Gentrification, White Encroachment, and the Policing of Black Residents in Washington, DC

When middle-class and White residents move into working-class and poor Black neighborhoods, are there increases in the frequency of arrests of Black people? There are a handful of published articles that examine quantitatively the relationship between gentrification and policing. These studies focus almost exclusively on Los Angeles and New York City and the focus on racialized policing is limited. The present study considers racialized policing in a city that was, until very recently, majority Black and explores the extent to which gentrification and racial change in Washington, DC are associated with enhanced policing of Black residents. A spatial regression analysis which models the association between gentrification, White encroachment, and the policing of Black residents using arrest data from the D.C. Metropolitan Police Department and a gentrification score based on American Community Survey data reveals clear racialized and spatial disparities in arrest rates in Washington, DC. We hypothesized that census tracts experiencing gentrification and White encroachment would have higher drug arrest rates of Black residents. We found support for our White encroachment hypothesis but not for our gentrification hypothesis.


Introduction
In 2018, the media reported 92 instances of White people calling the police on Black people for sitting in Starbucks, eating, sleeping, smoking cigarettes, going for walks, barbecuing, playing basketball, and a host of other everyday undertakings (McNamarah 2018)leading journalist Harriot (2018) to write an article for The Root titled "white caller crime" to describe the propensity of White people to call the police on Black people. residents over the age of 25 with a college degree (Smith 1987;Freeman 2005;Martin and Beck 2018;Owens 2012).

Racial Change and Policing
Studies of postindustrial policing are based on the idea that, as high-income residents move into a low-income neighborhood, policing tactics will change. This raises the question of whether policing tactics change as White residents move into primarily Black neighborhoods. Research in the racial threat tradition leads us to hypothesize that there would be an association between the arrival of White residents and increased policing of Black residents.
Understanding racialized patterns of arrest requires bearing in mind that arrest patterns are not solely a function of the level of crime but also of other factors, including police organizational practices (Eitle and Monahan 2009;Beckett et al. 2006); workgroup norms (Herbert 1997); and social and economic trends (Parker and Maggard 2005;Beck 2019;Donnelly et al. 2019). When seeking to understand variations in arrest rates, scholars often draw from conflict theory, which approaches racial disparities in arrest rates from the perspective that disparities in law enforcement activity are due in part to the choices officers make in terms of how to carry out their duties (Eitle and Monahan 2009). Racial threat theory is part of the conflict theory tradition and views policing as a tool White people and White power structures use to control a Black population they perceive as threatening (Eitle and Monahan 2009;Beck 2019;Engel et al. 2012).
Scholars in the racial threat tradition have drawn from patterns such as the propensity of police officers to over-police Black communities and the anti-Black attitudes of police officers to explain racial disparities in arrest patterns (Gaston 2019a;Donnelly et al. 2019). This perspective is summed up by Gaston (2019a: 426): The "unifying premise" of this body of theories "acknowledges a race-based power structure in which Whites mobilize social control when they perceive threats to their political, economic, or social dominance from nondominant groups." Translated to policing, this "unifying premise" suggests that when Whites mobilize the police against Black residents as part of a racial threat response, we can expect increases in drug arrests to occur because of the great deal of discretion officers are given to make such arrests.
Although Black and White people in the USA use and sell drugs at similar rates, Black people are more likely to be arrested on drug charges (e.g., Mitchell and Caudy 2015). These racial disparities in drug arrest rates are exacerbated in racially diverse neighborhoods. A study of racial disparities in drug arrests in Washington, DC revealed that racial disparities were highest in predominantly White areas (Fielding-Miller et al. 2016). These findings have been replicated in Saint Louis (Gaston 2019a); in Donnelly et al. (2019); and in a nationwide study where Beck (2019) found that suburbs with fewer Black people had greater racial disparities in arrests. Scholars have attributed these racial disparities to the choice of police to focus on drug sales in outdoor venues; the higher presence of buy/bust tactics in Black neighborhoods; and police officers' greater propensity to use discretionary stops on Black residents (Gaston 2019a;Beckett et al. 2006).
Most studies focus on racial disparities in drug arrests and on how neighborhood racial composition affects these patterns. There are fewer studies of how change in racial composition affects arrest rates. In a study using 2001 arrest data from Washington, DC, Kane et al. (2013) found that historically White census tracts with increases in the percentage of Black people in the population experienced concomitant increases in Black misdemeanor arrests. This study poses the converse question: When White people move into historically Black census tracts, are there increases in arrests for Black people? A recent quantitative study of New York City approaches this question and finds an increase in order-maintenance arrests (such as disorderly conduct) in neighborhoods when White people move in, but fewer stops and proactive arrests. Proactive arrests are driven by police activity rather than citizen complaints and include arrests on charges such as drug and weapon possession (Beck 2020). Another study of Seattle, Washington, found that racial disparities in arrest patterns are exacerbated in neighborhoods experiencing an influx of wealthier, Whiter residents (Lanfear et al. 2018).
In the year 2000, Washington, DC was a highly segregated, majority Black city. Seventy-two of the 188 census tracts in the city were over 90% Black. In the twenty-first century, however, the city began to change and White people began to move into neighborhoods that had been nearly all Black for decades. When White people move into historically Black neighborhoods, this could be associated with higher levels of arrests of Black residents for two reasons: (1) White residents may be more likely to call the police to address problems they experience in the neighborhood; and (2) police officers may be more likely to patrol neighborhoods when White people move into them. This paper thus explores the extent to which White encroachment in majority Black neighborhoods is associated with racialized policing practices.
We measure White encroachment by the percent change of White residents in each census tract between 2000 and 2015. We focus primarily on White and Black residents because D.C. has long been a city where the Black/White divide is the most significant and enduring-in terms of both economic inequality and spatial segregation. Although the Latinx population increased from 8% in 2000 to 11% in 2017, and the Asian population from 3 to 4%, the Black and White populations continue to constitute a significant majority (84%) of residents-with the Black population decreasing from 60 to 47% and the White population increasing from 28 to 37%. Whereas the 2000 Census revealed that 47 census tracts were majority White and 125 were majority Black in 2000 in DC, only one tract was majority Latinx, and only ten tracts (out of 188) had more than 1000 Latinx residents. There is an important story to be told about the policing of Latinx residents in DC, but this paper focuses primarily on the arrests of Black residents, who accounted for 85% of the total arrests in the period under study.
The Metropolitan Police Department (MPD) has 3,796 sworn officers. The majority of these officers are on patrol services and handle drug complaints. Drug-related investigations are handled by the 164 officers in the Narcotics and Special Investigations Unit, which provides proactive law enforcement aimed at reducing drug dealing and trafficking. According to a report produced by the National Police Foundation, nearly all of the targets of NSID actions are Black men: Between August 1, 2019, and January 31, 2020, 88% of the 3,680 people stopped by NSID were Black and 83% were men. Of these 3,680 people, 2,180 were arrested-91% of whom were Black, and 85% of whom were men (MPD Oversight Hearing 2020; National Police Foundation 2020).

White Encroachment and Gentrification in Washington, DC
In 1970, the Black population of Washington, DC reached a peak of 70% of the city's population and then began a steady decline. The city's population also shrunk beginning in 1970, reaching a nadir of 572,000 in 2000, before swinging up again to nearly 700,000 in 2017. With this upswing, DC also lost its Black majority-by 2017 only 47% of its residents were Black. Meanwhile, the White population increased from 28% in 2000 to 37% in 2017. The Asian and Latino populations also both doubled during this period, reaching 10.4% for Latinos and 3.7% for Asians in 2014 (Asch and Musgrove 2016;Prince 2014).
The overall trend of population decline shifted in 2006, when 900 more households moved into the city than moved out. This trend accelerated, and between 2009 and 2010, nearly 5,000 more households moved in than moved out (Sturtevant 2014). Between 2000 and 2010, the Black population declined by 38,000 while the White population increased by 45,000. The racial composition of this recent population gain was unique among US cities-DC and Atlanta are the only two cities where the growth in the city's White population was greater than the overall population change (Sturtevant 2014). These in-migrants were also significantly more well educated than residents who did not move or who moved out (Sturtevant 2014).
Recent qualitative studies of gentrification in Washington, DC make it clear that the city has changed, and that gentrification is a racialized process (Hyra 2017;Prince 2014;Asch and Musgrove 2016;Howell 2016;Gallagher 2016;Maher 2015;Summers 2019). Nevertheless, change has not occurred evenly across the city. There are many neighborhoods west of Rock Creek Park that have remained majority White and affluent. There are many areas east of the Anacostia River that have remained majority Black and low-income. And, even where changes have occurred, these changes have been uneven. Several neighborhoods along 16th Street and Georgia Avenue, for example, have been both majority Black and middle class since the 1960s. These neighborhoods have experienced an influx of White residents, but the socioeconomic status of the neighborhood has not changed. There are other neighborhoods such as Capitol View where housing projects have been demolished and replaced with mixed-income housing where the average home values and incomes have increased, yet these neighborhoods have remained over 90 percent Black. Thus, although most previous research in DC has conflated racial and class change, it is imperative to distinguish between these two processes. This article thus considers racial change separately from gentrification. We also consider the racial composition in 2000 because an increase in White residents in a neighborhood that is already majority White is different from the arrival of White residents into a majority Black neighborhood.
In some DC neighborhoods, such as Shaw, White in-movers are more affluent than the long-term Black residents. In other DC neighborhoods, such as Shepherd Park, White inmovers have about the same income as long-term Black residents. Given this context, we control for White/Black income competition, as previous studies have found that there are more arrests of Black people in neighborhoods where White residents feel economically threatened by Black residents, i.e., where there is less disparity between their incomes (Gaston et al 2020).

Analytical Strategy and Hypotheses
The present study uses police drug arrests as its dependent variable primarily because drug arrests represent a highly discretionary form of police coercion. Police departments choose where and how vigorously to engage in drug enforcement across space and time (e.g., Engel et al. 2012;Gaston 2019a). Moreover, drug arrest behaviors appear susceptible to racialized policing processes, as police departments have been shown to focus their drug enforcement arrest efforts primarily on communities and individuals of color (Beckett et al. 2005(Beckett et al. , 2006Eitle and Monahan 2009;Mitchell and Lynch 2011;Gaston 2019a). Collectively, these factors represent both the advantages and disadvantages of using drug arrests as our outcome measure: On the one hand, the discretionary nature of drug enforcement means that police departments and officers are relatively free to respond to real or perceived changes in local ecological conditions that make it seem as though more or less drug enforcement is justified. On the other hand, since drug enforcement has been identified as a "threat" response in the minority group-threat literature (e.g., Eitle and Monahan 2009;Mitchell and Lynch 2011;Gaston 2019a), it is not yet known if drug enforcement levels in historically Black communities are already so high as to not be responsive to gentrification processes hypothesized to lead to more drug enforcement. Still, given the discretionary nature of drug policing, as well as its historic sensitivity to neighborhood racial composition, we view it as a preferred and highly valued indicator of formal social control for the purposes of this study. We focus on the arrests of Black people because of our theoretical focus on racial threat theory and because of the fact that 85% of arrests in DC in the period under study were of Black people.
Hypothesis 1: Census tracts eligible to gentrify that have experienced gentrification will have higher arrest rates of Black residents on drug charges, net of controls.
Hypothesis 2: Historically Black census tracts that have experienced an influx of White residents will have higher arrest rates of Black residents on drug charges, net of controls.

Data and Variables
To measure drug arrests of Black residents in Washington DC, we first collected each offense charge from the D.C. Metropolitan Police Department. The department provides geocoded information on each crime and arrest incident by location, charge categories and by defendants' age, gender, and race/ethnicity. The information is publicly available through the DC Crime Card Application webpage. 1 We only included drug arrests where arrestees were eighteen years or older. Since the incidents of arrest vary across years, for rigorous measurement of each tract's arrests, we created five-year averages between 2013 and 2017. In addition to Black arrests, for comparison on the mechanism of gentrification and White encroachment on Black versus non-Black arrests, we also created five-year averages of tract-level non-Black arrests from 2013 to 2017. Since less than 11% of arrests in the District of Columbia from 2013 to 2017 were of non-Black residents (an average of 16.7 Black drug arrests per census tract versus 2.0 for non-Black drug arrests), we aggregated all non-Black arrests together. Figure 1 presents the 5-year averages of drug arrests by arrestee race across the 179 census tracts of the District of Columbia. It shows that there were fewer incidents of Black drug arrests in the upper northwest area of D.C. In the central and the eastern parts of the city, there were widespread Black arrests on drug offense charges. In light of the concentration of Black residents on the eastern side of the city and White residents on the western side, we use a rate variable as opposed to a count. Our dependent variables thus are: (1) the rate of Black arrests per Black population; and (2) the rate of non-Black drug arrests per non-Black population.
To measure gentrification, we focused on four sociodemographic indicators of neighborhood changes between 2000 and 2015: changes in the median home value, the median household income, and the median rent, and the percentage point change in residents over the age of 25 with a college degree (Smith 1987;Freeman 2005;Martin and Beck 2018;Owens 2012). Using "change scores" is an appropriate way for capturing these patterns (Allison 1990). For the change in median home value, household income, and rent, after adjusting in constant 2019 dollars reflecting the consumer price index, we calculated the percent changes in each tract from 2000 and 2015. And similarly, we calculated the percent point change of residents over the age or 25 with a college degree in each tract from 2000 and 2015. We then created an unweighted additive index of gentrification using those change scores (α = .796). In terms of neighborhood racial change, we accounted for White encroachment by estimating the percent point change of White residents in each tract between 2000 and 2015. For those variables, census tract-level data came from the 2010 Decennial Census 2 and the 2015 American Community Survey: 5-Year Data, provided by the IPUMS National Historical Geographic Information System (NHGIS). 3 Since the Beck (2020: 252) writes that it is important for gentrification research "not to compare tracts that were eligible for gentrification to those that were already gentrified or too wealthy to be gentrified." Following this logic, we coded a tract as already gentrified or ineligible for gentrification if, at the start of the given period, the tract's median household income was above the citywide median of $54,600 in 2000. There were 99 census tracts that were already gentrified or too wealthy to be gentrified and 80 tracts that were eligible to be gentrified in 2000 (see Appendix 1). And, in terms of the racial change variable, we coded 130 tracts as "historically Black" as they were more than 50% Black in 2000. The remaining 49 tracts were coded as "historically White." The city was highly segregated in 2000 and had small numbers of residents that were neither Black nor White. Thus, all tracts could easily be coded as historically White or Black. The high level of segregation is made clear in Appendix 1 (b) which shows that the entire western side of the city is historically White, and the eastern side is historically Black.
In addition, we controlled for a series of neighborhood conditions that other researchers have found to be related to arrests. First, population density was calculated using area of the land and total population of the tracts. We also accounted for the percentage of male and the percentage of young residents aged 18 to 34. These populations not only show the gender and age composition of the tracts, but also accounts for who is most likely to be arrested-young men (Snyder 2012;Donnelly et al. 2019;Beck 2020). An index for structural disadvantage for each tract is included, based on the social and economic conditions of disadvantage in a neighborhood (Parker and Maggard 2005;Kane et al. 2011;Donnelly et al. 2019). Based on Kane and colleagues' (2011) measurement, we constructed the structural disadvantage score by adding the following variables without weighing: residents below the poverty line, unemployed for the population 16 years and over, residents with no high school diplomat for the population 25 years and over, and the percentages of households on public assistance income and female householder with children under 18 years (α = .824). In addition, Gaston et al. (2020) argue that economic inequality between racial groups, as part of the racial context of a neighborhood, impacts racial conflicts that are connected to racialized policing. To capture the tract-level economic competition between racial groups, we included a measure for per capita income ratio between White and Black residents. The citywide mean of the White-Black per capita income ratio in 2015 is 2.1 with a standard deviation of 1.3. To account for residential stability, we included a variable for the percentage of households staying in their current residency for more than five years. The percentage of vacant housing units accounts for the housing situation of a census tract. These geocoded variables came from the 2015 American Community Survey: 5-year data (2011)(2012)(2013)(2014)(2015), provided by the IPUMS NHGIS (Manson et al. 2022).
Lastly, to examine the relationship between arrest charges, crime incidents, and demand for low-level policing in a neighborhood, we added variables for 5-year averages of violent crime incidents and the number of nonemergency 311 calls. From the DC Crime Card Application webpage, we gathered all crime incidents and nonemergency 311 calls between 2013 and 2017 and calculated the 5-year averages violent crime incidents and 311 calls for each census tract.
We also include a measure of 311 calls in the models for two distinct-and to some extent, competing-reasons. First, at the aggregate level, 311 calls can represent a community's attempt to mobilize municipal government on its behalf to address problems of physical disorder. In this way, an aggregate measure of 311 calls can indicate "custodianship" (O'Brien 2015: 304), which is not precisely collective efficacy but rather an indicator of the kinds of physical conditions community members will not tolerate in their public spaces. To the extent that custodianship empowers communities to either call the police or send signals to drug users that drug use behaviors are not tolerated, we might expect the aggregate measure of 311 calls to be positively related to drug arrests. Similarly, 311 calls may also represent actual physical disorder on the street block (e.g., O'Brien et al. 2015;Wheeler 2018), which may lead to more arrests due to more crime as predicted by broken windows processes (e.g., St. Jean 2008;Wheeler 2018;Golash-Boza et al. 2022).
Conversely, a relative absence of 311 calls in certain communities can also suggest that some local residents-particularly those residing in structurally disadvantaged communities-do not trust the local government to respond to their needs as related to the physical conditions on the street block. Such underreporting may create "systematic underrepresentation" (Theall et al. 2021: 263) of communities experiencing some of the highest levels of physical disorder on the street block (e.g., Crawford 2013)-e.g., the very communities that also experience drug arrests. In this case, we could expect an inverse relationship between aggregate 311 calls and drug arrests, but for the opposite reason of custodianship. This relationship would likely be explained by something analogous to "legal cynicism." (Sampson and Bartusch 1998). We make no a priori prediction of the potential relationship between 311 calls and drug arrests; we simply presume that based on prior theorizing and empirical study, 311 calls may be important community-level covariates of drug arrests and thus control for 311 calls in our analyses. Table 1 reports descriptive statistics of the tract-level variables for Black arrests, gentrification, White encroachment, and other variables for neighborhood conditions. Among covariates, we did not find any pair of correlations that potentially increases the risk of multicollinearity in our multivariate analyses (see Appendix 2). 4

Analytic Strategy
We used spatial regression analysis for modeling the association between neighborhood gentrification, White encroachment, and the policing of Black residents. As Fig. 1 shows, the distribution of Black residents' drug arrests in D.C. is related to geographical locations-there were fewer incidents in the northwest versus more in the central and southeast areas. And the Moran's I test results indicate that there is a positive spatial autocorrelation in the dependent variables, indicating that drug arrests in DC census tract are spatially clustered and influenced by neighboring tracts (for Black drug arrests, I = .034, p < .000, for non-Black drug arrests, I = .003, p < .000).
Adding a spatial lag, the method enables us to assess the impacts of independent variables on the dependent variable by considering the spatial dependence between units. The general spatial regression model is given by: where X is matrix of observations on the covariates, W and M are spatially weighted matrices on the distance between observations, is a spatially correlated residual, ε is identically distributed disturbances, λ and ρ are scalar parameters showing the dependence of y n on its nearby y , and the spatial correlation in the errors, respectively (Cliff and Ord 1973;Ord 1975;Kelejian and Prucha 2010;Drukker et al. 2013). In our full dataset of 179 DC census tracts, we have 11.2% (20 out of 179) of missingness in the gentrification score and 9.5% (17 out of 179) for the White encroachment variable. In a spatial regression analysis, missing values make models inestimable, because the outcome for one unit relies on outcomes for other units. To deal with missingness, listwise deletion is not appropriate, because it does not only drop such missing observations, but also eradicates the spatial dependence between units (Wang and Lee 2013). Instead, we replaced those missing observations with multiple imputation. Considering the uncertainty around missing information, multiple imputation combines the parameter estimates from multiple draws of the simulated values using a statistical method of interest (herein, linear regression), along with other covariates (Wang and Lee 2013).
We then created two separate contiguity weight matrices along with our two sets of dependent variables-gentrification or White encroachment-and their eligibility criteria-eligible/ineligible to gentrify or historically Black/non-Black-based on the locations of the neighborhoods and the distances among them. Herein, we assumed that census tracts will be considered as neighbors only if they share a common boundary. Using spatial regression models with those matrices, we first examined the effects of gentrification on drug arrests by arrestee race. In doing so, spatial regression models are separated by eligibility: eligible to be gentrified or already gentrified. Then, we ran models for predicting the impacts of White encroachment on the policing practice by arrestee race. We ran models for historically Black census tracts and for non-Black census tracts, respectively. Stata (version 13) was used for our statistical analyses.

Findings
To test our first hypothesis on the relationship between gentrification and Black arrests, Table 2 presents the results from the spatial regression models assessing the relationship between gentrification and drug arrests. The first two models regress the drug arrest rate of Black residents on the gentrification score. To adjust for racial differences in exposure to arrest, we used the rate of Black drug arrests per Black residents as the outcome variables. For those models, we did not find evidence of a positive association between gentrification and drug arrests.
Model 1 includes 80 census tracts that were eligible to be gentrified-those tracts that reported median household incomes below the citywide median of $54,600 in 2000. It reveals that all other things being constant, an increase in the gentrification score yields a decrease in incidents of drug arrests of Black residents (b = − .016). However, the association is not statistically significant. Among neighborhood conditions in 2015, residential stability (b = − .816, p < .05) yields a decrease in the policing of Black residents. On the other hand, increases in the percentages of young residents (b = .933, p < .05), the incidents of crime (b = .376, p < .05) and nonemergency 311 calls (b = .035, p < .001) have a positive influence on increasing Black residents' drug arrests.
Model 2 examines the association between gentrification and Black residents' drug arrest rates in 99 census tracts that reported median household incomes over the citywide median of $54,600 in 2000. For affluent neighborhoods, we found that an increase in the gentrification score is associated with 0.23% decrease in the Black drug arrest rate (b = − .105, p < .05). 5 What this is showing is that increases in socioeconomic indicators in neighborhoods that are already affluent are associated with fewer Table 2 Results from the spatial regression models predicting impacts of gentrification on drug arrests in DC neighborhoods, arrestee race and eligibility criteria *p < .05; **p < .01; ***p < .001 a Missing values were replaced by multiple imputation. Standard error in parentheses arrests of Black residents on drug charges. It is consistent with the prevailing literature that there would be fewer drug arrests in affluent neighborhoods, and thus not surprising that there would be even fewer arrests as neighborhoods become more affluent. Among covariates for neighborhood characteristics, we found that increases in population density (b = 3.364, p < .05), violent crimes (b = 1.860, p < .001) and nonemergency 311 calls (b = .029, p < .001) are linked to increases in arrests of Black residents on drug offense charges. For comparison, in addition to Black drug arrests, the other two models regress the rate of non-Black drug arrests per non-Black residents on the gentrification score, by eligibility criteria. For those models, we did not find a statistically significant relationship between the drug arrest rate and gentrification.
We then ran spatial regression models estimating the impacts of White encroachment on the drug arrest rate of Black residents. We predicted that historically Black census tracts that have experienced an influx of White residents will have higher arrest rates of Black residents on drug offense charges. As reported in Table 3, our spatial regression models provide evidence supporting that there is a positive association between White encroachment and Black drug arrest rates. More specifically, Model 1 regresses the rate of Black drug arrests per Black residents on the variable for White encroachment in 130 census tracts that were historically Blackthose that had more than fifty percent of Black residents in 2000. It reveals that an increase in the percentage of White residents yields a 2.86% increase in the drug arrest rate of Black residents (b = .648, p < .05). 6 Similarly, we found that increases in violent crimes (b = .831, p < .001), and nonemergency 311 calls (b = .018, p < .01) are associated with increases in Black drug arrests.
For the 49 historically non-Black tracts, we also found that an increase in the number of White residents in a neighborhood is associated with more arrests of Black residents for drug offense charges (Model 2, b = .068), but the association is not statistically significant. The model also indicates that an increase in violent crimes (b = .533, p < .001) leads to an increase in Black drug arrests. Herein, we found that some of the coefficients among covariates for neighborhood characteristics flip from positive to negative when we compare the tracts that were historically Black in 2000 to those that were non-Black.
The latter two models estimate the association between White encroachment and drug arrests of non-Black populations for historically Black (Model 3) and non-Black neighborhoods (Model 4). Both models use the rate of non-Black drug arrests per non-Black residents as the outcome variable. These results do not yield a significant relationship between racial change and the arrest rates of non-Black residents in either Black or non-Black neighborhoods-providing further support for our claim that race is important for understanding policing practices.
In sum, these findings indicated that our first hypothesis is not supported. Although we hypothesized a positive association between gentrification and Black drug arrests, results from the spatial regression models do not align with our expectations. And yet, we found that our second hypothesis on White encroachment is supported. We predicted that White encroachment would yield increases in the arrest rate of Black residents on drug offense charges in historically Black neighborhoods. Our findings from the spatial regression models showed that the influx of White residents is connected to more policing of Black residents.

Conclusion and Discussion
There are clear racialized and spatial disparities in drug arrest rates in Washington, DC. This paper endeavored to disentangle how neighborhood change affects these disparities. We hypothesized that census tracts experiencing gentrification and White encroachment would have higher drug arrest rates of Black residents. We found support for our White encroachment hypothesis but not for our gentrification hypothesis.
We hypothesized that as White people move into neighborhoods that were majority Black in 2000, Black people would be at higher risk for arrest on drug charges. Insofar as Black people have a long and troubled history with the police, they are less likely to trust police officers. This in turn means that people who commit crimes in predominantly Black neighborhoods are less likely to be arrested-as their neighbors are less likely to call the police on them (Kirk and Matsuda 2011). Whereas Black people are less likely to cooperate with police, White people are more likely to perceive the police as protectors of their community than as aggressors. When residents of a neighborhood perceive the law to be just, legitimate, and responsive, they are more likely to cooperate with police (e.g., [deleted self-citation] 2005)-which could lead to more arrests relative to crime. Thus, as White people move into historically Black neighborhoods, it would follow that the number of arrests of Black people would increase. And, this is what we found: increases in the percentage White in historically Black neighborhoods were associated with increases in drug arrests of Black residents, showing clearly the racialized nature of drug law enforcement.
The control variable "White/Black income competition" also was consistently negatively associated with Black arrest rates. Those tracts where the White/Black income ratio was higher had lower rates of arrest of Black residents. Gaston et al (2020: 9) argue "racial income competition manifests when White-Black economic inequality wanes and Black people are not economically subordinate to Whites." Our findings align with Gaston et al.'s (2020) argument that there will be fewer arrests of Black people in those areas where White residents do not feel economically threatened by Black residents. Our findings thus indicate a higher likelihood of arrest on drug charges of Black people in census tracts where Black people's earnings are on par with Whites. This indicates that the story is not just one of White cooperation with the police but also one of neighborhoods where there is more economic parity between White and Black residents seeing higher numbers of Black arrests.
Notably, we also found a positive relationship between the White/Black income ratio and non-Black arrests in historically Black neighborhoods as well as in neighborhoods eligible to gentrify. Those historically Black and historically low-income tracts where the White/Black income ratio was higher had higher drug arrest rates of non-Black residents. The number of non-Black arrests is much smaller than that of Black arrests, so we will not read too much into this. However, it does point to the need to further explore the nature of racialized policing, and particularly how it might affect Latinx residents.
Although we expected to find a positive association between gentrification and the policing of Black residents in gentrifying neighborhoods, we did not. What we find most interesting about this finding is its stark opposition to our findings regarding racial change. For us, this points to the need to separate out racial change from socioeconomic change in analyses of gentrification and policing. Gentrification is a socioeconomic process where middle-class people move into historically disinvested neighborhoods. This process may happen alongside racial turnover-when one racial group displaces another-but the two processes are distinct. Our findings reveal that gentrification and racial turnover have different implications for the policing of Black residents and should be analyzed separately. Our findings point very clearly to racialized policing practices, showing a positive association between increases in White residents in historically Black neighborhoods and arrests rates of Black residents on drug charges. In contrast, our findings with regard to socioeconomic change are not significant. We suggest further study to parse out these findings as well as to consider how neighborhood-level changes affect Asian, Latinx, and Native American residents.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.