Racial Disparities in Crime, Arrests, and Police Interaction: NYC, Chicago, LA, Schemes and Mind Maps of Criminology

Download Racial Disparities in Crime, Arrests, and Police Interaction: NYC, Chicago, LA and more Schemes and Mind Maps Criminology in PDF only on Docsity! 1 Race, Crime, and Police Interaction John MacDonald Professor of Criminology and Sociology University of Pennsylvania September 26, 2021 Conference Paper Federal Reserve Bank of Boston Economic Research Conference Series: Racial Disparities in Today’s Economy, 64th Economic Conference Abstract This paper examines the relationship between racial disparities in poverty, arrests, crime, and police interactions. This study reviews research on racial/ethnic disparities concentrated poverty and its association with disparities in crime victimization and official police interactions. An analysis of 221 large U.S. cities in 2014-2018 examines the association between racial disparities in poverty, unemployment, and arrests. The results show that arrest rates for blacks relative to whites are significantly higher even after controlling for the level of concentrated disadvantage. Even among the cities that rank highest in concentrated disadvantage for whites, blacks on average have higher unemployment and are more likely to live below the poverty line. These findings confirm that blacks and whites on average in large U.S. cities live in largely different environmental contexts. An examination of the spatial concentration of economic disadvantage, crime, and arrest patterns in New York City, Chicago, and Los Angeles for years 2014-2019 shows that black and Hispanic monthly arrest rates are significantly higher in census block groups with greater levels of concentrated economic disadvantage. Arrest rates for whites and other groups are either negatively associated or have no relationship with the level of concentrated disadvantage. In all three cities, the level of reported crime is more strongly associated with black and Hispanic disparities in arrest rates than concentrated disadvantage. In New York and Chicago, a substantial share of disparities in black arrest rates are driven places in the top five percentiles of reported crime, and the same pattern holds for Hispanic arrest rates in Los Angeles. The results suggest high crime places located in areas with concentrated poverty help explain a significant share of black and Hispanic disparities in arrest rates. The paper concludes that investing in place-based programs that improve public safety could reduce racial disparities in police contact. Introduction Serious crime, poverty, and police activity are highly concentrated by place. Black Americans are on average more likely than white Americans and other groups to live in neighborhoods characterized by concentrated disadvantage, reflecting higher spatial concentrations of poverty, unemployment, joblessness, family disruption, and geographic isolation (Sampson & Wilson, & 2 Katz, 2018). Just three to five percent of places and street segments in a given city generate at least fifty percent of crime (Sherman et al., 1989; Weisburd, 2006). Racially isolated neighborhoods of concentrated disadvantage are more likely to have these hot spots of crime and police contact (Braga & Weisburd, 2010; Sampson, 2011). The substantial spatial inequality in the concentration of poverty, violent crime, and social resources connect to historic and contemporary patterns of racial residential segregation (Massey & Denton, 1993). The number of calls for service and crime typically influences patterns of police deployment in U.S. cities. The extra-allocation of police to high crime areas is particularly evident in cities like New York that adopted the “new policing model” of linking officer assignments to crime analytics (MacDonald, Fagan, & Geller, 2016). These disparities by place are fundamentally important for thinking about who is most likely to encounter a police officer, especially in the context of highly discretionary activities like the decision to stop and question someone suspected of a crime or make an arrest. In this paper, I examine whether concentrated disadvantage at the city and census block group level explains a significant share of the racial disparities police arrests. I review some of the empirical research on racial disparities in police stops, arrests, and use of deadly force. I discuss how spatial patterns of concentrated disadvantage may help explain a substantial share of racial disparities in the police interactions, like the decision to stop, question, and frisk someone or to make an arrest. The review focuses on published studies that examine racial disparities in stops and arrests, with some discussion of police use of deadly force. An analysis of city level data on arrests for serious crimes examines how much differences in concentrated poverty explain the gap in arrest rates for blacks relative to whites. An analysis of micro data from New York, Chicago, and Los Angeles estimates how the disparities in arrest rates for blacks and Hispanics relative to whites and other groups is accounted for by the level of concentrated poverty and reported crime between census block groups. Finally, I discuss the consequential role of historic and contemporary fractured police-minority relationships and the need for more research on testing how police can collaborate with other municipal service agencies and community groups to address problematic crime hot spots that generate a disproportionate share of arrests. An evidence-based policing model that focuses on places may help reduce racial disparities in police contact and improve public safety in the neighborhoods with the greatest levels of concentrated disadvantage in the U.S. 5 The spatial concentration of disadvantage is also important for helping explain some patterns in racial disparities in police contact and arrests. Sampson (1986) shows that even after controlling for self-reports of serious delinquency youth in Seattle who are black and living higher poverty neighborhoods are more likely to experience a police arrest. These findings suggest that exposure to police and discretion by place and race may condition police discretion in deciding whether to arrest a youth for a crime. Kirk (2006) found in a longitudinal sample of youth in Chicago that the probability of arrest at age seventeen was 29 percent for blacks compared to 12 percent for whites, but that black youth were significantly more likely to live in areas of concentrated poverty that were racially segregated.1 The expected black-white disparity in arrest rates is 21 percent lower after accounting for neighborhood differences in concentrated poverty, racial segregation, and other factors. There is a considerable body of research suggesting that police deployment and interactions with citizens vary considerably by neighborhood environments. Klinger (1997) argues that the deployment of police by geography in cities exposes officers in different units to varying levels of crime and disorder. Within patrol areas, norms develop among police officers on the style of policing and their propensity to enforce the law. Research has found that police discretionary decisions to stop a suspect or make an arrest vary considerably by neighborhoods (Fagan and Davies, 2000; Gelman, Fagan, and Kiss, 2007; Smith, 1986). National estimates from the Police Public Contact Survey (PPCS) in 2015, a supplement to the National Crime Victimization Survey, find 14.55 per 1,000 black people report experiencing a street stop in the prior year compared to 9.07 for whites.2 Here the data suggests that the disparities are greater for street stops than traffic stops, consistent with the fact that police deployment, crime, and poverty are highly concentrated in urban cities in racially segregated neighborhoods. A primary challenge with research on racial disparities in police contact is establishing the benchmark for who should be at risk for a police stop and/or arrest. Ridgeway and MacDonald (2010) and Neil and Winship (2019) provide a summary of the methodological challenges with establishing who is at risk for being stopped by the police and why most approaches do not provide credible inference. Setting aside the issue of the appropriate 1 For black youth in the sample on average 78 percent of the population of their neighborhoods were comprised of black residents. For white youth on average 49 percent of their neighborhoods were comprised of white residents. 2 https://bjs.ojp.gov/content/pub/pdf/cpp15.pdf 6 benchmark for the population at risk for police stops, research shows that stop rates are higher in neighborhoods with a higher percentage of black residents, even after controlling for neighborhood levels of poverty and crime (Fagan et al, 2010; MacDonald and Braga, 2019). Fryer (2019) shows that population level black-white disparities in the ratio of stop rates declines from 4.23 to 1.43 after controlling for crime and arrest rates across police precincts in New York City, suggesting that a substantial share of the disparity in stop rates is explained by differences in crime across places. MacDonald and Braga (2019) show that in New York City the stops rates are no longer associated with the percentage of black residents in neighborhoods after the police began a series of reforms as part of a federal court settlement. In comparison to estimates of disparities in police stops, less research on racial disparities in police arrest rates examines variation by place. Studies typically examine how concentrated disadvantage and other factors are associated with city level differences in racial disparities in police arrest rates. Parker et al. (2008), for example, find that economic disadvantage as measured by a composite measure of rates of poverty, unemployment, and educational attainment is associated with higher black and white arrest rates in large U.S. cities in 2001, though the association is larger for white rates than it is for blacks. In one of the only studies to examine how arrest rates vary by neighborhood environments, Smith (1986) finds that police were more likely to make investigatory stops and arrests in neighborhoods with greater concentrations of poverty, though the study does not control for actual reported crime in neighborhoods and relies on residents’ perceptions. Smith (1986) also finds that the probability of arrest is higher for black suspects in neighborhoods with majority black populations. To my knowledge existing research has not examined how the levels of disadvantage and crime by place are associated with racial disparities in arrest rates. One likely reason is that arrest data until recently was not readily available to scholars with detailed geographic coordinates. In terms of racial disparities in use of force by the police, there is a paucity of empirical work that examines whether poverty and the level of crime by place is associated with higher risks for blacks and Hispanics relative to whites. Fryer (2019) offers one of the most comprehensive studies and finds that blacks suspects are 46 percent more likely in a stop to have forced used than white suspects in New York City, but that this disparity is reduced to a difference of 18 percent after controlling for precinct and year. MacDonald and Braga (2019) report similar disparities in use of force in New York City after controlling for encounter and 7 location characteristics of stops, but that the disparities reverse by 2015 after court settlement reforms. Fryer (2019) shows in a national sample of public police contacts that black respondents are 18 percent more likely than white respondents to report having any use of force in a police interaction in the past year, and that general location and encounter-related factors do not substantially reduce the disparity. An important limitation in this analysis is insufficient base rates and location information to estimate how much racial disparities in force are associated with levels of crime and concentrated disadvantage by places. When it comes to estimating racial disparities in police use of deadly force there are few studies that offer any assessment of the role of place-related factors. Police use of deadly force is rare relative to stops and arrests, so estimates of racial disparities in deadly force that attempt to control for location related factors are likely to be statistically underpowered. Studies have attempted in recent years to estimate disparities in officer involved shootings by comparing rates of shootings for black, Hispanic, and white suspects relative to arrests deemed at greater risk for a shooting (e.g., aggravated assault, robbery, attempted murder of a police officer). Fryer (2019), for example, finds that officers in Houston are less likely to shoot black suspects than white suspects relative to random draw of arrests for aggravated assault against a police officer, attempted murder of a police officer, resisting arrest, evading arrest, interfering in an arrest, and arrests with tasers used. Adding suspect, officer, and encounter related variables does not change the association. Fryer (2019), however, does not assess the associations between shootings and location related factors like crime or concentrated poverty. Klinger et al. (2015) attempt to assess the association between police shootings in general, concentrated poverty, crime, and the percent of black residents of neighborhoods in St. Louis.3 They find that officer involved shooting rates per neighborhoods are highest in areas with higher levels of gun violence, and that percent of black residents of neighborhoods is not associated with shooting rates. With a total of 230 officer involved shootings over 355 census block groups and a correlation of .69 between firearms violence and percent of black residents, this study is under powered to test for differences across these covariates. Legewie and Fagan (2016) provide one of the only recent city-level (n=266 cities) studies of black-white disparities in fatal police shootings (collected from crowd source data) per population or per arrests. They find a small association between city-level difference in 3 In these data over 90 percent of police shootings involved black suspects, and there is no reference group for cases that did not involve shootings, making it impossible to make any inferences about racial disparities. 10 D. Summary In U.S. cities crime is highly correlated with the concentration of poverty, such that the two go hand in hand. A few studies suggest that street stops are disparate in the places that generate higher levels of serious crime, but few studies examine what share of the racial disparity in arrests is attributable to the environmental context of locations. Additionally, there is the potential that crime is actually a poorly used proxy by the police. Grunwald and Fagan (2019), for example, find during the height of the use of stop, question, and frisk activity in New York City there was very little correlation between an officer indicating suspicion based on the legally permissible indicator of high crime area and the actual level of crime in that area. While criminal behavior in high crime locations may influence a significant share of racial disparities in police stops, perceived suspicion based on loose heuristics of an area being high crime may produce unjustified police actions in stopping individuals. Research on racial disparities in police arrests is especially thin when it comes to understanding how much arrest rates are associated with area differences in reported criminal activity and the level of concentrated disadvantage. Focusing police activity in the highest crime street segments make sense from a crime control perspective, given that crime is highly concentrated by location (Weisburd, 2006), but we have little research that examines how much population level disparities in arrests are driven by the concentration of poverty and crime. II. Racial Disparities in Poverty, Crime, and Police Interactions A. Aggregate Disparities Racial disparities in poverty, crime, and police contact are an established fact in the United States. Data from the census American Community Survey (ACS) estimates of poverty in years 2015 to 2019, for example, shows that blacks and Hispanics consistently have a higher share of the population living below the poverty level. Table 1 shows that all groups there was some improvement between 2015 and 2019, but in general blacks and Hispanics are roughly 2 to 1.8 times more likely than whites to live in poverty in the United States. Table 1. Race/Ethnic Disparities in Percent Population Living in Poverty Year White Black Hispanic 2015 12.2% 25.4% 22.6% 2016 11.6% 23.9% 21% 2017 11.1% 23% 19.4% 2018 10.9% 22.5% 18.8% 11 2019 10.3% 21.2% 17.2% Mean 11.22% 23.20% 19.80% Source: American Community Survey, Census Bureau https://data.census.gov/cedsci/table?q=poverty%20status&tid=ACSST1Y2015.S1701&hidePreview=true Separate analyses examining ACS data by county shows that blacks and Hispanics are on average about 2 times more likely than whites to live below poverty in urban counties with populations of over 500,000 people. These statistics, however, mask how much the disparity in poverty varies by geographic concentration within cities. Table 2 shows the data from the National Crime Victimization Survey and the FBI’s Uniform Crime Reports averaged for years 2015-2019. From these descriptive data, we can compare the proportion of black, white, and Hispanics in the population to representation in race of victims of robbery and aggravated assault reported in the NCVS and arrests of suspected offenders in the UCR. Hispanics are not separately distinguished from racial categories so the percentages exceed 100% when including this group. The data show that a higher proportion of blacks are arrested for robbery and assault compared to their representation in the population or as crime victims. Hispanics and whites are arrested proportionally closer to their victimization proportions in the NCVS. While the black-white disparity is larger in arrests than victimizations, it is hard to draw strong conclusions about the sources of the disparities from these aggregate data. Table 2. Racial Disparities in Victimizations and Arrests for Robbery and Aggravated Assault, Average 2015-2019 Race/Ethnicity Population Robbery Victims Robbery Arrests Assault Victims Assault Arrests White 60.4% 47.3% 48.8% 59.5% 62.5% Black 12.5% 18.8% 48.8% 13.3% 33.2% Hispanic 18.3% 23.7% 23.1% 19.8% 24.9% Sources: Bureau of Justice Statistics, NCVS Victimization Tool and FBI, Uniform Crime Reports, 2015-2019. Assaults represent aggravated felony assaults. Given that most interpersonal offenses are intra-racial, the share of blacks arrested for robbery should be substantially lower if arrests are a random sample of those victimized. Data from the 2018 NCVS shows that blacks are about twice as likely to be offenders compared to their victimization percentages.5 The 2019 NCVS shows that around 46 percent of victims of 5 See https://bjs.ojp.gov/content/pub/pdf/cv18.pdf table 12. 12 non-fatal violent offenses report the offender’s race as black.6 These data suggest that the higher rate of arrests for blacks is likely a reflection of higher offending rates relative to their share of the population and victimizations. A recent report from the Bureau of Justice Statistics examines the micro-data from the NCVS for 2018 and compares the race and ethnicity of offenders observed by victims, as well as those reported to the police (Beck, 2018). Here we have estimates for the race of offenders from the perspective of the victims of aggravated assault, robbery, and sexual assault, and how that compares to UCR arrest data for these same offenses. Table 3 shows that in 2018 arrest percentages for whites and blacks is closely comparable to the perceived race of reported offenders in nonfatal serious violent crimes. A higher proportion of Hispanics are arrested by the police relative to the perceived ethnicity of offenders in victimization data. A challenge with these comparisons, however, is that the race and ethnicity of the offender is what the victim perceives and may be hard for victims to determine. Table 3. Race or Ethnicity of Offenders in NCVS and Persons Arrested for Serious Violent Crime, 2018 Race/Ethnicity Offenders in NCVS Offenders in NCVS Reported to Police UCR Arrests White 43.8% 40.9% 38.7% Black 35.9% 42.8 % 36.1% Hispanic 15.5% 12.0% 21.4% The lack of disparities between reported race of offenders in the NCVS and UCR arrests, however, should not be surprising given the differences in the spatial concentration of poverty, race/ethnicity, and crime in cities. Table 4 presents some descriptive data on racial disparities in homicides caused by firearms as reported in the FBI’s Supplementary Homicide Reports for years 2015-2019 and the Washington Post data on police shootings. Here the focus on black and white disparities, because the SHR homicide statistics across cities are not consistent in reporting the ethnicity of known offenders or victims. 6 https://bjs.ojp.gov/content/pub/pdf/cv19.pdf Table 15. 15 these years to the census block. Census blocks represent blocks in the same contiguous block group and often correspond to a city block. Stops by race/ethnicity and crimes were aggregated to the level of block (month-year). The data shows there is a close connection between the location of stops and total reported crimes per month to the police. The rank order correlation shows that the total number of stops for years 2013 to 2015 are highly associated the total number of reported crimes (r= 0.2038; p<.0001; n= 99,703). Results Table 5 shows that when examined by the ratio of stops to crime the burden of stops still falls disproportionately on blacks and Hispanics. On average blacks and Hispanics are stopped at a higher rate relative to the crime reported in a given census block. However, the disparities in these ratios diminishes over time as the NYPD reduced its use of stop, question, and frisk, as is evident from the declining differences between 2013 and 2015. Table 5. Ratio of Stops by Race to Reported Crime in NYC, 2013-2015 Stop to Crime Black White/Other Hispanic White/Other 2013 1.284 (.041) .591 (.013) .953 (.021) .665 (.016) Diff .692 (.040) ** .287 (.022) ** N= 3,267 3,253 2014 .716 (.036) .459 (.021) .691 (.033) .560 (.024) Diff .256 (.031) ** .130 (.029) ** N= 566 513 2015 .634 (.062) .442 (.033) .635 (.045) .564 (.040) Diff .192 (.055) ** .070 (.040) ** N= 194 214 All Years 1.173 (.034) .566 (.011) .902 (.017) .646 (.013) Diff .607 (.033) ** .255 (.018) ** N= 4,027 3,890 Note: Standard errors in parentheses. **p<.001 An analysis of these patterns shows that the differences in the racial disparities in stops to crime ratios are greatest in areas with lower reported criminal offenses. Figure 1 shows the marginal estimates of these disparities from a regression model that examines differences at twenty quantiles of reported crime. One can see that the disparities in ratios between blacks and Hispanics relative to whites and others diminishes as the level of crime reported in a census block increases. This finding suggests that disparities in stops relative to crime are highest in 16 places with the least amount of crime reported. The findings suggest that there are racial disparities in who is committing crime in relatively low crime blocks or that police are engaged in racial profiling in deciding whom to stop and question for suspected criminal activities. C. City Level Arrest Disparities Given the paucity of research in recent years examining the association between racial disparities in concentrated poverty and police arrest rates, the present analysis re-examines this issue with recent data. Data and Measures The data for the city level analysis of arrest disparities between blacks and whites comes the Chalfin et al. (2020) study of police force sizes, crime, and arrests in 242 U.S. cities with populations greater than 50,000 in 1980 and regularly report data to the U.S. Census Bureau Annual Survey of Government (ASG). These data combine city level measures of crime and arrests captured by the Uniform Crime Reports (UCR) system of the Federal Bureau of 17 Investigation. The final sample consists of 221 cities with complete data on crime and arrests for index offenses (murder, rape, robbery, aggravated assault, burglary, grand larceny, and motor vehicle theft) for blacks and whites for years 2014-2018. Index offenses measure seven felony crimes measured uniformly across cities as part of the FBI’s annual survey of crime. These data were combined with U.S. Census Bureau population for each city captured in the annual American Community Survey (ACS) (five year estimates for years 2014-2018). Race-specific measures of concentrated disadvantage for each city were measured by a standardized composite scale (mean centered at zero) of the black or white percentage of the population living below poverty, the percentage of the population unemployed, and the median household income from ACS data. Measures for population density from the ACS and the per capita public expenditures for each city from the ASG are also included. Region is measured for each city according to Federal Information Processing (fips) classifications (Northeast, Midwest, South, West).8 Empirical Model The empirical model examines the extent to which race-specific measure of concentrated disadvantage are associated with yearly city level disparities in black and white arrest rates for index offenses. Rates of arrest reflect the per capita population. A Poisson regression model estimates the arrests rate per city (i) for each group (j) (blacks or whites) separately, and includes the population of blacks or whites as exposure variable. This approach converts the counts of arrests to a rate per population (black or white). The model estimated takes the following form: log ( (𝜆𝑖𝑡 𝑗 ) 𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑖𝑡 𝑗 ) = 𝛽0 + 𝛼k%𝐵𝑙𝑎𝑐𝑘𝑖𝑡 + 𝜇k%𝐻𝑖𝑠𝑝𝑎𝑛𝑖𝑐𝑖𝑡 + ϒ𝐶𝑟𝑖𝑚𝑒 𝑅𝑎𝑡𝑒𝑖𝑡 + 𝜃k𝐶𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑒𝑑 𝐷𝑖𝑠𝑎𝑑𝑣𝑎𝑛𝑡𝑎𝑔𝑒𝑖𝑡 𝑗 + 𝜎𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝐷𝑒𝑛𝑠𝑖𝑡𝑦𝑖𝑡 + 𝜋𝑃𝑒𝑟 𝐶𝑎𝑝𝑖𝑡𝑎 𝐸𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒𝑠𝑖𝑡 + 𝜂𝑟 + 𝛿𝑡 In each model the black or white arrest rate (λ) per month is estimated assuming a Poisson distribution after controlling for crime rates (reported index offenses), the race/ethnicity percentages of the population (% Black, % Hispanic, % Other), per capita public expenditures, 8 https://www.census.gov/geographies/reference-files/2020/demo/popest/2020-fips.html 20 (0.156) (0.148) Crime rate 1.000 (0.0000201) Observations 1055 1055 1055 Exponentiated coefficients (Incidence Rate Ratio); Standard errors in parentheses; Reference groups are 1st (0-33 percentile) for Quantiles, 2014 for year, and Northeast for region. Concentrated Disadvantage represents average of percentage of blacks in poverty, percentage of unemployed, and median household income. * p < .05, ** p < 0.01 Table 8 presents the results from the estimates of the white arrest rate in each city before (1) and after including measures of concentrated disadvantage (2) and crime (3). Cities that rank higher in the proportion of black population have a significantly lower white arrest rates. Column 1 shows that white arrest rate declines by approximately 22.2 and 37.3 percent for cities that are in the second and third quantile relative to the first (0-33 percentile) in percentage of black residential population (3). The lower white arrest rate in cities with a majority black population closely mirrors the lower black arrest rate. Table 8 also shows that the white arrest rate for index offenses is 54.8% higher in cities that rank in the top quantile of white concentrated disadvantage, even after controlling for the crime rate (3). Table 8. City Level Index Arrest Rate for White, 2014-2018. (1) (2) (3) Index arrests, White Index arrests, White Index arrests, White Quantiles % Black=2 0.787* 0.768** 0.778** (0.0750) (0.0690) (0.0736) Quantiles % Black=3 0.671** 0.637** 0.627** (0.0951) (0.0873) (0.0865) Quantiles % White=2 0.732** 0.837* 0.818* (0.0803) (0.0710) (0.0688) Quantiles % White=3 0.784 0.875 0.863 (0.112) (0.114) (0.114) Quantiles % Hispanic=2 1.225* 1.197 1.215 (0.125) (0.121) (0.124) Quantiles % Hispanic=3 1.517** 1.430** 1.506** (0.204) (0.197) (0.208) Expenditures per 1,000 1.000 1.000 1.000 (0.0000263) (0.0000256) (0.0000258) Population density 1.000* 1.000 1.000 (0.0000183) (0.0000183) (0.0000182) Year=2015 0.956** 0.953** 0.952** (0.0114) (0.0115) (0.0106) 21 Year=2016 0.888** 0.884** 0.879** (0.0219) (0.0224) (0.0215) Year=2017 0.807** 0.803** 0.801** (0.0231) (0.0234) (0.0221) Year=2018 0.769** 0.766** 0.774** (0.0247) (0.0250) (0.0240) Midwest 0.729** 0.837 0.777* (0.0832) (0.0978) (0.0939) South 0.994 1.318 1.183 (0.138) (0.197) (0.198) West 1.106 1.342** 1.216 (0.122) (0.140) (0.143) Disadvantage, White=2 1.259** 1.181* (0.109) (0.0960) Disadvantage, White=3 1.668** 1.548** (0.186) (0.180) Crime rate 1.000* (0.0000254) 1055 1055 1055 Exponentiated coefficients (Incidence Rate Ratio); Standard errors in parentheses; Reference groups are 1st (0-33 percentile) for quantiles, 2014 for year, and Northeast for region. Concentrated Disadvantage represents average of percentage of whites living below poverty, percentage unemployed, and median household income. * p < .05, ** p < 0.01 22 Figure 2 shows the expected black and white arrest rates from the models estimated in Tables 7 and 8 (column 3) by level of concentrated disadvantage. The black arrest rate is significantly higher than the white arrest rate at every level of concentrated disadvantage. These findings highlight that the racial disparity in arrests is not simply a function of city level differences in poverty, unemployment, and median household income. In even the most economically disadvantaged cities for the white population the level of unemployment and poverty is considerably lower than it is for the black population. Specifically, for cities that rank in the top quantile of white concentrated disadvantaged the unemployment percentage is 11.59 for the white population compared to 19.74 for the black population. In this top quantile of white concentrated disadvantage 23.4 percent of the white population lives below the poverty line compared 34 percent for the black population. Flint, MI, Detroit, MI, Pontiac, MI, Camden, NJ, and Reading, PA rank in the top five of cities with the highest level of concentrated disadvantage for the white population. In these five cities, the percentage of the white population that is unemployed or living below the poverty line is considerably lower than it is for the black 25 In each model, the black or white arrest rate per block group (i) is estimated controlling for concentrated disadvantage and number of crimes or victims of the same race/ethnicity in a given month. The terms (η) and (𝛿) refer to region (r) and year (t) fixed effects. For New York City, Chicago, and Los Angeles regions are defined by the Borough, Ward, or LAPD Division in which the census block group is located. Standard errors are clustered at the block group to correct for over-dispersion and unmeasured dependence within cities (Wooldridge, 2010). Results Figure 3 shows the basic spatial pattern of arrests per census block group for black and Hispanic arrestees for years 2014-2019 for New York, Chicago, and Los Angeles. The figure shows that there is some spatial concentration in arrest patterns across each city. Spearman rank order correlations also indicates that arrest rates for blacks and Hispanics in each city are associated with the percentage of the population of the same race and ethnicity, number of reported criminal offenses, and the level of concentrated disadvantage. 26 Figure 3. Spatial Concentration of Arrests of Blacks and Hispanics 2014-2019, New York, Chicago, and Los Angeles 27 Table 9 examines the spatial concentration of arrests, crime, demographics of residential population, and concentrated disadvantage in each city using the Moran’s I statistic.13 The level of spatial concentration of arrests for blacks and Hispanics is the highest in Chicago, which may be a consequence of a higher spatial uniformity of crime and residential segregation for black residents. Table 9: Spatial Concentration of Arrests, Crime, and Concentrated Poverty Measure New York (n=6,291) Chicago (n=2,299) Los Angeles (n=2,579) Black arrests .073** .319** .224** Hispanic arrests .079** .349** .162** Crime .136** .220** .104** Percent Black Residents .609** .700** .559** Percent Hispanic Residents .594** .583** .553** Concentrated Disadvantage .531** .556** .541** 13 Moran’s I was calculated based on a power function of –distance2 (kilometers) between focal census block group (i) and other block groups (j). 30 and crime or same race/ethnicity of victims reported in each census block group. The white arrest rate of 0.78 and is approximately 16 percent lower after controlling for concentrated disadvantage and crime or same race victims of crime. A comparison of regression coefficients across models (Clogg et al., 1995) that adjusts covariance for clustering standard errors on block groups (White, 1982), shows that the estimate of concentrated disadvantage is significantly larger for the black (Chi-square (1) 237.23, p<.001) and Hispanic (Chi-square (1) 341.86; p<.001) arrest rates compared to white and other group arrest rates. For all groups the number of victims of the same race or ethnicity is substantially associated with the arrest rate, though the size of the association is significantly larger for black and Hispanic arrest rates than white and other groups. The results for Chicago show a black arrest rate of 1.20 per month and is 63 percent lower (.448) after controlling for concentrated disadvantage and the crime rate of census block groups. The relationship between concentrated disadvantage appears to be a major driver of a lowering of the expected black arrest rate. Concentrated disadvantage also has a significantly larger association (Chi-square (1)105.3; p<.001) with black arrest rates than it does with white and other groups and a somewhat larger association than for Hispanics (Chi-square (1) 3.68; p=.055). Levels of crime are associated with higher arrest rates for all groups, and the size of the association is only slightly larger for black and Hispanic rates relative to white arrest rates. Table 11: Rate of Arrests for Black, Hispanic, and White or Other Groups (1) (2) (3) (4) (5) New York Black Black Black Black Black Disadvantage 1.192** 1.226** 1.052** (0.00334) (0.0228) (0.0163) Criminal Offenses 1.030** (0.00308) Black victims 1.199** (0.00578) Average rate 2.050 1.904 1.805 1.630 1.462 Observations 417427 417390 387188 381527 381527 Hispanic Hispanic Hispanic Hispanic Hispanic Disadvantage 1.216** 1.241** 1.094** (0.00305) (0.0221) (0.0195) Criminal offenses 1.029** (0.00301) Hispanic victims 1.197** (0.0115) 31 Average rate 1.455 1.273 1.198 1.098 1.070 Observations 417427 417390 387188 381527 381527 White/Other White/Other White/Other White/Other White/Other Disadvantage 0.925** 0.941** 0.963 (0.00296) (0.0188) (0.0192) Crime offenses 1.027** (0.00296) White/other victims 1.030** (0.00359) Average rate 0.782 0.696 0.699 0.657 0.667 Observations 417427 417390 387188 381527 381527 Chicago Black Black Black Black Black Disadvantage 1.253** 1.213** (0.0241) (0.0201) Criminal Offenses 1.016** (0.00252) Average rate 1.201 0.512 0.460 0.448 Observations 154604 154595 147277 147277 Hispanic Hispanic Hispanic Hispanic Hispanic Disadvantage 1.188** 1.160** (0.0257) (0.0234) Criminal offenses 1.015** (0.00241) Average rate 0.264 0.133 0.132 0.129 Observations 154604 154595 147277 147277 White/Other White/Other White/Other White/Other White/Other Disadvantage 1.009 0.979 (0.0266) (0.0224) Crime offenses 1.013** (0.00208) Average rate 0.157 0.0905 0.0906 0.0894 Observations 154604 154595 147277 147277 Los Angeles Black Black Black Black Black Disadvantage 1.056 1.120 1.068 (0.0635) (0.0659) (0.0486) Criminal Offenses 1.016** (0.00384) Black victims 1.082** (0.0140) Average rate 1.550 0.911 0.874 0.866 0.865 Observations 127701 127701 123865 121328 121328 Hispanic Hispanic Hispanic Hispanic Hispanic Disadvantage 1.198** 1.236** 1.204** (0.0346) (0.0299) (0.0299) Criminal offenses 1.019** (0.00365) 32 Hispanic victims 1.055** (0.0100) Average rate 2.488 2.200 2.119 2.077 2.072 Observations 127701 127701 123865 121328 121328 White/Other White/Other White/Other White/Other White/Other Disadvantage 0.929 0.955 0.976 (0.0416) (0.0397) (0.0380) Crime offenses 1.020** (0.00385) White/other victims 1.040** (0.00587) Average rate 1.314 0.855 0.840 0.826 0.820 Observations 127701 127701 123865 121328 121328 Year fixed effects No Yes Yes Yes Yes Region fixed effects No Yes Yes Yes Yes Exponentiated coefficients (Incident Rate Ratios); Standard errors in parentheses clustered at census block group. ** p < 0.01 The results for Los Angeles show a black arrest rate of 1.55 per month, which is 44 percent lower after controlling for concentrated disadvantage and crime or victimizations of black residents. The Hispanic arrest rate of 2.49 per month is approximately 17 percent lower after controlling for concentrated disadvantage and crime or victimizations of Hispanic residents. The white and other groups arrest rate of 1.31 is substantially lower after controlling for regions of the city and year, though there is very little association with concentrated disadvantage. Figures 4-6 examine the association between arrest rates for blacks, Hispanics, and whites or other groups by level of concentrated disadvantage and reported crime. Each figure displays the marginal effects (expected rate) across twenty quantiles (0-5…95-100 percentile), controlling for all other measures specified in column (4) of Table 11. Figure 4a shows that in New York City the higher black and Hispanic arrest rates association with concentrated disadvantage is driven by the upper 75 percentile. By comparison, there is no association between the arrest rate of whites and others and the level of concentrated disadvantage. Figure 4b shows that the nonlinear relationship between levels of crime and arrests, and that the arrest rates are driven by the 95 percentile of census block groups. While the general increase is similar for all groups, the base rate is substantially higher for the black and Hispanic arrest rate, implying that the higher arrest rate in the highest crime areas has a larger population level impact on black and Hispanic arrest disparities. 35 Figure 6a shows that for Los Angeles the association with concentrated poverty and a higher Hispanic arrest rate is driven by the upper 30 percentile. There is effectively no association with concentrated disadvantage and variation in the arrest rate for blacks or whites and other groups after controlling for crime. Figure 6b shows that the 95 percentile has a disproportionate relationship with higher arrest rates for all race and ethnic groups. The relative disparity in rates for Hispanics relative to others in Los Angeles is very comparable to black arrest rate disparities in Chicago. 36 The results from the regression estimates and figures reveal a few key takeaways. First, the racial and ethnic disparities in arrest rates are strongly associate differences between areas in concentrated disadvantage and crime. Crime rates are more associated with racial disparities in 37 arrest rates than concentrated disadvantage. Census block groups in New York, Chicago, and Los Angeles with the highest rates of crime are substantial contributors to the population level disparities in arrest rates between blacks and Hispanics relative to whites and other groups. Second, while city level differences in black and white poverty appear to have little association with racial disparities in arrest rates, the hyper-concentration of crime and poverty within in America’s three largest cities appears to play a substantial role in explaining racial disparities in the arrest rates. Third, while the causes of what drives the role of place in shaping disparities is beyond this scope of this study, the results imply that place-based disparities are important population drivers of official contacts between the police that results in racial disparities in arrests. Population level disparities in arrest rates for blacks could be cut by 30 percent in New York (2.05 to 1.43), 25 percent in Chicago (1.20 to .894), and 29 percent in Los Angeles (1.55 to 1.10) by moving the 95th percentile of census blocks from their observed arrest rate to the 50th percentile.14 While the evidence presented here suggests that the variation in the level of concentrated disadvantage and crime in New York, Chicago, and Los Angeles are important for understanding population level racial disparities in arrests, there are important limitations to acknowledge. These analyses of cities do not account for repeat arrests of the same individuals in estimating racial disparities in arrest rates by census block groups. Given that criminal behavior is highly concentrated among a subset of the population, it is likely that within the highest crime places there are a subset of individuals generating a disproportionate share of the arrests. Accounting for repeat arrests among the same individuals could be another source for population level racial disparities. Additionally, the arrest data were not disaggregated into seriousness of crime categories or other relevant contextual information that may influence arrest decisions. Research shows that arrests are more likely when a crime is more serious, suspects have a criminal history, the demeanor of suspects, the desires of victims, and whether a witness is present (Kochel, Wilson, and Mastrofski, 2011). Disaggregating arrest counts by the seriousness of the offense and relevant contextual information may influence the size and direction of population level racial disparities in arrests. The analyses of New York, Chicago, and Los Angeles also does not 14 In New York City, Chicago, and Los Angeles this represents an area of approximately 275 block groups (population of roughly 470,000 residents), 100 block groups (population of roughly 163,000 residents), and 112 block groups (population of roughly 228,000) respectively. 40 to communities. Place-based approaches to addressing public safety offer some guide for how to improve the well-being of communities and reduce racial disparities in police contact. References Ba, Bocar A., Dean Knox, Jonathan Mummolo, and Roman Rivera. "The role of officer race and gender in police-civilian interactions in Chicago." Science 371, no. 6530 (2021): 696-702. Beck, Allen J. Race and Ethnicity of Violent Crime Offenders and Arrestees. U.S. Department of Justice Office of Justice Programs Bureau of Justice Statistics January 2021 Statistical Brief NCJ 255969, 2018. Braga, Anthony A., and Brenda J. Bond. "Policing crime and disorder hot spots: A randomized controlled trial." Criminology 46, no. 3 (2008): 577-607. Braga, Anthony Allan, and David Weisburd. Policing problem places: Crime hot spots and effective prevention. Oxford University Press on Demand, 2010. Branas, Charles C., Eugenia South, Michelle C. Kondo, Bernadette C. Hohl, Philippe Bourgois, Douglas J. Wiebe, and John M. MacDonald. "Citywide cluster randomized trial to restore blighted vacant land and its effects on violence, crime, and fear." Proceedings of the National Academy of Sciences 115, no. 12 (2018): 2946-2951. Chalfin, Aaron, Benjamin Hansen, Emily K. Weisburst, and Morgan C. Williams. Police Force Size and Civilian Race. No. w28202. National Bureau of Economic Research, 2020. Chalfin, Aaron, Benjamin Hansen, Jason Lerner, and Lucie Parker. "Reducing crime through environmental design: Evidence from a randomized experiment of street lighting in New York City." Journal of Quantitative Criminology (2021): 1-31. Chetty, Raj, Nathaniel Hendren, Patrick Kline, and Emmanuel Saez. "Where is the land of opportunity? The geography of intergenerational mobility in the United States." The Quarterly Journal of Economics 129, no. 4 (2014): 1553-1623. Clogg, Clifford C., Eva Petkova, and Adamantios Haritou. "Statistical methods for comparing regression coefficients between models." American journal of sociology 100, no. 5 (1995): 1261- 1293. Crutchfield, Robert D. "From slavery to social class to disadvantage: An intellectual history of the use of class to explain racial differences in criminal involvement." Crime and Justice 44, no. 1 (2015): 1-47. 41 Fagan, Jeffrey, and Garth Davies. "Street stops and broken windows: Terry, race, and disorder in New York City." Fordham Urb. LJ 28 (2000): 457. Fagan, Jeffrey A., Amanda Geller, Garth Davies, and Valerie West. "Street stops and broken windows revisited." In Race, ethnicity, and policing, pp. 309-348. New York University Press, 2010. Fagan, Jeffrey, Anthony A. Braga, Rod K. Brunson, and April Pattavina. "Stops and stares: Street stops, surveillance, and race in the new policing." Fordham Urb. LJ 43 (2016): 539. Fryer Jr, Roland G. "An empirical analysis of racial differences in police use of force." Journal of Political Economy 127, no. 3 (2019): 1210-1261. Gelman, Andrew, Jeffrey Fagan, and Alex Kiss. "An analysis of the New York City police department's “stop-and-frisk” policy in the context of claims of racial bias." Journal of the American statistical association 102, no. 479 (2007): 813-823. Goel, Sharad, Justin M. Rao, and Ravi Shroff. "Precinct or prejudice? Understanding racial disparities in New York City’s stop-and-frisk policy." The Annals of Applied Statistics 10, no. 1 (2016): 365-394. Grunwald, Ben, and Jeffrey Fagan. "The end of intuition-based high-crime areas." Calif. L. Rev. 107 (2019): 345. Jean, Peter KB St. Pockets of crime. University of Chicago Press, 2008. Kirk, David S. "The neighborhood context of racial and ethnic disparities in arrest." Demography 45, no. 1 (2008): 55-77. Klinger, David A. "Negotiating order in patrol work: An ecological theory of police response to deviance." Criminology 35, no. 2 (1997): 277-306. Klinger, David, Richard Rosenfeld, Daniel Isom, and Michael Deckard. "Race, crime, and the micro‐ecology of deadly force." Criminology & Public Policy 15, no. 1 (2016): 193-222. Kochel, Tammy Rinehart, David B. Wilson, and Stephen D. Mastrofski. “Effect of suspect race on officers’ arrest decisions.” Criminology 49, no. 2 (2011): 473-512. Krivo, Lauren J., and Ruth D. Peterson. "Extremely disadvantaged neighborhoods and urban crime." Social Forces 75, no. 2 (1996): 619-648. Krivo, Lauren J., Ruth D. Peterson, and Danielle C. Kuhl. "Segregation, racial structure, and neighborhood violent crime." American journal of Sociology 114, no. 6 (2009): 1765-1802. 42 Legewie, Joscha, and Jeffrey Fagan. "Group threat, police officer diversity and the deadly use of police force." Columbia Public Law Research Paper 14-512 (2016). MacDonald, John, and Anthony A. Braga. "Did post-Floyd et al. reforms reduce racial disparities in NYPD stop, question, and frisk practices? An exploratory analysis using external and internal benchmarks." Justice Quarterly 36, no. 5 (2019): 954-983. MacDonald, John, Charles Branas, and Robert Stokes. Changing Places. Princeton University Press, 2019. MacDonald, John, Jeffrey Fagan, and Amanda Geller. "The effects of local police surges on crime and arrests in New York City." PLoS one 11, no. 6 (2016): e0157223. MacDonald, J. M., & Fagan, J. (2019, May). Using shifts in deployment and operations to test for racial bias in police stops. In AEA Papers and Proceedings (Vol. 109, pp. 148-51). MacDonald, John, and Steven Raphael. An Analysis of Racial/Ethnic Disparities in Stops by Los Angeles County Sheriff’s Deputies in the Antelope Valley. Working Paper: https://lasd.org/wp- content/uploads/2020/10/AVDOJ_Analysis_of_LASD_Stops_in_the_AV_January- July_2019.pdf (2020). MacDonald, John, Viet Nguyen, Shane T. Jensen, and Charles C. Branas. "Reducing crime by remediating vacant lots: the moderating effect of nearby land uses." Journal of Experimental Criminology (2021): 1-26. Massey, Douglas, and Nancy A. Denton. American apartheid: Segregation and the making of the underclass. Harvard university press, 1993. Neil, Roland, and Christopher Winship. "Methodological challenges and opportunities in testing for racial discrimination in policing." Annual Review of Criminology 2 (2019): 73-98. Parker, Karen F., and Patricia L. McCall. "Structural conditions and racial homicide patterns: A look at the multiple disadvantages in urban areas." Criminology 37, no. 3 (1999): 447-478. Parker, Karen F., Brian J. Stults, and Stephen K. Rice. "Racial threat, concentrated disadvantage and social control: Considering the macro‐level sources of variation in arrests." Criminology 43, no. 4 (2005): 1111-1134. Ridgeway, Greg, and John MacDonald. "Methods for assessing racially biased policing." In Race, ethnicity, and policing, pp. 180-204. New York University Press, 2010. Sampson, Robert J. "Effects of socioeconomic context on official reaction to juvenile delinquency." American Sociological Review (1986): 876-885.