Examining the neighborhood effects on police performance to assault calls.
0 General Document comments
0 Sentence and Paragraph comments
0 Image and Video comments
Using agency-generated data collected from the Houston Police Department (HPD) and the 2000 census statistics, this article examines the relationship between police performance and neighborhood disorganization through an analysis of the distribution of police response to in-progress assault calls across different Houston neighborhoods. The results of multilevel analyses suggest that concentrated disadvantage, immigrant concentration, and residential stability are significantly related to the distribution of the HPD's response time patterns. More specifically, police responses were quicker to in-progress assault calls in disorganized neighborhoods. The implications drawn from the current study's results on police response time patterns can be useful in improving police service levels, the police-community relationship, and patrol strategies.
Keywords: rapid response; neighborhood disorganization; calls-for-service; assault calls
Introduction
Police patrol encompasses numerous roles (i.e. addressing crime, delivering emergency services, engaging in rapid response, and providing a visible police presence), the most vital traditionally being crime prevention, combat, and arrests (Eck & Rosenbaum, [15]; Hoover, [ 6]). These functions are crucial for both deterrence purposes and the reassurance of the public (Hoover, [ 7]). Although existing research indicates that unfocused (random) patrol in general and rapid response, in particular, are not profound strategies for reducing crime (see for example, Kelling, Pate, Dieckman, & Brown, [ 9]; Sherman & Eck, [19]), rapid response beyond its deterrent effect is still important for reducing harm, collecting evidence and witness information, and fulfilling citizen expectations (Sherman & Eck, [19]).
Despite the importance of rapid response in pthe olice patrol operations, studies tend to neglect the geographical distribution of police response. This is an important oversight because the distribution of rapid response reflects the availability of police resources and performance across different neighborhoods. Additionally, examining police performance at the neighborhood level would be beneficial to practitioners, policy-makers, researchers, and academics in improving police performance and police-community relations both inside and outside the USA Using the 2000 US Bureau of Census statistics and 2006 Houston Police Department's (HPD) calls-for-service data, the current study seeks to address the gap in the literature by examining the application of social disorganization theory to the geographical distribution of rapid police response. More specifically, this article uses hierarchical linear modeling to examine how the level of neighborhood calls and disorganization (concentrated disadvantage, immigration, and residential mobility) influence the distribution of police response time to in-progress assault calls.
Literature review
Neighborhood structures and distribution of police response
The disparity in police response time patterns across different neighborhoods (i.e. high crime and disorganized) is likely to occur because police departments manage limited resources over the strategic deployment of patrol officers in response to citizen demand for police services and the geographic distribution of crime (Hoover, [ 7]). Only a few studies attempted to assess the distribution of police response practices across poor and minority neighborhoods (Cihan, [ 4]; Cihan, Zhang, & Hoover [ 5]; Mladenka & Hill, [13]; Stevens, Webster, & Stipak, [22]). In York, Pennsylvathe nia, for example, Stevens et al. ([22]) examined the link between rapid response, patrol workload, time of day, and location of the crime. Their results suggested that in-progress calls or more serious crimes (e.g. murder and manslaughter) received faster police response than non-emergency calls. They also reported that commercial places and high-crime areas are more likely to be associated with shorter response times, while the time of day (shift change and noon) is more likely to be related to a delay in response time.
Mladenka and Hill ([13]) examined the effects of neighborhood socioeconomic characteristics on the distribution of police response time to calls-for-service in Houston. Based on the analysis of a small number of calls reported (N = 660), they concluded that there was no systematic response bias against poor and black neighborhoods. This study used meaningful predictions about neighborhood characteristics expected to influence police response time, but they were not able to sample a large number of calls in their analysis.
In a recent study, Cihan et al. ([ 5]) sought to test the hypothesis that the distribution of in-progress burglary calls is related to neighborhood structural characteristics. The study was based on the analysis of 4917 in-progress burglary calls across 420 Houston's census tracts in 2007. Cihan et al. ([ 5]) reported that there is a negative relationship between police response time, concentrated disadvantage, and immigrant concentration, whereas the relationship between police response time and residential stability was positive. More specifically, shorter police response time was observed across socially disorganized neighborhoods (see also Cihan, [ 4]).
Overall, response time studies appeared nearly a half-century ago, but later work downplayed the effectiveness of rapid response on patrol operations (Sherman & Eck, [19]). Studies generally measured response time as the time interval between a call received by dispatcher and the arrival of the unit to the scene (e.g. Blake & Coupe, [ 1]; Cihan, [ 4]; Cihan et al., [ 5]). With a few exceptions, however, these studies were generally unable to specify the different call priorities. Some recent work has attempted to show how neighborhood characteristics influence police response patterns (Cihan, [ 4]; Cihan et al., [ 5]) The current study expands previous research through an examination of the effects of neighborhood structural characteristics on rapid response to in-progress (priority one) assault calls using a large sample of incidents and census tracts.
Theoretical considerations
Social disorganization theory focuses on the relationship between neighborhood structural characteristics, social control, and the nonrandom distribution of crime across the city (Kubrin & Weitzer, [10]; Sampson, [16]). A considerable amount of research has advanced the theoretical underpinnings of social disorganization theory, incorporating several community-based factors, such as informal control, social network, social capital, social ties, and collective efficacy (see Sampson, [16]). In fact, the revival of theory formulates these key concepts in a systemic causal model of social control that mediates the relationship between neighborhood structural conditions (i.e. rates of poverty, unemployment, and female-headed households, racial and ethnic heterogeneity, residential stability), and crime (e.g. Bursik, [ 2]; Bursik & Grasmick, [ 3]).
Studies testing social disorganization theory have focused on neighborhood social control wherein emphasis is placed on the role of informal control (such as family and friends) through an examination of private control (relationships among the family and groups) and parochial control (interpersonal networks amidst residents, friends, and local groups) (Bursik & Grasmick, [ 3]). However, research examining social disorganization theory has understudied formal control, a form of social control that refers to practices of government, particularly police agencies. From the social disorganization perspective, it is expected that disorganized communities are unable to sustain police resources to address the crime problem (Kubrin & Weitzer, [10]; Triplett, Gainey, & Sun,[23]). In other words, neighborhoods with a high level of disorganization are expected to have insufficient patrol deployment, police protection, and other police services to fight crime. Therefore, it is important to examine the neighborhood level of police practices and services to understand the distribution of formal control across neighborhoods.
The degree to which level of social disorganization exists across neighborhoods can be assessed by the level of concentrated disadvantage, immigrant concentration, and residential stability (Sampson, Raudenbush, & Earls, [18]). These components may limit the capacity of a neighborhood to maintain both formal and informal social control and allocation of governmental resources. From the social disorganization perspective, one would assume that disorganized neighborhoods are more likely to receive longer police response time as compared to more affluent and stabilized communities.
Overall, there is limited research examining the relationship between a police response and neighborhood structural characteristics. While many hypotheses concerning macro-sociological explanations of police response can be derived from the literature, this study expands previous research by exploring the distribution of police response to in-progress assault calls, taking into account the characteristics of socially disorganized neighborhoods (i.e. concentrated disadvantage, immigration, and residential stability).
Hypotheses
Based on previous research and social disorganization theory, the following hypotheses were developed:
Hypothesis 1: The level of social disorganization influences the geographical distribution of police response time. There is a difference in the distribution of rapid response time between socially disorganized neighborhoods (high level of concentrated disadvantage, immigrant concentration, and low residential stability) and wealthier, stabilized neighborhoods.
Hypothesis 2: Police patrols are expected to have a higher workload in neighborhoods with high call rates. Therefore, the level of neighborhood call rates influences police response time.
Data and method
The current study uses 2006 agency-generated calls-for-service data obtained from the HPD and the 2000 US Bureau of Census statistics. The current study focuses on Priority One calls that require rapid response when 'a potential threat to life or the potential threat of serious bodily injury is in-progress' (HPD, [ 8], p. 2). The data screening process revealed that the HPD received 1275,622 calls-for-service in 2006. Selecting only Priority One calls or excluding lower-level calls from the analysis also minimizes differences in dispatch protocol because a quicker police response to in-progress calls is less likely to be influenced by police discretion and biases about the incident (Sherman, Gartin, & Buerger, [20]). Of all calls, 33,179 (2.6%) calls in the HPD were classified as Priority One.
The HPD call for service data includes information regarding a variety of call characteristics, including the address of the incident (i.e. street and apartment number), geographic coordinates (X and Y) of the incident, call priority (e.g. priority 1, 2, 3, and 4), exact timing of dispatch, unit arrival time, and weapon use. Screening all Priority One calls revealed 14,070 in-progress assault cases in 2006. After excluding 248 calls (186 outliers and 62 missing cases) from the in-progress assault calls, the final sample included 13,822 in-progress assault cases.
This study also utilized the 2000 US Bureau of Census statistics (Level 2) to examine the effects of neighborhood structural characteristics (race, ethnicity, immigration, poverty, unemployment, and residential stability) on the distribution of police response. Consistent with the social disorganization literature (Sampson & Raudenbush, [17]; Sampson et al., [18]), the current study measured the three dimensions of the level of social disorganization by using the 2000 US Bureau of Census statistics. Examining Houston neighborhoods using census tract-level data was expected to yield a better analysis of social disorganization theory with a more heterogeneous population (Sampson, [16]; Sampson & Raudenbush, [17]; Sampson et al., [18]). Houston was the fourth largest metropolitan city in the USA with 2000 census populations (49% White, 37% Latino, and 25% Black) of over one million residents (US Census Bureau, [24]).[ 1] This incredible population growth in recent decades in Houston makes the natural boundaries of neighborhoods difficult to determine as was that in some historic cities such as Boston, Chicago, or Detroit.[ 2]
Dependent variables
The dependent (outcome) variable was obtained from the HPD's agency-generated calls-for-service data, which was officially recorded into the CAD system. The outcome variable, response time to in-progress assault calls, was measured using the CAD screen that shows the types of calls already engaged by units, a unit's responding progress (i.e. en-route, on-scene, and reporting process), and a unit's availability for deployment. Once the dispatcher receives a citizen call, the CAD system begins to record the exact time of a call, the dispatch, travel, and arrival times for each patrol unit. Consistent with previous literature, patrol response (travel) time interval (the time from dispatch to the arrival of the scene) was calculated.
Level 1 measures
All Level 1 variables were obtained from the HPD calls-for-service data. Using incident time and date, day of the incident variable was calculated as weekends (1 = Friday and Saturday) and weekdays (0 = Sunday through Thursday). It is important to control the effects of police workload because police are busier on Fridays and Saturdays than other days of the week (Mladenka & Hill, [13]). To control the effects of time of the day, night-time was measured as whether the call was received during the daytime (0) or during night-time ( 1). The legal definition of night-time in Texas is the 'time between thirty minutes after sun set and thirty minutes before sun rise' (Laws v. State, [11], p. 655). Based on this definition, the average monthly sunset and sunrise times were calculated for each incident time recorded in the CAD (US Naval Observatory, [25]). Source of the call was also controlled (911 calls = 1 and 10-digit calls = 0). It is also important to control whether or not the incident involved a weapon because an assault (in some states defined as battery and aggravated assault) with the use of a weapon is considered a more serious crime than one without the presence of a weapon (Smith, [21]). Therefore, the use of a weapon in an assault case may possibly influence police response (weapon = 1 and no weapon = 0).
Level 2 measures
To measure neighborhood social disorganization (see e.g. Sampson & Raudenbush, [17]; Sampson et al., [18]), concentrated disadvantage, immigrant concentration, and residential stability were measured using factor scores (see Table 2) extracted from 10 census tract-level items (percent below poverty line, on public assistance, female-headed families, unemployment, less than age 18, Black, Latino, foreign-born, living in the same house as in 1995, and owner-occupied household). Call rates were also measured using the aggregated number of calls-for-service reported in a year at each census tract divided by each census population with this quotient multiplied by 1000 (McEwen & Taxman, [12]; Sherman et al., [20]). Calls-for-service rates should be a central consideration in social disorganization theory because they are inherently related to police workload and are disproportionately distributed across neighborhoods. Calls rates were used as a proxy measure for police workload (Sherman et al., [20]).[ 3]
Measurement and analytic strategies
To test the hypotheses, data-sets (agency-generated calls-for-service data and census statistics) were merged into one nested data structure using the geographical identifiers in the data-sets. Because police response (travel time) was measured as a continuous variable (outcome variable) in a multilevel data structure, hierarchical linear modeling was chosen to determine the effects of Level 1 and Level 2 variables on response time.
Results
Table 1 presents descriptive statistics for dependent and independent variables. The descriptive results for the in-progress assault cases revealed that the average response time was about three and a half minutes, with a standard deviation of 2.41 and a range from.02 to 29.85. The majority of in-progress assault incidents (52%) occurred during night-time. The use of a weapon was reported in a small percentage of the incidents (18%). In 2006, the average calls-for-service rate (per 1000 citizens) per census tract in Houston was about 660 calls with a standard deviation of 701.
Table 1. Descriptive statistics for the HPD calls-for-service characteristics (N = 13,822).
| Variables | Frequencies (%) | Mean | SD | Range |
| Response time | 3.65 | 2.41 | .02–29.85 | |
| Day of the incident (weekends) | 4379 (32) | 0–1 | ||
| Source of the call (911 dials) | 6520 (47) | 0–1 | ||
| Night-time | 7150 (52) | 0–1 | ||
| Weapon | 2531 (18) | 0–1 | ||
| Calls rates per 1000 citizens | 660 | 701 | 6–9, 925 |
It is meaningful to examine response time categories further, and the breakdowns of the percentages that fall into those categories. Response time to assault calls was categorized into asymmetrical minute intervals. The breakdown of response time categories included 0–1, 1.01–2, 2.01–3, 3.01–4, 4.01–5, 5.01–10, and 10.01 min or longer. According to Figure 1, almost 80% of in-progress assault calls fell within the 0–5 min response time category, only fewer than 3% required longer than 10 min for a patrol unit to arrive on the scene. In other words, the HPD responded to a greater percentage of calls as response times became less than five minutes.
Graph: Figure 1. Distribution of response to in-progress assault calls by the HPD in minutes.
To measure neighborhood social disorganization, 10 items (poverty, race and ethnicity, immigration, employment status, age composition, family structure, homeownership, and residential stability) were obtained from 2000 US Bureau of Census statistics (Cihan et al., [ 5]; Sampson et al., [18]). The Varimax rotated factor patterns for concentrated disadvantage, immigrant concentration, and residential stability for assault calls that occurred in 444 Houston census tracts are presented in Table 2.[ 4] Results from the analysis suggested that concentrated disadvantage, immigrant concentration, and residential stability could be retained for the analysis. Six items (below the poverty line, on public assistance, female-headed families, unemployed, less than age 18, and Black) were loaded on the factor of concentrated disadvantage. The factor loading of these items was .75 and greater, except the percentage of population less than age 18, which had a factor loading of.53. The percentage of Latino and foreign-born was loaded on the factor of immigrant concentration (factor loadings were.90 and greater). The last two items, the percentage of the population living in the same house as in 1995 and owner-occupied house, were loaded on the third factor of residential stability (factor loadings were.91 and.92, respectively).
Table 2. Varimax rotated factor patterns in Houston neighborhoods (United States Census Bureau, 2000).
| Variables | Factor 1 | Factor 2 | Factor 3 |
| Concentrated disadvantage | |||
| Below poverty line | .87* | .28 | −.09 |
| On public assistance | .88* | .03 | .08 |
| Female-headed families | .78* | −.18 | −.28 |
| Unemployed | .85* | .16 | .10 |
| Less than age 18 | .53* | .44 | .43 |
| Black | .75* | −.51 | .12 |
| Immigrant concentration | |||
| Latino | −.03 | .90** | −.28 |
| Foreign born | .11 | .95** | .05 |
| Residential stability | |||
| Same house as in 1995 | −.29 | −.17 | .91*** |
| Owner-occupied house | .18 | .17 | .92*** |
1 Note: Statistics are from the United States Census Bureau ([24]).
- 2 Factor 1 (Concentrated disadvantage).
- 3 Factor 2(Immigrant concentration).
- 4 Factor 3 (Residential stability).
Multilevel analysis
Hierarchical linear models were conducted to examine the effects of Level 1 (call characteristics) and Level 2 (neighborhood characteristics) variables on police response time to assault calls (the outcome variable) in Houston.[ 5] Table 3 shows the results of a one-way ANOVA (analysis of variance) test with a random effect or a baseline model. The baseline model provides preliminary information regarding the amount of variation in the outcome variable (response time) that lies within and between neighborhoods in Houston. The results of the one-way ANOVA suggest that police response time significantly varied across neighborhoods, accounting for 9.6% of the variance between neighborhoods.
Table 3. Hierarchical linear models of neighborhood and response characteristics on police response time to in-progress assault calls by the HPD.
| Estimated coefficients and errors | |||
| Parameters | Coefficient | SE | t-value |
| (LN) Police response time between tracts | |||
| Grand mean | 1.46 | .01 | 166.84** |
| Conditional error variance components | Variance |
χ
2 |
df |
| Between neighborhoods | .022 | 1835** | 443 |
| Within neighborhoods | .205 | ||
| Full model with neighborhood effects between neighborhoods | |||
| Grand mean | 1.49 | .01 | 166.62** |
| Concentrated disadvantage | −.02 | .01 | −2* |
| Immigration concentration | −.04 | .01 | −6.15** |
| Residential stability | .02 | .01 | 2.07* |
| (Ln) Call rates per 1000 citizens | −.09 | .01 | −5.97** |
| Within neighborhoods | |||
| Weekends | −.02 | .01 | −2.50* |
| 911 calls | .02 | .01 | 2.68** |
| Night-time | −.04 | .01 | −4.31** |
| Weapon | −.01 | .01 | −.02 |
| Conditional error variance components | Variance |
χ
2 |
df |
| Between neighborhoods | .015 | 1385.06** | 439 |
| Within neighborhoods | .205 |
- 5 p < .05.
- 6 p < .01.
A random-intercept model, including neighborhood-level variables and incident-level variables, was constructed to test the assumption that police response time to in-progress assault calls varied across Houston neighborhoods. Level 1 variables, including weekends, 911 calls, and night-time variables, were added to the model. The intercept of the Level 1 model, which measures the true mean level of police response time to in-progress assault calls, was assumed to vary at Level 2 (neighborhood level). In this model, the intercept of the Level 1 model became the outcome variable of the Level 2 model. All Level 2 variables (concentrated disadvantage, immigrant concentration, residential stability, and call rates per 1000 citizens) were added to the model as predictors of the mean response time to in-progress assault calls. Call characteristics were centered around the group mean, and the Level 2 variables were centered around the grand mean to disentangle the Level 2 effect from the Level 1 effect (Raudenbush & Bryk, [14]). In other words, this model isolates the neighborhood effects from the Level 1 effect and estimates the net effect of the level of neighborhood social disorganization on the distribution of response time. The second part of Table 3 shows the results from an examination of the random-intercept model incorporating community-level effects. All community-level indicators and neighborhood call rates per 1000 citizens were significantly associated with the response time. However, this model only accounted for 7% of the variance between neighborhoods.
Concentrated disadvantage (B = −.02, p < .05), immigration concentration (B = −.04, p < .01), and call rates per 1000 citizens (B = −.09, p < .01)were negatively related to police response time, suggesting that police response time was faster for neighborhoods with a high level of concentrated disadvantage, immigrant concentration, and increased call rates. However, residential stability (B = .02, p < .05) was positively associated with the outcome measure, indicating that police response time was slower in stable neighborhoods. Although the percentage of the variance between neighborhoods was very small, overall this model reduces the variance between neighborhoods (obtained from the ANOVA analysis, in Table 3).022 to.015, a 60% reduction in error variance between neighborhoods.
Results show that call characteristics were also related to police response time (see Table 3). While night-time (B = −.04, p < .01) and weekends (B = −.02, p < .05) were negatively related to an increase in police response time, 911 calls (B = .02, p < .01) were positively associated with response time. In other words, response time to assault calls was shorter during weekends and nights. Response time to 911 calls was slower when compared to 10-digit calls. Contrary to expectations, the presence of a weapon did not attain significance.
Discussion
Police response can be viewed as vital to police patrol operations. Based on previous research and theory, the current study examined the effects of neighborhood social disorganization on the distribution of police response. From the social disorganization perspective, neighborhoods with a high level of social disorganization are expected to receive disproportionate police activities and resources. Although some research has attempted to measure neighborhood-level forces that drive police activities and behavior, little is known about the linkage of neighborhood social disorganization with the distribution of response time. This can be viewed as an important oversight because the level of social disorganization is considered to be related to the distribution of police responses and activities. Analysis of the calls-for-service data suggests that social disorganization indicators (concentrated disadvantage, immigrant concentration, and residential stability) and neighborhood call rates are important predictors of the distribution of in-progress assault calls in Houston. Contrary to the social disorganization perspective, the results suggest that police responses were quicker to in-progress assault calls in neighborhoods with high disorganization and call rates. This is consistent with research that examined in-progress burglary response time patterns in Houston (Cihan et al., [ 5]). The distribution of police response patterns by neighborhood characteristics and call rates can provide police agencies with the feedback they can use to enhance police responsiveness by reconfiguring their patrol strategies across different neighborhoods.
The results also show that, except for the presence of a weapon, call characteristics were also related to the rapid police response. Police response time was faster during night-time and weekends, while 911 calls received a longer police response. Regarding the effect of night-time on police response time, the results were consistent with prior research (e.g. Cihan et al., [ 5]). It is possible that, during night-time, patrol units respond faster to the scene due to less traffic and more uncommitted patrol time. Overall, detailed response time analysis is crucial for agencies to improve patrol strategies and the police-community relationship.
Finally, it is important to note the limitations of the current study. The generalizability of results for the current study is limited because other major city police departments may have different policing strategies and response practices (Sampson, [16]; Sherman et al., [20]). Police response can also be influenced by a variety of factors, including neighborhood structural dynamics, allocation of police resources, police patrol deployment, the location of a patrol unit, traffic congestion, and police behavior (organizational and individual). Future studies may consider examining these factors that may influence police response.
Conclusion
Using agency-generated calls-for-service and census data, this article tested hypotheses about linkages among neighborhood disorganization (high level of concentrated disadvantage, immigrant concentration, and low residential stability), the level of neighborhood calls, and police response patterns. Findings, contrary to theoretical expectations, indicate that higher levels of neighborhood disorganization and calls are inversely related to police response patterns – higher levels of neighborhood disorganization and call rates are related to the shorter police response. Presumably, one would expect that higher levels of neighborhood disorganization and call rates reduce the availability of patrol units and increase police workload, which was not the case in Houston. Perhaps the outcome variable (response time) is context-specific, impacting the aforementioned relationship. In other words, the HPD seems to deploy patrol units proportional to demand, reducing response time in stressed neighborhoods. Overall, the results suggest that more attention ought to be paid to the theoretical relationship between neighborhood macro-sociological characteristics and the distribution of patrol activities (emergency availability, rapid response, and visible police presence) as formal control.
General Document Comments 0
icons on the left to see existing comments.
0 archived comments