Community-Level Social Processes and Firearm Shooting Events: A Multilevel Analysis

Abstract

Firearm violence is a leading public health issue that contributes to significant health inequalities within communities. Relatively little is known about the community-level social processes that occur at the street segment level and contributed to the community variation of firearm violence. This study examines the spatial patterns of firearm shooting events on street segments and the associated community-level social processes at both the street segment and neighborhood level. Multilevel mixed-effects Poisson regression was used to assess the relationship between measures of social disorder, physical disorder, and collective efficacy at the street segment-level and neighborhood-level measures of social disorganization theory. The results demonstrate that firearm shooting events occur on a small number of street segments across the city. Street segments with higher levels of social and physical disorder, along with lower levels of collective efficacy, are expected to have higher rates of firearm shooting events when accounting for neighborhood-level measures. Overall, the findings indicate specific street segments are experiencing higher rates of firearm shooting events and that these events are influenced by social processes. Prevention efforts should be focused on street segments experiencing higher rates of shootings.

Electronic supplementary material

The online version of this article (10.1007/s11524-020-00424-y) contains supplementary material, which is available to authorized users.

Introduction

Firearm violence is a public health crisis in the USA. The rate of homicides continues to increase across most of the USA [1] and nonfatal shootings occur at a higher rate than homicides [2]. Young, minority males are at highest risk for firearm-related injuries, leading to large disparities within urban environments [3, 4]. Homicides are known to cluster across counties, cities, and neighborhoods [5–7]. The community context of these “hot spots” [8] contributes to individuals’ risk for violent victimization, as firearm violence spreads through communities much like an infectious disease [9, 10].

Firearm violence is concentrated within larger communities on individual street segments or street blocks. For instance, Braga and colleagues in Boston found gun assaults were spatially concentrated with 75% of the gun assaults occurring on only 5 % of the city’s street segments [11]. Additional work by Koper and colleagues examined shootings at the street segment and found approximately 8% of the street segments accounted for 64% of the shootings in Minneapolis. Higher numbers of shootings were even more concentrated; for example, only 2.8% of the city’s streets experienced more than ten shootings over the 24-year period [12]. These studies, however, did not examine the community social processes occurring on these street segments with high rates of firearm violence.

Community-level characteristics such as concentrated disadvantage, residential instability, ethnic heterogeneity (i.e., social disorganization theory), neighborhood disorder, and low levels of collective efficacy [13] are associated with higher levels of violence [14–16]. Furthermore, as broken windows theory [17] posits, disorder is prevalent in communities with higher concentrated disadvantage, as they often lack resources for citizens to invest back into their communities [15, 18]. Small signs of disorder can indicate a lack of informal social control [19], therefore, increasing fear among residents, and residents may be less willing to act as guardians of the neighborhood [20]. The majority of community-based research, however, has been conducted at the neighborhood level (i.e., census tract, county level) [5, 15], which may miss the variation in social processes across street segments within neighborhoods, as street segments have their own behavior settings [21].

A challenge to studying community social processes at the street segment level is a lack of available data at the individual address or street segment level. Criminologist have historically utilized police 911 calls to measure community social disorder [22, 23], since these calls indicate a lack of informal social control, as residents are unable to handle interpersonal disputes themselves and must involve the police to resolve issues. Police 911 calls were recently validated as a measure of social disorder in Boston [24]. Similarly, scholars have begun utilizing large administrative datasets, such as, 311 calls, the number of active voters, and rain collector barrels to measure physical disorder and collective efficacy at the micro place [25, 26]. In Seattle, Weisburd and colleagues examined community-level characteristics, such as informal social control, disorder, and collective efficacy at the street segment level, and findings demonstrated that community-level characteristics vary from street to street [26]. More recently, Wheeler demonstrated a positive relationship between 311 calls for physical disorder and crime in Washington, D.C. [25] These studies, however, utilized overall crime trends and did not examine firearm shooting events, which may be driven by different social processes given the complexity of interpersonal violence [27].

To fill this gap, this study utilized police data on firearm shooting events and other community-level measures in Indianapolis, Indiana. Using data over a 2-year period, this study examined the spatial patterns and social processes associated with firearm shooting events (e.g., fatal and nonfatal shootings) at the street segment level while accounting for variation in community-level characteristics across neighborhoods. The purpose is to understand the social processes that occur on street segments with higher rates of firearm shooting events. Identifying these social processes is key to designing effective intervention programs to prevent future violence.

Methods

The current study takes place in Indianapolis, Indiana, and draws data from multiple sources: the Indianapolis Metropolitan Police Department (IMPD), the Indianapolis Mayor’s Action Center (MAC), the City of Indianapolis (data.indy.gov), and the US Census Bureau. Indianapolis is an ideal city for this research due to the level of firearm violence that occurs annually and the unique number of available data sources which allows for measurement of different physical and social constructs. According to the Federal Bureau of Investigations, in 2015, Indianapolis was ranked the tenth most violent city in the nation for cities with a population over 200,000 people [28]. The homicide rate was 17 per 100,000 population, and the violent crime rate was 1288 per 100,000 population.

Dependent Variable

The outcome measure in this study is a firearm shooting event. Shootings in 2016 were chosen to ensure that the effects of the community measures (in 2015) would be independent. Nonfatal shooting data were obtained from the Indianapolis Nonfatal Review database. Local researchers’ hand entered, coded, and cleaned data, which was obtained using information from both IMPD’s police incident reports and internal police documents when an individual was injured by a powder discharge weapon [2, 29]. Fatal shooting data were obtained from the IMPD homicide records management database. In 2016, there were 561 (n = 135 = fatal and n = 426 nonfatal) firearm shooting events. These events were combined, as the difference between a fatal and nonfatal shooting may only be the distance to a trauma center [30]. Firearm shooting events were measured at the count level.

Community-Level Measures

Neighborhood-level measures of concentrated disadvantage, residential instability, and ethnic heterogeneity were used to measure social disorganization. Data were drawn from the American Community Survey for 2015 and defined using principal components analysis as done in prior studies [13]. Concentrated disadvantage was defined from four socioeconomic measures: percent unemployment, median household income, percent living in poverty, and percent of female-headed household [13, 31, 32]. Residential instability was defined from the percentage of population residing in the same residence for the last year, percent of owner-occupied homes, and the percent moved within the last year [13]. Ethnic heterogeneity was defined from the total Hispanic population and the percentage of the population that was foreign born [13]. The neighborhood proportion of African American residents was included at the neighborhood level to account for the racial composition of each neighborhood [4–6]. All four measures were included as continuous measures at the census tract level and were centered by the grand mean.

Street Segment-Level Measures

Collective efficacy was defined by using the number of calls residents made to the Indianapolis Mayor’s Action Center (MAC) (i.e., 311 calls) in 2015. The MAC receives calls about trash, tall weeds/grass, graffiti, zoning concerns, pot holes, broken traffic signals, and illegal dumping. This measure deviates from historical work on collective efficacy that utilized citizen surveys to measure resident engagement [13]; however, this study draws on Uchida and colleagues’ definition of collective efficacy, “the ability of residents to produce social action to meet common goals.” [33] Others have used 311 calls to measure physical disorder but suggest 311 calls also measures civic engagement because when a resident makes the decision to take action and call to report an issue, that individual is taking responsibility for that public space [34]. Utilizing 311 calls also follows and extends recent research which uses administrative records to measure collective efficacy by utilizing active voters, community gardens, and rain collector barrels [25, 26]. The number of citizens calls for public space (tall weeds/grass/ trash, graffiti, illegal dumping) were totaled, reverse coded, and aggregated to the corresponding street segment, then divided into quartiles and dichotomized into a binary variable with the top quartiles (lower collective efficacy) and the lower quartiles (higher collective efficacy) [35].

Social disorder was measured using data obtained from the IMPD 911 computer-aided dispatch (CAD) calls for service in 2015. Police 911 calls for narcotics, public intoxication, disturbances, and loud noise complaints are included as indicators for social disorder [23, 24]. All call types were combined into one index variable to operationalize social disorder and divided into quartiles and then dichotomized into the upper (higher social disorder) and three lower quartiles (lower social disorder) [35]. Physical disorder was obtained from the City of Indianapolis (data.indy.gov) and was defined as the number of abandoned homes broken into the top (higher physical disorder) and three lower quartiles (lower physical disorder) [35].

Units of Analysis and Geocoding

This study uses two levels of spatial aggregation as the units of analysis: [1] street segments (n = 53,922) and [2] census tracts (n = 205). Street segments are defined as “the two block faces on both sides of a street between two intersections” [36]. Arterial and residential streets are included, and highways¹ are excluded due to the lack of human activity that occurs on such segments, as well as intersections [26, 36]. Census tracts are the unit of analysis that define the neighborhood [37]. Not all street segments are nested within census tract boundaries; therefore, to ensure spatial interdependence, all street segments that cross boundaries with a census tract (n = 5547) were removed [38]. This follows prior work and still includes 96% of all street segments within Indianapolis [39].

All data measures were geocoded through ArcGIS (version 10.4.1), which involves assigning X and Y coordinates to each individual address. All measures were automatically matched, and those that did not automatically match were matched by hand. Of the 561 firearm shooting events, 32 were unable to be geocoded due to unknown or unmatched addresses, resulting in a 94% geocoding rate. The majority of these unmatched addresses were nonfatal shootings, where the victim was unable or unwilling to give a specific location of where the shooting occurred. These unknown locations speak to a larger issue of uncooperative victims [40]. Next, using a street address file, a database was developed to identify and maintain all street segments within the city of Indianapolis, not just those that have a reported firearm shooting event, indicator of social disorder, physical disorder, or call into the MAC center [25]. Each measure was then aggregated to the corresponding street segment. Street segments have a mean length of 391.5 ft (SD = 380.9).

Spatial Weights and Analyses

Given the spatial nature of these data, a spatial lag variable was created for each of the street level variables. Spatial lags were based on the count of the total of calls in neighboring locations and were calculated using a queens contiguity spatial weights matrix [25]. Residual spatial autocorrelation was also assessed using Moran’s I, and 99 permutations were run to determine statistical significance [41]. A lower value of Moran’s I indicates little to no spatial autocorrelation in the residuals, which is ideal.

Descriptive statistics were calculated for all community social processes in 2015 and firearm shooting events in 2016 (Supplement A). The number of firearm shooting events on each street segment was calculated, and the rate per street segment was determined. Given that the mean and variance of firearm shooting events was nearly identical, a multilevel mixed-effects Poisson² regression analysis was performed to assess how community mechanisms at both street segment and neighborhood influence the likelihood of a firearm shooting event occurring. All analyses were conducted in Stata [42].

Results

Spatial Patterns of Firearm Shooting Events

There were 481 street segments that experienced at least one firearm shooting event. The majority of street segments never experienced a firearm shooting event (99%), and another 439 (0.814%) only experienced one firearm shooting event. There were 38 street segments that experienced two shooting events (0.070%), and only 2 street segments that experienced three and four shooting events (0.004%) (Table 1). These results display that firearm shooting events are occurring on less than 1 % of street segments in Indianapolis.

Table 1.

Number of firearm shooting event per street segment

# of Firearm violence Incidents (n = 529)	# of street segments	Total % of street segments
> 4	2	0.004
3	2	0.004
2	38	0.070
1	439	0.814
0	53,441	99.1
Total	53,922	100

Figure 1 displays the distribution of firearm shooting events by street segment overlaid on the rate of firearm violence for each neighborhood.³ Several patterns begin to emerge as you examine the spatial concentration of street segments. First, a large number of street segments with more than one shooting are concentrated within the center of the city. Secondly, street segments with more than four shooting incidents (highlighted in purple) are spatially dispersed across the city and not concentrated within neighborhoods with high levels of firearm violence. Lastly, in neighborhoods with high firearm violence rates, not all street segments experience a shooting event, and indeed, most do not. Further, seemingly isolated street segments across the city experience single shooting events and may be attributed to other confounding factors such as incident motive and other crime generators (e.g., gas stations, bars, and liquor stores), which are not examined in this study.

Fig. 1.

Firearm shooting events on street segments in Indianapolis, Indiana, 2016

Multivariable Factors Associated with Firearm Shooting Events

Three multivariable models were examined: one including only street segment level variables (Model 1); one including both street segment and neighborhood level variables (Model 2); and one including all prior measures with the addition of the percent of African American residents (Model 3). These models included spatial weights for each of the community social process measures and assessed the odds of a firearm shooting incident occurring as a function of both the community social processes at the street segment and neighborhood (Table 2).

Table 2.

Multilevel Poisson regression model of firearm shooting events

N = 53,922	Model 1		Model 2		Model 3
	IRR	Std. error	IRR	Std. error	IRR	Std. error
Street segment level
Social disorder	10.7	1.51	10.5	1.49	10.2	1.43
Social disorder neighbor	1.08	0.021	1.08	0.012	1.08	0.012
Physical disorder	2.56	0.385	2.23	0.327	2.02	0.289
Physical disorder neighbor	1.43	0.453	1.15	0.361	0.971	0.308
Collective efficacy	0.584	0.069	0.627	0.074	0.597	0.070
Collective efficacy neighbor	1.04	0.084	1.03	0.083	1.05	0.083
Neighborhood level
Concentrated disadvantage			1.73	0.150	1.37	0.107
Residential mobility			0.943	0.083	0.969	0.073
Ethnic heterogeneity			0.891	0.067	1.07	0.071
Percent African Americans					1.02	0.002
Moran’s I	0.04		0.05		0.06
Constant	0.002	.000	0.002	0.000	0.002	0.000

Statistically significant values at p < 0.05 level in bold

For Model 1, street segments with higher calls for social disorder (IRR 10.7; p < 0.01) and a larger number of abandoned homes (IRR 2.56; p < 0.01) are expected to have a higher rate of firearm shooting events compared to those with lower social disorder and abandoned homes, whereas streets with higher levels of collective efficacy (IRR 0.584; p < 0.01) are expected to have a lower rate of firearm shooting events.

For Model 2, neighborhood level measures of social disorganization were included. Higher rates of firearm shooting events were ten and a half times greater for street segments with higher calls for social disorder (IRR 10.5; p < 0.01), approximately two times greater for streets with more abandoned homes (IRR 2.23; p < 0.01), and for neighborhoods with higher levels of concentrated disadvantage (IRR 1.73; p < 0.01), respectively. Additionally, higher levels of collective efficacy at the street segment level (IRR 0.627; p < 0.01) decreased the expected rate of shooting events. Residential mobility and ethnic heterogeneity did not reach the level of statistical significance (Table 2).

Lastly, for Model 3, the percentage of African American residents were included to account for the spatial concentration of violence within minority neighborhoods. Findings are similar to the prior two models, higher levels of social (IRR 10.2; p < 0.01) and physical disorder (IRR 2.02; p < 0.01) at the street segment level, and higher neighborhood concentrated disadvantage (IRR 1.37; p < 0.01) are expected to have higher rates shooting events, whereas higher levels of collective efficacy decrease the expected rate of a shooting occurring (IRR 0.597; p < 0.01), respectively. Across all three models, the spatial effect of social disorder had a smaller effect (IRR 1.08; p < 0.01) than the local effect, however, the spatial effect of physical disorder and collective efficacy did not reach statistical significance. These results suggest there may be confounding variables missing from the models; though, given the lag incident rate ratios are smaller than the local incident rate ratios, the bias is assumed to be small.

Discussion

This study used multiple data sources linked at both the street segment and neighborhood level to assess the association between community social processes and firearm shooting events. Overall, the findings indicate specific street segments are experiencing higher rates of firearm shooting events and that these events are influenced by social processes. These findings confirm prior work that identified a small number of street segments were responsible for a large number of firearm violence [11, 12] and extends prior work on micro place behavior [26, 43], by demonstrating the importance of the social processes on street segments with higher rates of firearm shooting events. These findings have implications for policy and practice.

As expected, neighborhoods with higher levels of concentrated disadvantage and higher proportions of African American residents were associated with higher rates of firearm shooting events at the street segment level. Prior work demonstrates homicides are concentrated in areas with high African American populations and with racial isolation [6, 44]. Similarly, recent work displayed racial segregation may increase the disparity in firearm homicide at the state level [45]; moreover, when examining modern day firearm assault injury, rates were highest in historically racialized areas [46]. Within urban environments, African Americans disproportionally represent individuals living in disadvantaged communities, which increases their exposure to guns, violent injury, and the psychological harms of violence [47–50]. The findings from this study speak to a larger issue of the racial disparities that exist in violent victimization, specifically shootings and how the street segment and neighborhood may be just as important as the individuals involved [8]. These are clear directions for future research.

Further, findings from this study indicate community-level social processes are occurring on individual street segments within neighborhoods. For instance, higher levels of social and physical disorder and lower levels of collective efficacy were associated with higher rates of firearm shooting events at the street segment level. These findings may indicate street segments that experience firearm shooting events have low guardianship or residents who are not willing to use informal social control; therefore, the community must engage the police to address common issues such as fights, public intoxication, and drug dealing, whereas street segments where residents are calling the MAC to request governmental resources are associated with lower rates of shootings and may be an indicator of informal social control at the street segment level [51].

The behavior of these street segments are no doubt influenced by the larger neighborhood in which they reside [52]. Understanding where these shooting events are occurring can help focus prevention efforts to street segments experiencing shootings events and not unnecessarily designate the entire neighborhood as problematic, which may cause unnecessary harm [51, 53]. Research demonstrates prevention strategies are most effective when they are focused on high-risk people and places, involve alternatives to arrest, and engage multiple community partners [54]. For instance, hot spot policing, which focuses police attention to specific locations with high crime, has demonstrated reductions in crime across multiple studies and decades [55]. Other programs, such as community greening projects, which encourages community involvement through greening of vacant lots has also demonstrated reductions in violent crime [56, 57]. A combined notion, collective efficacy policing, suggests officers can engage in community-building strategies at the street segment to improve crime, disorder, and facilitate informal social control [58]. A more recent model, partnership-oriented crime prevention, combines police, residents, local businesses, and other city agencies to help prevent violence within hot spots and has demonstrated promising results [59]. These are all promising interventions that can be focused on street segments with high rates of shooting events.

Limitations

While this study identified the spatial patterns and social processes associated with firearm shooting events, there are a number of limitations. There are likely other event circumstances and victim demographics that contribute to these events [2]. This study examined the shooting incident location using police data, which is not available in most clinical data sources; [16] however, all shooting events may not be reported to the police [60]. Additionally, this study did not explore the differences of why shootings are not occurring on all streets at the same rate and should be addressed in future studies. This study did not account for differences between fatal and nonfatal shooting events, as seen in prior studies; [2, 30] however, this should be a direction for future research. These results are also limited by a resident’s willingness and ability to call both 911 and the MAC. Prior research indicates residents are willing to engage the police based on their sex, age, prior arrest history, and crime type [61, 62]; furthermore, residents within disadvantaged communities believe the police make communities safer and should be involved to help resolve neighborhood issues [61, 63]. It is possible that only engaged residents are the ones calling the MAC and that not all streets experience the same need to call for city services; however, using MAC calls indicates community engagement and is a potential data source that is readily available to police departments and other city agencies as they design interventions. Lastly, prior research has utilized 311 calls as indicators of physical disorder and suggests a positive correlation with crime, as in the more 311 calls, the more physical disorder, whereas this study interprets more calls for public space as an indicator of more resident engagement or collective efficacy.

Based on this research and other studies, community-level social processes are a key factor in identifying and addressing firearm violence within the community. The conclusions of this study point to the importance of identifying where firearm shooting events are occurring at a micro street segment level and more importantly understanding the community-level social processes occurring at these micro hot spots in order to design effective interventions to reduce firearm violence within the community.

Funding Information

This work was partially supported by award No. 2013-R2-CX-0015, awarded by the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect those of the Department of Justice.