Examining the neighborhood effects on police performance to assault calls

Abstract

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

Acknowledgments

The author would like to thank Larry Hoover and Yan Zhang for their help and providing data; the editor, managing editor, and anonymous reviewers for their valuable comments; and Kimberly Chism for editing this article.

Notes

1. The sum of Houston’s population percentage exceeds 100% because of considerations for individuals with multi-ethnic backgrounds.

2. It is also important to note that a smaller unit of analysis (i.e. street segments and block groups) could also be used as an alternative geographic unit for the current study. However, it is highly unlikely that calls-for-service data provide the exact location of the incident being reported. According to Sherman et al. (1989), in most instances, the location of the actual incident differs from the exact location of the caller or the location to which police are being dispatched. For this reason, using calls-for-service data at census tract level led to a better a proxy measure of neighborhood level social disorganization than using smaller geographic units (Sherman et al., 1989).

3. The calls-for-service per 1000 individuals ratio was calculated with the aggregated number of calls-for-service at each census tract divided by census tract population and the quotient multiplied by 1000. While this measure can be limited because officer or shift-specific calls-for-service information and neighborhood-level crime rates were not available at the census tract level, previous literature suggests that certain neighborhoods produce disproportionately more calls-for-service than others (see Sherman et al., 1989). Future research should consider the effects of actual crime rates and shift-specific call rates on police response time.

4. The smallest census tract has 288 people, while the largest tract has 15,411 people. The final sample includes 444 census tracts with an average tract population of 5091.

5. It is also important to note that the measurement for calls-for-service rates per 1000 citizens (a Level 2 variable) and police response time (the outcome variable) were highly skewed. In order to improve ‘the normality of error distribution’ (Raudenbush & Bryk, 2002, p. 267), and to meet the assumptions of multilevel modeling, these variables were transformed through natural logarithmic function.