Skip to main content

Building collective action at crime hot spots: Findings from a randomized field experiment

Abstract

Objectives

The study examined whether Assets Coming Together (ACT), a policing intervention directed at increasing collective action and collective efficacy at crime hot spots in Brooklyn Park, Minnesota, would have impacts on these outcomes, as well as police legitimacy, crime and fear of crime.

Methods

We used a block-randomized experimental design in which hot spots of crime were randomly allocated to treatment and control conditions. The treatment condition received the ACT program, and the control condition received normal police response. We analyzed crime data using an ANOVA approach, taking into account treatment and block. We analyzed survey data collected at each hot spot using mixed-effects linear regression models with robust standard errors to account for the nesting of responses within hot spots.

Results

We find that the intervention increased citizen reporting of collective actions (including collaboration in problem solving and contacts with the police) at hot spots, but it had little impact on general measures of collective efficacy or police legitimacy. Fear of crime increased at the treatment sites. We found that crime reporting was significantly inflated in the treatment sites. Crime outcomes were non-significant without accounting for this reporting inflation, but the treatment areas had a significant crime decrease when adjusting estimates based on reporting inflation.

Conclusions

Our experimental findings show that collective actions at hot spots can be encouraged through programs like ACT and that ordinary policing resources—patrol officers in this case—can be successfully used to carry out such programs. We find preliminary evidence that the program also impacted crime. At the same time, our study points to a bias in using official crime data to assess outcomes in programs that encourage community collaboration.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3

Notes

  1. All population and demographic information are taken from the American Community Survey 2012–2016 5-year estimates, “Community Facts,” http://factfinder.census.gov (accessed September 7, 2018).

  2. Our analysis of dispatch data during the planning phase of the project showed that BPPD officers have on average 3 h 51 min of discretionary time during a 12-h shift (approximately one third of the shift).

  3. The intervention was initially intended to last for a year, but BPPD extended the implementation using their own funds to provide additional time for problem-solving and data for analysis.

  4. In a block-randomized experiment, blocks can be removed from the study without affecting the overall assumptions of an experimental design (see Weisburd and Gill 2014).

  5. For a detailed description of the study methods, see the technical report for this project (Weisburd et al., 2018a).

  6. However, we did include some civil complaints that reflected social disorganization and informal social control, including disorder, neighbor disputes, fighting, animal complaints, and noise violations.

  7. Commercial streets were included as long as there was a business community to work with (i.e., a block that contained a large national chain retailer where staff would not be able to work with police without permission from corporate headquarters was not included).

  8. This variable did not reduce the heterogeneity in the other factors and was therefore unnecessary. Each blocking factor reduces the degrees of freedom and can limit statistical power (Weisburd and Gill 2014).

  9. Mean monthly citizen-initiated CFS in the treatment sites = 12.2 (SD = 18.5); control sites = 11.4 (SD = 10.1). Mean monthly crime incidents in treatment sites = 4.3 (SD = 6.8); control sites = 4.1 (SD = 4.0).

  10. We used an intra-class correlation of .40 to estimate the power of the block randomized study (based on Gill and Weisburd 2013) using CRT Power software (Borenstein et al. 2012) and assumed equal block size because of computational limitations.

  11. Again, we assume an ICC of .40 for the block level, but an ICC of only .02 for the hot spot level (these are all hot spots of crime, and we expect them to have relatively disadvantaged residents living across the sites).

  12. Independent samples t tests indicate no significant differences at baseline between study conditions for both CFS and crime incidents. t42 = − .442, p = .661 and t42 = − .639, p = .527, respectively.

  13. See Weisburd et al. (2018a) for full technical details of the survey.

  14. Some business locations, such as small strip malls, had fewer addresses than our target number of surveys (for example, one of our hot spots consisted of a street segment with just three businesses). In these cases, the researchers attempted to interview multiple employees within the same business.

  15. We drew a new sample of addresses for the follow-up survey, but six addresses were surveyed in both Wave 1 and Wave 2. Seventeen respondents in the follow-up survey said they remembered taking the survey before, although we cannot verify they were correctly remembering our survey. In most cases, we only conducted one survey per address; however, a few hot spots had fewer than seven addresses on the street, and at others, multiple people were willing to take the survey (this was usually the case at business addresses where a number of employees were working). Only 24 individuals (7.6% of all individuals surveyed) were nested in addresses at Waves 1 and 2 (3.7%) at Wave 2. As a sensitivity analysis, we also ran the models including random effects for individuals and addresses. Due to the lack of variability, the random effect for individuals created instability in the models. The random effect for address also did not contribute to several of the models, but where we were able to include it, the results were very similar to the models presented here.

  16. The inflation factors were almost identical when we calculated them excluding Block 2 (1.67 in the treatment group and 1.26 in the control group), so we use the above inflation factor in later analyses, whether or not Block 2 is included.

  17. Because of the very high correlation between block and logged pre-intervention crime, we also ran the models for the adjusted crime outcomes with block or logged pre-intervention crime excluded. For the model with block excluded, the observed one-tailed p value was .047 for the full sample. The p value excluding pre-intervention crime was .099, still significant at the .10 level.

  18. Nagin and Sampson (2019) have raised the question of whether scaling up interventions like ACT creates interference between units, making it difficult to assess jurisdictional impact. We think it important to note that we expect that these types of programs implemented on a large scale would still be restricted to the most intractable hot spots in a city. The resources involved in applying treatment make scaling up much beyond that prohibitive.

References

Download references

Acknowledgments

The opinions, recommendations, and conclusions herein are those of the authors and do not necessarily reflect the position of the U.S. Department of Justice. We would like to thank Chip Coldren, Craig Uchida, Robert Sampson, and Shellie Solomon for their advice and support in developing and implementing this study.

Funding

This research was funded by the U.S. Department of Justice, Bureau of Justice Assistance under Award Number 2013-DB-BX-0030.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to David Weisburd.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Weisburd, D., Gill, C., Wooditch, A. et al. Building collective action at crime hot spots: Findings from a randomized field experiment. J Exp Criminol 17, 161–191 (2021). https://doi.org/10.1007/s11292-019-09401-1

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11292-019-09401-1

Keywords

  • Community policing
  • Collective efficacy
  • Collective action
  • Hot spots
  • Block randomized
  • Experiment