Abstract
Research has increasingly moved toward a consensus that violent crime declines as neighborhoods gentrify, yet some studies find the direction of this relationship varies by type of violent crime. This finding becomes even more important when connected with recent research that finds the structural influences of gang and non-gang homicide are disparate. The current study engages with research in each of these areas by examining the relationship of gentrification with levels of total, gang, and non-gang homicide in Los Angeles Police Department’s (LAPD) Hollenbeck Community Policing Area. We find gentrification was not associated with variation in total or gang homicide, but was positively associated with non-gang homicide.
Introduction
Cities across the United States continue to look to urban revitalization strategies such as gentrification to address the social ills of disadvantaged neighborhoods despite continued debate about the implications of such strategies (Boggess & Hipp, 2016; C. M. Smith, 2014). As defined by N. Smith (1998), gentrification refers to “the process by which central urban neighborhoods that have undergone disinvestment and economic decline experience a reversal, reinvestment, and the in-migration of a relatively well-off middle- and upper-class population” (p. 198). The research on gentrification and crime partially supports arguments that gentrification helps to address urban social ills. Specifically, research has increasingly found gentrification was associated with long-term declines in violent crime, but there was also evidence that the relationship of gentrification with violent crime varied by crime type (Barton, 2016b; Papachristos, Smith, Scherer, & Fugiero, 2011). Much of this research examined trends in broader definitions of crime (i.e., homicide, robbery, and assault), but C. M. Smith (2014) extended research on the association of gentrification with crime to gang homicide. C. M. Smith (2014) found private investment gentrification (proliferation of coffee shops) was negatively associated with gang homicide, but state-based gentrification (demolition of public housing) was positively associated with gang homicide. C. M. Smith’s (2014) study remains the only one to empirically assess the relationship of gentrification with gang crime despite recent efforts to better understand the temporal and spatial distribution of gang crimes (e.g., see Costanza & Helms, 2012; Tita & Greenbaum, 2009; Tita & Radil, 2011; Valasik, 2014; Valasik, Barton, Reid, & Tita, 2017).
Much of the literature on gang homicide focused on ascertaining the disparities of micro-level characteristics between gang and non-gang homicides, such as features of the participants or incident locations (e.g., Maxson, Gordon, & Klein, 1985). Other studies such as that of Pyrooz (2012) examined structural covariates with gang homicide in large U.S. cities. Given the localized nature of gangs and gang violence, macro-level analyses potentially conceal “sub-area and neighborhood cycles . . . that cancel each other out in the aggregate” (Klein, 1995, p. 223). Valasik and colleagues (2017) sought to bridge this gap by examining patterns in gang homicide at the neighborhood level in the Los Angeles Police Department’s (LAPD) Hollenbeck Community Policing Area (hereafter Hollenbeck) between 1980 and 2010. Their findings highlighted the protracted nature of gang violence as being over-represented and geographically clustered in socioeconomically disadvantaged neighborhoods (Valasik et al., 2017). This study provided a much-needed step forward, but the authors were unable to explain why neighborhood social structural features were associated with variation in gang violence over time. Gentrification has been frequently suggested as a potential causal mechanism for the broader changes in violence in cities (see Barton, 2016b; Kirk & Laub, 2010; C. M. Smith, 2014). Yet, research examining the relationship of gentrification with variation in gang homicide has only been explored one other time in the literature (see C. M. Smith, 2014), thereby limiting our understanding of the impacts of such policies on gang-related violence.
The current study builds on research investigating the association of gentrification with crime and research on neighborhood correlates of gang homicide by examining the association of gentrification with tract-level variation in total homicide, gang homicide, and non-gang homicide in the Hollenbeck area of Los Angeles during the 1980s, 1990s, and 2000s. Hollenbeck is unique from other gentrifying areas in Los Angeles or in other large cities because it experienced gentrification (Ahrens, 2015; Loukaitou-Sideris, Gonzalez, & Ong, 2017) while continuing to struggle with a history of entrenched gang violence (Brantingham, Tita, Short, & Reid, 2012; Moore, 1991; Valasik et al., 2017; Vigil, 2007). Our results echo Kreager, Lyons, and Hays’s (2011) argument that efforts by urban policy makers to combat entrenched problems such as neighborhood violence by encouraging gentrification may be limited at best or misguided at worst.
Gentrification and Crime
Gentrification has often been heralded as the reversal of urban decline because gentrifying neighborhoods experience improvements in amenity access, city services, and reductions in crime due to the increased representation of middle-class residents (Balzarini & Shlay, 2016; Barton, 2016a). Critics of gentrification argue that the process may encourage class or race-based tensions because the process introduces more affluent White households into lower class minority neighborhoods. Such tensions were often associated with the fear of residential displacement (Balzarini & Shlay, 2016; Brown-Saracino, 2017). This fear may deter incumbent residents from building relationships with gentrifiers, which is important because multiple studies found that the influence of changes in neighborhood structural characteristics on crime were mediated by collective efficacy in neighborhoods (Armstrong, Katz, & Schnebly, 2015; Sampson, 2012). We recognize the potential importance of displacement for the relationship of gentrification with crime, but are unable to engage with this debate in the present study due to restrictions of the data.
Research on gentrification and crime has typically drawn upon social disorganization/collective efficacy and routine activities (Barton & Gruner, 2016). Social disorganization/collective efficacy argued neighborhoods characterized by concentrated disadvantage, racial/ethnic heterogeneity, and low residential stability were more likely to feature higher crime because these characteristics decreased the potential for community development. Residents of such neighborhoods were less likely to use or respond to informal means of social control and, therefore, were more likely to rely on the police (Sampson, 2012). Gentrification presents a problem for this framework because it breaks up concentrations of poverty by bringing middle-class residents into disadvantaged areas, which should lead to reductions in neighborhood crime. Conversely, gentrification was also associated with increased racial heterogeneity and low residential stability, which were positively associated with crime.
The other theory frequently used in previous research was routine activities theory, which predicts crime will be more likely to occur when a suitable target (high value) lacking a capable guardian converged in space and time with a motivated offender (Cohen & Felson, 1979). Given these tenets, gentrification should be positively associated with crime because the process introduced young, middle-class professionals into disadvantaged neighborhoods populated by impoverished residents who may be resentful of neighborhood changes. Gentrifiers were more suitable targets than incumbent residents because they were more likely to possess high-value goods and were not necessarily familiar with techniques for protecting themselves in public (Anderson, 1990; Covington & Taylor, 1989). The motivation to offend among incumbent residents may be guided by wealth disparities or by resentment of gentrifiers (Freeman, 2006; Taylor & Covington, 1988). Furthermore, gentrification disrupted local guardianship as new residents were less familiar with the neighborhood, and incumbent residents may be less willing to act as capable guardians due to resentment of the gentrification process (Anderson, 1990; Freeman, 2006; Taylor & Covington, 1988). Alternatively, gentrification may be associated with increased guardianship as gentrifiers become more vigilant and responsive, in turn crime would be expected to decline (Kreager et al., 2011).
Similarly, neighborhood-level explanations were regularly used to understand the relationship between neighborhood characteristics with gang homicide (Kirk & Laub, 2010). Given that “a differential causal process operating in the social production of criminal homicide” is taking place, it is important to not “analyze the homicide rate as if it were a homogeneous category of lethal violence” (Williams & Flewelling, 1988, p. 422). Kubrin (2003) further pointed out disaggregating homicides was important for understanding how the social structure of community associates with different types and frequencies of homicide. Drawing upon the social disorganization framework, Curry and Spergel (1988) found gang homicides were an ecologically distinct phenomenon impacting local communities and adhered to traditional theories of poverty and social disorganization. Subsequent research confirmed neighborhood-level covariates were associated with different homicide types, including gang homicide (Kubrin & Wadsworth, 2003; Mares, 2010; Pyrooz, 2012; Valasik et al., 2017). Research also found gang violence was prevalent in or around public housing complexes (Fremon, 2008; Vigil, 2007). Furthermore, Valasik et al. (2017) found gang homicide remained clustered in disadvantaged neighborhoods over a 35-year period, suggesting that neighborhood features exert considerable influence on producing and sustaining gangs with violent tendencies.
While previous research on the association of gentrification with crime drew upon similar theoretical arguments, empirical analyses produced mixed results. Research on gentrification and crime in Baltimore during the 1970s found a positive association with assault and homicide (Taylor & Covington, 1988) and robbery (Covington & Taylor, 1989). Similarly, Lee (2010) found gentrification in Los Angeles was positively associated with assault and robbery, but not significantly associated with homicide or rape. Furthermore, analyses of gentrification with violent crime in Chicago during the 1990s and early 2000s identified a negative association with homicide and robbery (Aliprantis & Hartley, 2015; Papachristos et al., 2011), and mixed findings with gang-related homicide depending on how gentrification was measured (C. M. Smith, 2014). Together, these findings suggest gentrification may be differentially associated with specific types of violent crime. Additional reasons for the contradictory findings among these studies include differences in the operationalization of gentrification, length of time assessed, and the analysis strategy employed (Barton, 2016b; Kreager et al., 2011).
The strategy used to operationalize gentrification had implications for which areas were included in analyses (Barton, 2016a; Brown-Saracino, 2017). For example, Covington and Taylor (1989) operationalized gentrification through a single measure of property value, which treated all areas in Baltimore as gentrifiable. In contrast, the more restrictive measure used by Papachristos et al. (2011) only allowed areas where coffee shops could be built to be classified as gentrifiable. C. M. Smith (2014) built upon this definition by stating that the opening of a coffee shop represented a form of private investment gentrification while the razing of public housing complexes was a form of state-based gentrification. Most other studies used multivariable census-based strategies to identify gentrified neighborhoods (Barton, 2016a). This was primarily due to the availability of census data, but it also addressed the multiple dimensions of the gentrification process. The current study operationalized gentrification through the replication of the census-based strategy used by Loukaitou-Sideris and colleagues (2017). Just as the definition of a street gang varied across the existing literature (Curry, 2015; Esbensen, Winfree, He, & Taylor, 2001; Papachristos, 2005), there is also no agreed upon operationalization of gentrification (Barton, 2016a; Brown-Saracino, 2017). We chose Loukaitou-Sideris and colleagues (2017) strategy because it was a grounded approach that analyzed variation in four neighborhoods in Los Angeles, one of which was in Hollenbeck.
The contradictory findings of previous research were also the result of how temporal variation was controlled for in the analyses. Study periods ranged from 5 to 30 years with an average of about 10 years (Barton, 2016b; Barton & Gruner, 2016). Within the overarching study periods, previous research analyzed annual variation in crime (Lee, 2010), variation in 3-year crime averages (Papachristos et al., 2011; C. M. Smith, 2014), and variation in crime between decennial censuses (Barton, 2016b; Boggess & Hipp, 2016; Covington & Taylor, 1989; Kreager et al., 2011; McDonald, 1986). Using smaller intervals makes intuitive sense but may not be necessary because neighborhood processes such as gentrification typically occur gradually (Hipp & Wickes, 2016). Focusing on longer intervals and a broader study period may highlight more of the impact of gentrification on violence, particularly gang homicide.
Previous studies also controlled for temporal variation in different ways. Several studies incorporated a temporally lagged measure of the dependent variable (Kreager et al., 2011; Papachristos et al., 2011; C. M. Smith, 2014). Kreager et al. (2011) found gentrification was not associated with violent or property crime controlling for a temporally lagged version of the dependent variable. Papachristos et al. (2011) found gentrification was negatively associated with homicide and robbery controlling for counts of each crime during the preceding 3-year period. C. M. Smith (2014) also controlled for counts of gang homicide during the preceding 3-year period and found gentrification was positively and negatively associated with gang-related homicide depending on how gentrification was measured. Kreager et al. (2011) also assessed within-unit changes in total, property, and violent crime over time and found gentrification was associated with short-term increases in aggregate property and total crime rates but not associated with changes in the aggregate violent crime rate.
Gentrification and Hollenbeck
Hollenbeck is a 15.2 square mile region east of Los Angeles’s urban center. Approximately 170,000 residents resided in eight neighborhoods: Boyle Heights, El Sereño, Hermon, Hillside Village, Lincoln Heights, Montecito Heights, Monterey Hills, and University Hills (LAPD, 2018). The eight neighborhoods were divided among 50 census tracts used as the units of analysis in the current study. Census statistics for the study period indicated this area remained over 80% Latino and continued to be a disadvantaged part of Los Angeles with over 25% of the population living below the poverty line (Minnesota Population Center, 2011). The intensity of disadvantage in this area was associated with an extended history of intergenerational gangs, which residents viewed as a fundamental social problem (Maxson, Hennigan, & Sloane Ranney, 1999; Moore, 1991; Vigil, 2007).
The quasi-institutional nature of gangs in Hollenbeck allowed them to adapt, evolve, and maintain influence in their barrios (Moore, 1991). Moore (1991) portrayed Hollenbeck as overrun with gang feuds where “violence seems to be endemic” (p. 67). Tita and colleagues (Tita & Radil, 2011; Tita et al., 2003) ascertained that political boundaries and physical barriers cloistered the neighborhoods in Hollenbeck to produce an environment where gang rivalries were restrained and interactions with groups in adjacent areas were limited, producing a natural field site. Features of the natural (hills, rivers, valleys) and built (highways, railways) environments also created buffers between neighborhoods and subsequent gang turfs within Hollenbeck (Smith, 2014; Valasik et al., 2017).
Like the rest of Los Angeles, Hollenbeck experienced a substantial reduction in serious crime and violence during the 1990s and 2000s (Valasik et al., 2017). The total number of active street gangs in Hollenbeck varied over the last three decades, yet the number of gangs remained relatively consistent since the late 1990s, with approximately 30 active street gangs that each claimed a geographically defined territory (Brantingham et al., 2012; Tita & Radil, 2011; Valasik et al., 2017). That the number of gangs remained relatively unchanged suggests the decline in violent crime was due to other changes to neighborhoods in this area, such as gentrification.
Tension among residents in Hollenbeck regarding gentrification are high due to recent influxes of younger Latinos with higher education levels and incomes than the incumbent residents. Given the Hispanic nature of the area, this process has been relabeled as “gentefication” (Ahrens, 2015) with newcomers being called “Chipsters,” a Chicano play on the word hipsters, or “gentefiers” (Medina, 2013; Mejia, 2016). Many of these “gentefiers” express the importance of preserving Hollenbeck’s Chicano culture, but many long-standing Chicano residents in the area disapprove of perceived and actual changes in the area. This strain between “gentefiers” and local residents in Boyle Heights has been so virulent over the years that it is even being used as a principal plotline on Starz’s drama Vida and Netflix’s comedy Gentefied (Betancourt, 2019; Reyes-Velarde, 2018). Friction has grown considerably in recent years about gentrification because a growing number of non-Latino-owned art galleries and business opened throughout the Boyle Heights area specifically. Resentment and opposition to the changing character of the community resulted in increased public demonstrations and vandalized businesses (Auge, 2017; Vives, 2017). Other important changes include the state-based gentrification of the 1,257-unit Pico-Aliso public housing complex, the largest public housing complex west of the Mississippi River, during the 1990s (Leavitt & Ochs, 1998). Prior to the state-based gentrification efforts, there were at least eight turf-based gangs claiming territory within this public housing complex (Fremon, 2008; Valasik, 2014; Vigil, 2007). The demolition, reconstruction, and selective re-admittance of residents in the early 2000s displaced most of these gangs. Presently, only three gangs retain their originally claimed turfs within the newly built housing complexes of Pueblo del Sol, Pico Gardens, and Las Casitas (see Brantingham et al., 2012; Radil, Flint, & Tita, 2010; Valasik, 2014; 2018; van Gennip et al., 2013).
Current Study
The current study examines the relationship of gentrification with total, gang, and non-gang homicide levels in Hollenbeck. These relationships were assessed with longitudinal negative binomial models to account for the influence of neighborhood changes that may have influenced each type of homicide. Previous longitudinal studies of homicide used a fixed-effects or random effects analysis strategy to control for unmeasured variable bias (Barton, 2016b; Papachristos et al., 2011; C. M. Smith, 2014). Fixed and random effects models were assessed, but results of the Hausman test indicated random effects regression models were preferable. While unable to control for unmeasured variable bias in the way fixed effects regression would, random effects regression was an improvement over ordinary least squares (OLS) regression because the standard error estimates adjust for the within-unit correlation that occurs with repeated measurements in longitudinal research (Allison, 2005). Tract population was treated as an exposure variable in all analyses.
Focusing on changes over an extended period of time within a small area such as Hollenbeck is advantageous for several reasons. The localized and geographically contained nature of gang violence means that focusing on the meso-level allows for a more nuanced insight into gang violence persistence across areas (Klein, 1995; Valasik et al., 2017). In addition, Maxson and Klein (2002) suggested examining gang homicide within a local jurisdiction in Los Angeles, due to its “multiple nucleated character” (p. 253) of urban villages that orbit a city center, allows for increased generalizability with other municipalities that lack Los Angeles’s sheer scale and size. Finally, using census tracts as the units of analysis diverges from the neighborhood clusters used by Papachristos et al. (2011) and C. M. Smith (2014) and the sub-borough areas used by Barton (2016b). This was an advantage of the current study because gentrification scholars recognize gentrification typically does not occur in an entire urban center but throughout particular sections, such as census tracts within Hollenbeck, which are situated within greater Los Angeles (Barton, 2016a).
Data and Variables
Dependent Variables
Homicide information was collected from the Homicide Detective Unit of LAPD’s Hollenbeck Community Policing Area. The data included information about 1,171 known homicides that occurred between 1979 and 2011 that was manually culled from individual homicide case files stored at LAPD’s station in the Hollenbeck Community Policing Area (see Valasik, 2014; Valasik et al., 2017). Incident-level information on the number of homicides in Hollenbeck were aggregated to 2010 Census tract boundaries. The dependent variables reflect 3-year homicide counts centered on 1990, 2000, and 2010 to account for annual fluctuations in lethal violence (Morenoff, Sampson, & Raudenbush, 2001).
A key contribution of the current study was the disaggregation of homicide counts into gang and non-gang homicide counts. This disaggregation was possible because the data collected from the LAPD were manually collected allowing for each homicide to be coded as gang-related or not. Gang-related events were coded under both member- or motive-based 1 definitions. A motive-based definition required a homicide to result from a direct function of gang activity (e.g., retaliation, territoriality, recruitment, etc.), while a member-based designation was more broadly defined to include any homicide in which a participant, victim, or suspect was affiliated with a gang. Debate remains about how a “gang” homicide should be appropriately designated (see Maxson & Klein, 1990, 1996). The current study used the member-based definition, erring by capturing incidents that may be motivated solely by an individual member’s purpose, “after all, gang members can and do act of their own accord” (Papachristos, 2009, p. 86). In contrast, a motive-based definition errs by “sampling too heavily on the dependent variable by capturing only those cases in which a group motive was determined” (Papachristos, 2009, p. 86). Thus, motive-based gang homicides were just a subsample of member-based gang homicides. Furthermore, restricting the dependent variable to a subsample may potentially discard valuable information (Pyrooz, 2012). Previous research found the same variables statistically distinguish a gang event from a non-gang event regardless so whether a member- or motive-based definition of gang homicide was used (Maxson & Klein, 1996).
Independent Variables
Data for the independent variables came from the National Historical Geographic Information System (NHGIS) (Minnesota Population Center, 2011). Data for 1980, 1990, and 2000 were gathered from the corresponding U.S. Censuses. Data for the 2010 cross-section were gathered from the American Community Survey (ACS) 2008-2012. For brevity, we refer to the ACS sample by the middle year of its time window (i.e., 2010).
There is no commonly agreed upon strategy for identifying gentrified neighborhoods, but much of the previous research identified gentrified census tracts through changes in population and housing characteristics (Barton, 2016a; Brown-Saracino, 2017). The current study uses the strategy described by Loukaitou-Sideris and colleagues (2017) who studied gentrification in four Los Angeles neighborhoods, one of which (Mariachi Plaza) is in the Boyle Heights community of Hollenbeck. Using this strategy, census tracts were classified as vulnerable to gentrification at the start of a given decade if they met at least three of four criteria: the percentage of low-income households was above the 40th percentile for the county, the percentage of residents with bachelor’s degrees or higher was below the 40th percentile for the county, the percentage of renters was above the county median, and the percentage of non-Hispanic White residents was below the county median. Census tracts were then classified as gentrified at the end of each decade if they were classified as vulnerable to gentrification and experienced all of the following: an increase in the percentage of residents with a bachelor’s degree greater than the average change for the county, an increase in median household income greater than the average for the county, an increase in percentage of non-Hispanic Whites greater than the county average, and an increase in median gross rent greater than the county average. Using this strategy, tracts were dichotomized as gentrified during the 1980s, 1990s, or the 2000s. Figure 1 shows the spatial distribution of gentrified tracts and the decade during which they gentrified.
Gentrification in Hollenbeck from 1980 to 2010.
Confirmatory factor analyses were conducted to create indices for concentrated disadvantage and residential stability using data from the 1990 and 2000 censuses and the 2008-2012 ACS 5-year estimates. Note that this is different from the gentrification measures, which reflected neighborhood changes between census cross-sections. This was done to help control for endogeneity because gentrification often occurs in socioeconomically depressed neighborhoods.
The measure of concentrated disadvantage was created for each census cross-section by taking the average of the standardized values for the percentage of the population living below poverty line, unemployment rate, percentage of female-headed households, and percentage of residents receiving public assistance. The residential stability index was created by averaging standardized values of the percentage of the population who lived in the same home for at least 5 years and the percentage of owner-occupied homes for each census. Percentage foreign-born was included as a control to account for ethnic heterogeneity. Percentage of the youth population was included to control for the population at greatest risk of offending or being victimized and was calculated by dividing the population aged 15 to 24 years by a census tract’s total population for each cross-section. Our analyses also controlled for the presence of public housing within each tract so that tracts that contained public housing were coded 1 and those that did not were coded 0. This variable was time variant for a single tract as the Pico-Aliso complex was demolished in 1997 and re-opened in 2004 (Ohland, 2004). We also controlled for temporal lag of each dependent variable.
The analyses controlled for spatial autocorrelation because homicide, especially gang homicide, tends to exhibit patterns of spatial dependence (Valasik et al., 2017). The assessment of spatial autocorrelation required the selection of a spatial weights matrix, which was a decision influenced by the data and theoretical framework guiding the analyses (Anselin, 2002). Given Hollenbeck’s distribution of census tracts, a queen’s contiguity first-order spatial weights matrix was initially developed using the GeoDa software package. The spatial weights matrix was adjusted to account for elements of the urban environment and connections between tracts that featured physical barriers (i.e., highways or railways) that limited interactions among proximate yet bounded gangs by preventing travel between tracts (L. M. Smith, Bertozzi, Brantingham, Tita, & Valasik, 2012; Valasik et al., 2017). Results of Moran’s I analyses indicated significant spatial clustering for all three cross-sections.
Results
Table 1 presents descriptive statistics for the dependent variables. These statistics indicate the average census tract experienced declines in total homicide and in gang homicide during the study period. For example, the mean number of total homicides declined from 3.140 in 1990 to 1.220 in 2010. Similarly, the mean number of gang homicides declined from 1.860 in 1990 to 0.880 in 2010. The trend for non-gang homicide deviated slightly from total and gang homicide as the mean number of non-gang homicides declined from 1.280 in 1990 to 0.340 in 2000 and remained at 0.340 in 2010. Analysis of variation in the number of homicides by tract in 2000 and 2010 showed the identical descriptive statistics for non-gang homicide to be purely coincidental.
Summary Statistics for Homicide Variables.
Note. N = 50 Census tracts.
The descriptive statistics for the gentrification measure show that one tract (2% of Hollenbeck) gentrified during the 1980s, six tracts (12% of Hollenbeck) gentrified during the 1990s, and two tracts (4% of Hollenbeck) gentrified during the 2000s. Descriptive statistics for concentrated disadvantage and residential stability feature a mean of zero for all three cross-sections because these measures were created by averaging standardized values of the component variables (see also Pyrooz, 2012). Statistics for the percent foreign-born indicate the average tract experienced a small decrease in the percent foreign-born between 1990 and 2000 and then again between 2000 and 2010. Descriptive statistics for the youth population variable show the average tract experienced small declines in the percentage of residents between the ages of 10 and 24 years between 1990 and 2000 and a very small increase between 2000 and 2010. Statistics for the public housing variable show that about 12% of tracts contained public housing in 1990 and in 2000, but only 8% of tracts contained public housing in 2000. This was because the Pico-Aliso housing complex was closed in 1997 and then re-opened in 2004, albeit in a different form.
Table 2 shows how the mean count of each dependent variable at the start of each decade varied in tracts classified as gentrified versus those not classified as gentrified. The results for the 1980 cross-section show that homicides were substantially more common in tracts that did not gentrify during the 1980s. Results for the 1990 cross-section, however, show that gang and non-gang homicides were more common in tracts that would go on to gentrify during the 1990s. Results for the 2000 cross-section show counts of gang homicide were slightly lower in tracts that would go on to experience gentrification during the 2000s, while counts of non-gang homicide were substantially lower in tracts that would go to experience gentrification than in non-gentrifying tracts. In summary, results in Table 1 suggest the average tract experienced declines in total, gang, and non-gang homicide, while results in Table 2 suggest spatial variation in the prevalence of total, gang, and non-gang homicide. That homicide levels at the start of each decade varied substantially among tracts that would and would not go on to experience gentrification provides further evidence that the relationship of gentrification with gang homicide should be assessed independently of gentrification’s relationship with non-gang homicide.
Mean Count of Homicides, Gang Homicides, and Non-Gang Homicides in Gentrified Versus Non-Gentrified Tracts at the Beginning of Each Decade.
Table 3 presents results of random effects negative binomial regressions. The results of the Total Homicide model indicate the number of homicides for the 2010 cross-section was significantly lower (β = −0.599, p < .05) than the count of total homicides in the average tract in 1990. The results also indicate gentrification was not significantly associated with homicide levels during the study period. Of the census-based controls, only the measure of concentrated disadvantage was significant. Tracts characterized by greater concentrated disadvantaged were more likely to feature homicides (β = 0.252, p < .05). The coefficients for the temporal and spatial lags of homicide were not significant.
Random Effects Negative Binomial Regression Results.
Note. N = 150 (50 Census tracts in each of the three periods). Standard error in parentheses.
p < .05. **p < .01. ***p < .001.
Results of the Gang Homicide model were similar to the Total Homicide model. First, results of the Gang Homicide model indicate the number of gang homicides for the 2010 cross-section was significantly lower (β = −0.574, p < .05) than the count of gang homicides in the average tract in 1990. Also, like the results of the Total Homicide model, gentrification was not significantly associated with variation in gang homicide levels. Of the remaining independent variables, only concentrated disadvantage was significantly associated with gang homicide. The positive coefficient for concentrated disadvantage (β = 0.329, p < .05) indicates gang homicides were more likely to occur in tracts characterized by concentrated disadvantage. Like total homicide, the temporal and spatial lags of gang homicide counts were not significant.
Results for the Non-Gang Homicide model indicate the number of non-gang homicides in the average tract was significantly lower in in 2010 (β = −0.804, p < .01) than in 1990. The most interesting finding of these results, however, was that gentrification (β = 0.689, p < .05) was positively associated with the number of non-gang homicides controlling for other neighborhood covariates. This suggests support for arguments made by routine activities that population changes in gentrifying neighborhoods may have been associated with changes in the availability of suitable targets (more high value targets) or changes in guardianship (gentrifiers less familiar with how to protect themselves in public; Anderson, 1990; Freeman, 2006; Taylor & Covington, 1988). The results also show concentrated disadvantage was positively associated with non-gang homicide levels (β = −0.317, p < .05). The coefficient for the temporal lag of non-gang homicide (β = 0.060, p < .05) was also significant, which indicates non-gang homicides were more common in tracts that featured higher counts of non-gang homicide at the start of a given decade.
Discussion and Conclusion
Gentrification, including the demolition of public housing communities, has been routinely used by municipalities to uproot gang activity and reinvigorate disadvantaged neighborhoods (Hagedorn & Rauch, 2007; C. M. Smith, 2014). Yet, our knowledge about the impact of gentrification on neighborhood crime, particularly gang homicide, remains limited (Barton, 2016b; Kreager et al., 2011; Papachristos et al., 2011; C. M. Smith, 2014). Much of the extant research focused on broader definitions of crime, but C. M. Smith’s (2014) findings in connection with findings presented by Papachristos et al. (2011) suggested the relationship of gentrification with gang homicide was unique from the relationship of gentrification with more broadly defined violent crime types. The current study contributed to research on the association of gentrification with violent crime by examining the relationship of gentrification with total, gang, and non-gang homicide in Hollenbeck, an area saturated with intergenerational gangs and which has experienced tensions associated with gentrification (Auge, 2017; Vives, 2017).
In contrast with recent research on the relationship of gentrification with homicide broadly defined (Barton, 2016b; Papachristos et al., 2011), we did not find gentrification was associated with variation in total or gang homicide during our 30-year study period, but we did find gentrification was positively associated with non-gang homicide. Research on the relationship of gentrification and crime drew upon the social disorganization and routine activities frameworks to develop competing hypotheses about whether gentrification would be positively or negatively associated with crime. Until recently, much of this research used cross-sectional research designs that were unable to fully assess the importance of neighborhood changes such as gentrification. The results of our analyses, while limited, should encourage researchers to conduct more longitudinal studies of social disorganization and routine activities so that these theories can be refined to better account for neighborhood change.
Our results are also important for policy makers because they suggest gentrification was not a major driver for the decline in either gang or non-gang homicide in Hollenbeck. Supplemental analyses, not included, assessed the importance of variation in gentrification over time by incorporating interactions of the time dichotomies with the gentrification measure. Results showed the relationship of gentrification with each dependent variable did not vary significantly over time. This is important because Valasik et al. (2017) suggested gentrification as a potential causal mechanism for declines in gang and non-gang homicide in Hollenbeck. Like Kreager et al. (2011) who found gentrification was not significantly associated with changes in violent crime and C. M. Smith (2014) who reported state-based gentrification was not associated with changes in gang homicide controlling for previous levels of gang homicide, we found gentrification was not associated with variation in total or gang homicide. This suggests that gentrification may not be a “silver bullet” policy for reducing neighborhood crime and violence especially in areas where intergenerational gangs have strong connections to their claimed turf and are committed to exert their presence regardless of urban planning or police interventions (Moore, 1991; Valasik, 2014; Vigil, 2007).
While we did not find gentrification was associated with total or gang homicide, we did find gentrification was positively associated with non-gang related homicide. This supports suggestions that homicide disaggregation is warranted when examining the influence of neighborhood change on homicide levels (Costanza & Helms, 2012; Kubrin & Wadsworth, 2003; Valasik et al., 2017). We were unable to identify why gentrification was differentially associated with gang and non-gang homicide, but previous research suggests a few possibilities. The lack of association of gentrification with gang homicide may be the result of areas with more entrenched gang activity being less likely to gentrify as gang members persist in the neighborhoods claimed by their gang despite change in the local neighborhood (Moore, Vigil, & Garcia, 1983; Valasik, 2014; Valasik & Tita, 2018). The positive association of gentrification with non-gang homicide may be a product of gentrifiers being at greater risk of experiencing a robbery due to being perceived as a more suitable target who lacks guardianship (Anderson, 1990; Covington & Taylor, 1989). Furthermore, because gentrifiers may not be as familiar with how to protect themselves on the street, these robberies have a greater likelihood of going “sideways” and ending in homicide. The increase in homicide in gentrifying areas may also be a product of increased violence more broadly defined as increased financial strain among incumbent residents encourage them to engage in illicit activities to meet basic needs (Boggess & Hipp, 2016; Taylor & Covington, 1988; Vigil, 2007). This does not mean that such individuals would engage in violence themselves, but instead that their illicit activities may attract violent offenders from other areas (Boggess & Hipp, 2016; Tita & Griffiths, 2005). Our findings in combination with these potential causal mechanisms should encourage policy makers to consider the possibility that gentrification may have unintended consequences in addition to the prevailing consequence of residential displacement.
Related to the importance of decomposing homicide types is the acuteness of our findings for research on the relationship of gentrification with gang homicide. Prior to the current study, only C. M. Smith (2014) had assessed the relationship of gentrification with gang homicide. C. M. Smith (2014) found a population-based measure of gentrification was negatively associated with gang homicide, while a state-based measure (demolition of public housing) was positively associated with gang homicide. Our results contrast with both findings as neither our population-based measure of gentrification or our state-based measure (presence of public housing) was associated with variation in gang homicide. These findings suggest the relationship of gentrification with gang homicide may vary from city to city, so we encourage future research on the relationship of gentrification with gang homicide in other cities with protracted histories of gang crime.
Our analyses also contribute to the broader gentrification literature. While discussions of gentrification in Hollenbeck have only recently picked up speed, our descriptive results show Hollenbeck experienced more gentrification during the 1990s than during the 2000s. This finding highlights inconsistencies in where gentrification was believed to have occurred versus where population trends indicate it occurred (Barton, 2016a; Brown-Saracino, 2017).
While the current study advanced research on the relationship of gentrification with violent crime broadly and gang homicide specifically, several limitations must be acknowledged. Like Valasik et al. (2017), we are unable to identify the causal mechanism(s) driving the decline in homicide in Hollenbeck. This was beyond the scope of this study, but we suggest this was due to local anti-gang interventions. This includes civil gang injunctions (Valasik, 2014), Community Law Enforcement and Recovery (Parks, 2005), Gang Reduction Program (Cahill & Hayeslip, 2010), L.A. Bridges (Building Resources for the Intervention and Deterrence of Gang Engagement) (Klein & Maxson, 2006), and Operation Ceasefire (Tita et al., 2003) any of which could have contributed to the decline in gang homicides. Valasik and colleagues (2017) attested that such policies were most likely to produce temporally brief reductions in gang homicide because of the stubbornness of intergenerational gang violence (see also Papachristos, 2013).
A second limitation is the reliance on census data to measure gentrification which required the unit of analysis to be the census tract. Smaller units of analysis (e.g., block, block-group, egohood, street segment) would have been desirable, but the variables needed to operationalize gentrification for the entire study period were unavailable for units smaller than census tracts prior to 1990. It was important to include the 1980 cross-section so that we had a data point before the great crime decline. Gang scholars also warn census tracts may be too spatially extensive of an areal unit to investigate the localized nature of gang violence (Blasko, Roman, & Taylor, 2015; Tita, Cohen, & Engberg, 2005). That said, prior neighborhood-level studies of crime and violence regularly utilize the census tract as their areal unit, which remains substantially more compact than either neighborhood-clusters (C. M. Smith, 2014) or cities (Pyrooz, 2012) used to examine the relationship of gang homicides with neighborhood-level structural features.
Finally, there are aspects of Hollenbeck that make our case study site unique from previous research. Mostly importantly, Hollenbeck is part of the city of Los Angeles, where other studies investigating the relationship of gentrification with total homicide or gang-homicide analyzed variation in all of parts of a city such as Chicago (Papachristos et al., 2011; C. M. Smith, 2014), Los Angeles (Boggess & Hipp, 2016; Lee, 2010), New York City (Barton, 2016b), or Seattle (Kreager et al., 2011). In addition, Hollenbeck has a long history of entrenched gang activity and concentrated disadvantage (Valasik et al., 2017). Given the potential distinctiveness of Hollenbeck and the dearth of research on the relationship of gentrification with gang activity, additional research needs to be conducted before a definitive statement about the impact of gentrification on gang violence can be made.
Policy Implications and Conclusions
The findings from this study reveal important considerations for urban planning and its influence on homicide in general and on gang homicide specifically. Hagedorn and Rauch (2007) warned that intergenerational gangs have a “symbiotic relationship with the community in which they exist cannot easily be eliminated through criminal justice measures” (p. 446). Crime, especially gang violence, tends to be a stubborn phenomenon with which local communities must engage (Hagedorn & Rauch, 2007; Sampson, 2012). The results of the current study highlight this “stickiness” and suggest efforts by policy makers to reduce neighborhood violence by encouraging gentrification or other forms of urban renewal may be limited at best or at worst misguided. Given that municipalities have increasingly provided greater access to crime data, we encourage further study of the relationship between gang violence and a neighborhood’s structural covariates.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
