![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
The current study examined the situational and community factors that affect police use of nonlethal force in New York City. Although a number of studies explored either situational factors at the incident-level or community factors at the neighbourhood level, only a few tested for interaction between the two levels. Using multiple data sources, the current study features hierarchical generalised linear modelling for both independent and cross-level interaction effects of situational and community factors. The findings suggest that the police use of nonlethal force was associated with suspect's race and other situational factors. Notably, the effects of race and seriousness of the offence appear to hinge upon the community characteristics. Findings are further discussed for policy implications.
Notes
1. The author thanks an anonymous reviewer for the advice to compare the findings from sub-samples with the results from the original sample.
2. Because incidents were previously linked to corresponding census tracts, they were automatically linked to corresponding community districts.
3. Concerned about biased estimation of regression coefficients, it was determined that female suspect cases (6.9%) be excluded. For the same reason, Asians (2.9%) and Native Indians cases (0.4%), and cases involving police use of firearm (0.5%) were excluded.
4. In total, there are 59 community districts in NYC. However, three community districts in Manhattan were excluded because of the overlap between the boundary of a few precincts and community districts.
5. Findings concerning the prevalence of police use of force were consistent with previous findings on police use of force in NYC and other settings, showing that the police rarely use physical force during an encounter. When police use physical force, they commonly use only the low level of force such as the use of bare hands to search or control the suspect (Garner et al. Citation1995, Terrill and Mastrofski Citation2002, Schuck Citation2004, Hickman et al. Citation2008).
6. The percentages of black and of Hispanic suspects are examined separately because the Hispanic population in NYC has increased rapidly in the past decade. Thus, this population change may have some effect on police behaviour (e.g., police use of force). In addition, previous research found that the number or percentage of Hispanic people had some effect on police use of force in some areas (Smith and Holmes Citation2003).
7. Census data provide a few socio-economic indicators (e.g., % of residents under poverty line, % of residents receiving public assistance), and each of these indicators as well as their composite scale has a strong correlation with % of Black (Pearson's r > 0.6) and % of Hispanic population (Pearson's r > 0.6). More importantly, supplementary regression analysis confirmed the presence of multicollinearity that incorporating both socio-economic indicator and race proportion measures was problematic. Accordingly, the current study elected to use violent crime rate as a proxy for neighbourhood disadvantage.
8. Because regression-based modelling is sensitive to multicollinearity between variables, its test was conducted. The variance inflation factors of all independent variables were <5, suggesting the absence of multicollinearity.
9. Logit function does not allow error term at the incident-level.
10. u0j was assumed to be normally distributed around a mean of zero and variance of τ00.
11. Initially, proportions of other races were included at the community-level to assess the log-odds of nonlethal force use in reference to a wide range of community characteristics. However, as it appears in the , NYC appears to have an interesting pattern of racial/ethnic concentration across communities. That is, a strong negative correlation exists between races at the community-level. While such correlation may seem plausible in that these indices are proportions and add up to 1 (100%), three-way classification (i.e., white, black and Hispanic) could illustrate different patterns where low or non-significant correlation exists between two of the three races. Given the racial concentration and resultant strong correlations between races, slight modification of modelling was inevitable such that only the proportion of suspect's same race in the community is included at the community-level. Nonetheless, because high proportion of a race translates to low proportions of the other races, the modelling still addresses how racial dominance (or lack thereof) interacts with the suspect's race in the matter of police use of nonlethal force.
12. Although race dummy variables (i.e., black and Hispanic) were structured to represent white suspects at the intercept, the use of other multiple dummy variables in the current modelling undesirably over-specified the intercept. Specifically, the intercept represents a white suspect, who was compliant and was not suspected of Part I offences, in an incident that occurred during the daytime and outside of the residence that the officer perceived to be a safe area. Therefore, alternative modelling was administered with white and Hispanic dummy variables such that the intercept would indicate the log-odds of use of nonlethal force for a black suspect in the aforementioned specification. Findings from this alternative modelling also suggested that a white suspect was less likely to experience nonlethal force. Taken together, a police officer was less likely to use nonlethal force in an incident with a white suspect.
13. The average here refers to a mean of violent crime rate because violent crime rate at the community district level was mean-centred.
14. To confirm the reduction in variance component, an ad-hoc test was administered to examine cross-community variance in the nonlethal force use with intercept (white), black and Hispanic at the incident-level. Variance components were significant at p = 0.05 for the effects of white (i.e., intercept) and black, while not for Hispanic.