborhood racial and ethnic composition changes.Given the short two-week window of time in which we observe the neighborhoods before and after a homicide, we expect the population demographics to be stable.
A related issue pertains to whether crime levels in the neighborhood are on a trajectory that we do not observe.For instance, if crime is increasing in the neighborhood, policing activity may be growing over time.Alternatively, if crime in the neighborhood is decreasing, police activity may also be decreasing.In either scenario, the police response to changing levels of crime may lag actual crime levels, if neighborhoods labeled at one period of time as "high crime," and subsequently assigned a heightened policing strategy, are not reevaluated with regularity.In our data, we only observe whether the justification for a stop is that the person was in a "high-crime area"; we do not have the underlying information on which this designation was made.Though this measure of a "high-crime area" reflects the police perception of the neighborhood, and may not align with actual crime levels as they change over time, results are robust to a high-crime designation based on the number of violent crimes.Finally, if a suspect for the homicide is identified, routine police work may include stopping individuals in the neighborhood who have characteristics that match the identity of the suspect, for instance, the suspect's race and ethnicity.Therefore the suspect's characteristics may be highly correlated with those of the residents in the neighborhood who are stopped.Conversely, if stops also increase for individuals who do not share the same characteristics as the suspect, this may be additional evidence of the collateral effect of homicide on neighborhood residents.Unfortunately, we only have information about the race and ethnicity of suspects for 24 percent of the homicides in our sample and are therefore unable to make meaningful observations about how suspect race and ethnicity is related to SQF activity.

Implications
The results from this study have implications for understanding the full cost of community violence and how this cost is distributed across the city's neighborhoods.The shock and fear associated with extreme violence are likely to affect people in all communities where violence occurs across NYC.This analysis reveals an additional consequence of extreme violence that is only experienced in some communities and not others.When a homicide occurs in predominantly white neighborhoods, we find no evidence that SQF activity increases.In neighborhoods that are not majority white, a homicide brings with it an escalation of police stops of community residents.The escalation of SQF activity is most pronounced in higher-crime neighborhoods, and it is greatest in block groups where the majority population is Hispanic or African American.
This descriptive analysis thus provides evidence on one additional pathway through which violence affects daily patterns of life within communities of color.In majority black and majority Hispanic neighborhoods, particularly those that have a high level of crime, a homicide changes the residential setting and alters the role of young people within the community.By virtue of their address alone, young people experiencing the fear and shock that come with extreme violence become potential witnesses, suspects, or accomplices, and the residential setting is turned into a crime scene.

Notes
1 Since the Terry v. Ohio ruling in 1968, police in the United States may legally stop, question, and frisk citizens based on a "reasonable suspicion" of illegal activity, based on an appearance similar to that of a crime suspect's, based on concern that the person is carrying a weapon, or if the person is leaving the scene of a crime (Jones-Brown, Gill, and Trone 2010).In NYC, SQF activity was further defined by People v. De Bour in 1976, which distinguishes stops due to reasonable suspicion from arrests, which require "probable cause" of the commission of a crime.
3 The average population of block groups in NYC in 2000 was 2,014 residents.The analysis excludes block groups with zero population.The use of block groups may mask some variation in criminal activity and policing that occur on the street segment/blockface level.Any increase in stops may be concentrated on a few street segments within a block group.Alternatively, an increase in stops following a homicide on one street may "spill over" onto the adjacent streets that make up the block group.The current analysis is not able to make these distinctions.9 In an alternative specification of "high crime," we stratify the sample by the total number of homicides during the study period: one homicide, two homicides, three homicides, or four or more homicides.However, in the body of the article, we focus on the "high-crime area" models because we think they are a better proxy for differences in total crime: total homicides may be less correlated with total crime because homicides can be somewhat random.Models using a second alternative "high-crime area" designation based on whether the total number of violent crimes in a census tract between 2006 and 2009 is above the census tract average for the period provide substantively similar results.
10 The small percentage of stops that result in arrest (5.5 percent) may have a crime-fighting benefit to the neighborhood.However, the vast majority of stops that do not result in arrest may have larger consequences for residents than benefits.We reestimate the main models using a measure of stops that did not result in arrests and find substantively identical results.Contact authors for table.
11 Supplementary Table 2 presents the results of the main models with two different dependent variables: the number of black stops and the number of nonblack stops.The sociological science | www.sociologicalscience.com first two columns show that following a homicide, the increase in the number of stops of black residents is much larger (0.24) than the increase in the number of stops of residents who are not black (0.09), although both are statistically significant.These estimates represent a 50 percent and 20 percent increase from the overall mean number of stops, respectively.
12 In Supplementary Table 2, columns 3 and 4, in majority black neighborhoods, the number of stops of black residents increases by 0.35 following a homicide (a 60 percent increase from the mean number of stops in majority black neighborhoods), but there is no change in the number of stops of nonblack residents.In majority Hispanic neighborhoods, the number of stops of all residents increases, but the number of stops of black residents increases more, on average (black stops increase by 47 percent from the mean stops in majority Hispanic neighborhoods, compared to an increase in nonblack stops of 27 percent).
14 See Supplementary Table 4.In high-crime majority black neighborhoods, the number of stops of black residents increases by 0.53 following a homicide (79 percent increase), whereas the number of nonblack stops does not increase.In high-crime majority Hispanic neighborhoods, the number of stops of black residents increases by 66 percent and the number of stops of nonblack residents increases by 23 percent following a homicide.
In lower-crime neighborhoods, the number of black stops increases in majority black, majority Hispanic, and mixed race/ethnicity neighborhoods by 56 percent, 36 percent, and 54 percent from the respective means.
16 During the study period, the average number of homicides in majority black block groups was 0.66, in majority Hispanic block groups 0.49, in majority Asian block groups 0.21, and in majority white block groups 0.10, and it was 0.19 in mixed block groups.

Figure 2 :
Figure 2: Average Stops in the Week Before and After a Homicide, by Neighborhood Composition.Note: * Indicates mean difference is significant at 0.001

Figure 3 :
Figure 3: Summary of Key Findings; notes: Each bar represents the coefficient from a separate regression model.All models include precinct fixed effects, controls for neighborhood composition, poverty rate, and day of the week, month, and year.All estimates presented are statistically significant at the p < 0.01 level.See Tables3 and 4, and Online Supplemental Tables B and C for full results.

Table 1 :
Distribution of Stops and Frisks byRace & Ethnicity (2006-2009) SQF activity are available to the public on the NYPD website and include the location of the stop; the actions taken by police (stop, frisk, and/or arrest); the reason for the stop; and information about the person stopped, including race, ethnicity, age, and gender.The total number of stops conducted by the NYPD increased by 17 percent from 2006 to 2009, our study period, with more than 580,000 stops occurring in 2009 (Figure

Table 3 :
Relationship between Local Homicides and Total Stops (2006-2009) a Controls include Day of the week, month, and year fixed effects.Reference category is majority white.Standard errors in parentheses.* † p < 0.01.

Table 4 :
Relationship between Local Homicides and Stops, High Crime Area designation (2006-2009) a Controls include Day of the week, month, and year fixed effects.Reference category is majority white.
The only way to investigate variation in policing by neighborhood demographics is to use U.S. Census geography.Although we recognize the utility and importance of investigating these relationships at a lower level of aggregation, this particular analysis necessitates the use of block group data.4 Via the New York Times website.Of the total homicides during this period, 38 percent have black or black/Hispanic suspects, 16 percent have other race suspects, and 46 percent are missing the race of the suspect.5 In the precinct fixed effect models, standard errors are clustered at the precinct level.6 In our sample of block groups that ever experienced a homicide between 2006 and 2009, there are very few majority Asian block groups; therefore we drop these from the rest of the analyses.See Supplementary Table 1 for means and sample sizes.7 Estimating equation: stops bp,mdy = β 0 +β 1 AfterHom bp,mdy + β 2 BefAfter bp,mdy + β 3 majBlack bp + β 4 majHispanic bp + β 5 MixedRace bp + β 6 PovRate bp + β 7 (AfterHom) × majBlack bp + β 8 (AfterHom) × majHispanic bp + β 9 (AfterHom) × MixedRace bp + β 10 (BefAfterHom) × majBlack bp + β 11 (BefAfterHom) × majHispanic bp + β 12 (BefAfterHom) × MixedRace bp + γ m + δ d + θ y + ϑ p + ε bp,mdy .8 Models using the most controlled specification and stratified by the age of the resident stopped (younger than 30 years or older than 30 years) indicate that twice as many stops of younger residents occur than stops of older residents, and for younger residents, stops are more prevalent among black residents compared to Hispanic residents or residents of other races.Contact authors for tables.