Neighbourhood context and unsolved murders: the social ecology of homicide investigations

ABSTRACT

Prior research indicates that victim and crime characteristics shape homicide investigations, yet less is known about the effects of neighbourhood demographics. Drawing upon several theoretical perspectives, this study examines the neighbourhood context of homicide investigations. Multilevel logistic regressions disentangle the effects of covariates at the agency, neighbourhood, and individual level on homicide investigations in Los Angeles County, California. While several non-racial factors influence clearance outcomes, homicides occurring in areas with larger Black and Latino populations are less likely to be cleared. These findings highlight the importance of neighbourhood racial composition beyond victim race effects and have implications for the community context of criminal justice.

KEYWORDS:

• Communities and crime
• community context and criminal justice
• homicide clearance
• homicide arrest

Acknowledgements

The author thanks the anonymous reviewers for their constructive feedback. I would also wish to thank Mona Lynch, Ann Hironaka, and Carroll Seron for their insightful comments on earlier drafts. Special thanks to Dick O'Reilly, Doug Smith, and Sandra Poindexter for helping me access the data.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 For brevity, ‘race’ is used instead of ‘race/ethnicity’, while the term ‘Latino’ rather than ‘Latino/a’ is used since most homicide victims are male.

2 Despite differences in the criminalisation of Blacks and Latinos, there are several points of convergence between the two groups. As such, I focus on the similarities between these groups given that ‘The public picture of Latinos and crime most closely resembles that of blacks. Latinos, too, are viewed as stealthy and criminal … ’ (Rome 2004, pp. 4–5).

3 While some studies find that racial composition does not affect crime stereotypes (Perkins and Taylor 2002, Mohan et al. 2011), these findings are fairly robust (Moeller 1989, Covington and Taylor 1991, Skogan 1995, Chiricos et al. 1997, Chiricos and Eschholz 2002, Welch 2007, Franzini et al. 2008, Scarborough et al. 2010, Quillian and Pager 2010, Drakulich 2012).

4 Riedel (2008) offers a comprehensive review of the homicide arrest literature. A number of other studies address these issues (e.g., Regoeczi et al. 2000, Puckett and Lundman 2003, Litwin 2004, Lee 2005, Addington 2007, Alderden and Lavery 2007, Jiao 2007, Litwin and Xu 2007, Roberts 2007, 2014, Regoeczi et al. 2008, Roberts and Lyons 2011, Lundman and Myers 2012, Regoeczi and Jarvis 2013).

5 Accidental, vehicular, and justifiable homicides are excluded from this analysis. Of the 9442 homicides that occurred during the study period, 1364 were not geo-coded due to missing or incomplete address information. These cases were handled using list-wise deletion.

6 Publicly available homicide data sets do not contain neighbourhood information (e.g. State Court Processing Statistics; Offender-Based Transaction Statistics; Pennsylvania Sentencing Data; Supplementary Homicide Reports; Changing Patterns of Homicide and Social Policy, etc.).

7 Recognising that neighbourhoods are difficult to operationalise because ‘the very notion of a neighbourhood has been assigned a range of conceptual and operational definitions’ (Bellair 2000, Wells et al. 2006, p. 531), I use census tracts as proxies for neighbourhoods. Census tracts not only offer a standardised geographic unit of analysis (Quillian and Pager 2001, Slocum et al. 2010), but also tap into police and resident perceptions of neighbourhoods (Brooks-Gunn et al. 1993, Greene 2000). Moreover, census tracts are commonly used in the neighbourhood literature, including studies of policing (Jackson & Boyd, Jacobs and Helms, 1996, Jacobs and O'Brien 1998), homicide victimisation (Krivo and Peterson 1996, Kubrin and Herting 2003, Griffiths and Chavez 2004), police–community relations (Weitzer 1999, Sampson and Raudenbush 2004, Sampson 2009, Slocum et al. 2010), and criminal justice (Clear et al. 2003, Lynch and Sabol 2004, Wooldredge and Thistlethwaite 2004, 2002).

8 Consistent with prior research (Pare et al. 2007, Ousey and Lee 2010, Roberts 2014), police workload was calculated using the total number of sworn officers, rather than the number of homicide investigators, since all sworn officers can potentially contribute to clearance outcomes (Chaiken 1975, Cordner 1989).

9 As diagnostic measures, variance inflation factor (VIF) statistics were calculated based on single-level OLS versions of Model 2 (mean VIF = 2.04) and Model 4 (mean VIF = 2.11) with robust standard errors. Given that the VIF statistics from these ‘generic’ models are below standard cut-off levels (Hair et al. 1995), they suggest that standard errors are not biased by multicollinearity.

10 Recognising the potential interplay between victim demographics and social contextual factors (Wooldredge 2007), a supplementary model, not included here but available upon frequent, examined cross-level interactions. In this supplementary model, cross-level interactions between victim race and racial composition were not significant, indicating that neighbourhood effects do not depend upon victim race.

Additional information

Funding

This project was partially supported by the Proteus Fund [grant number 102938].