Measuring the Impacts of Everyday Police Proactive Activities: Tackling the Endogeneity Problem

Abstract

Objectives

To examine how the practice of daily proactivity affects and responds to changes in crime at micro geographic and temporal scales.

Methods

Police calls for service and automated vehicle location data from a large suburban jurisdiction were used to create comprehensive measures of police proactivity. Panel data and the generalized method of moments framework were applied to tease out the endogenous relationships between crime and police proactivity and understand the unique impact of proactive patrol and crime upon one another.

Results

Daily police proactivity in this locality was highly stable at micro places, although police did intensify their activities very briefly in response to recent changes in crime. In turn, increases in proactive patrol generated immediate increases in crime reporting, followed by fleeting residual deterrent effects that were weaker and less robust. The patterns remained relatively consistent when varying the units of analysis or focusing on hot spots with different profiles of proactivity, but the deterrent effects appeared more sensitive to model specification. Of all measures of proactivity, patrols of medium length and non-traffic enforcement activities were associated with stronger evidence of crime reduction effects.

Conclusions

Short-term adjustments in hot spot patrols appear to produce both reporting effects and temporary residual deterrent effects as measured through calls for service and police vehicle location data. Police could potentially enhance and prolong their deterrence by adopting more deliberate strategies with their daily proactive behaviors, including making their proactive activities more targeted and sustained.

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

Notes

  1. 1.

    Information retrieved from Koper et al. (2020).

  2. 2.

    Assuming that an officer calls in for a proactive activity (as noted, this can be uncertain), the length of a call as recorded by CAD could overestimate or underestimate the actual length of officer presence on scene for various reasons. Officers might close out a call after they have left the scene while writing their report elsewhere. They might also close out a call yet continue to stay in the area. Conversely, officers might mark out for proactive work at a hot spot as they are driving to the location. To provide some tentative insight into this issue, we looked at instances in which CAD suggests that officers were in the hot spot on a proactive call, but the AVL shows that they were somewhere else (either a non-hot spot location or another hot spot) for at least part of the “on scene” time record in CAD. We found that this happened in about 55% of proactive calls recorded in CAD. Determining the precise extent of this error in the measurement of time on-scene would require extensive calculations and may be a useful subject for future methodological research. Nonetheless, this basic illustration underscores the caveat that CAD-based time measures should be viewed as approximations of time on scene.

  3. 3.

    Hence, our focus is on police actions and impacts in the hot spots that were most active and stable during the study year.

  4. 4.

    Under the standard definition of Granger (1969) causality, for each individual \(i\in [1, N]\), a stationary variable \({x}_{i,t}\) is deemed to be “causing” another stationary variable \({y}_{i,t}\), if past values of \({x}_{i,t}\) increase the predicting power of the model on \({y}_{i,t}\) compared to using only past values of \({y}_{i,t}\). The original Granger Causality test was spelled out for individual time series data only, but scholars have extended the formula to panel data given the proliferation over time of panel data with large N and large T (Dumitrescu and Hurlin 2012). Individual effects are fixed in the time dimension, controlling for the effect of time-invariant variables (Allison 2005). Lag orders are assumed to be identical for all cross-section units of the panel and the panel should be balanced with an equal number of time points for each cross-sectional unit.

  5. 5.

    Panels with small T and large N are particularly susceptible to such biases. Judson and Owen (1999) suggest a 20% downward bias in the coefficient of a lagged dependent variable using within group transformation even when T = 30. That being said, the current study is likely less affected by the dynamic bias due to the relatively long time series (T = 52).

  6. 6.

    Depending on the underlying autocorrelation structure in the outcome series, one or several lagged terms of the dependent variable are included in the GMM models. Two test statistics are conducted specifically to assist with model identification. The Arellano-Bond test for ARs in first differences detects autocorrelation left in the differenced residuals. The test for an AR(1) process in first differences is expected to be significant, since the residuals of the differenced equation should possess serial correlation by construction (i.e., \({y}_{i,t-1}\) is correlated with \({v}_{i,t-1}\) in the error term in differences, \(\Delta {\varepsilon }_{it}={v}_{it}-{v}_{i,t-1}\), by construction). The test for an AR(2) process, however, should not reject the null hypothesis because it tests for serial correlation in levels (e.g., when the AR(2) term is significant, \({y}_{i,t-2}\) would be endogenous to \({v}_{i,t-1}\) in the error term in differences, \(\Delta {\varepsilon }_{it}={v}_{it}-{v}_{i,t-1}\), and therefore becomes an invalid instrument). Adding higher-order lags of the dependent variable (and occasionally the independent variables) as regressors typically removes the serial correlation. In addition, the Sargan/Hansen and the difference-in-Hansen tests are part of the standard output following system GMM estimation and are typically used to understand the validity of the model specification and whether the instruments or subsets of instruments are exogenous. Insignificant test statistics are preferred as they suggest that the model specifications are valid.

  7. 7.

    Back-up activities, in particular, occurred in about 17% of officer-initiated calls in the jurisdiction, with an average of 1.3 officers attending a proactive officer CAD call (these back-ups are not included in the CAD measure of proactivity). Adjusting for this, roughly 24 of the 31 lengthier AVL visits might have been independent in nature.

  8. 8.

    Note that any impact of the lagged independent variable is a partial effect, namely, the effect of a one-unit change in past week’s crime or proactivity on current proactivity or crime, holding everything else in the model unchanged. Past proactivity, for example, might also affect past crime, which, in turn, affects current crime. While it may be possible to quantify the long run effect of the predictor, the study focuses on the direct impact of crime and proactivity that go beyond the first week.

  9. 9.

    The grid cell approach is tested with CAD data only to limit to a reasonable extent the extensiveness of the robustness analysis.

References

  1. Allison PD (2005) Fixed effects regression methods for longitudinal data using SAS. Sas Institute

  2. Allison PD (2009) Fixed effects regression models, vol 160. SAGE publications, Thousand Oaks

    Google Scholar 

  3. Arellano M, Bover O (1995) Another look at the instrumental variable estimation of error-components models. J Econ 68(1):29–51

    Article  Google Scholar 

  4. Ariel B, Weinborn C, Sherman LW (2016) “Soft” policing at hot spots—do police community support officers work? A randomized controlled trial. J Exp Criminol 12(3):277–317

    Article  Google Scholar 

  5. Barnes, G. C., Williams, S., Sherman, L. W., Parmar, J., House, P., & Brown, S. A. (2020). Sweet spots of residual deterrence: A randomized crossover experiment in minimalist police patrol. https://osf.io/preprints/socarxiv/kwf98/.

  6. Blundell R, Bond S (1998) Initial conditions and moment restrictions in dynamic panel data models. J Econom 87(1):115–143

    Article  Google Scholar 

  7. Braga AA, Bond BJ (2008) Policing crime and disorder hot spots: a randomized controlled trial. Criminology 46(3):577–607

    Article  Google Scholar 

  8. Braga AA, Turchan BS, Papachristos AV, Hureau DM (2019) Hot spots policing and crime reduction: an update of an ongoing systematic review and meta-analysis. J Exp Criminol 15(3):289–311

    Article  Google Scholar 

  9. Braga AA, Welsh BC, Schnell C (2015) Can policing disorder reduce crime? A systematic review and meta-analysis. J Res Crime Delinq 52(4):567–588

    Article  Google Scholar 

  10. Carter JG, Piza EL (2017) Spatiotemporal convergence of crime and vehicle crash hotspots: additional consideration for policing places. Crime Delinq 64(14):1795–819

    Article  Google Scholar 

  11. Dumitrescu EI, Hurlin C (2012) Testing for Granger non-causality in heterogeneous panels. Econ Model 29(4):1450–1460

    Article  Google Scholar 

  12. Gorr WL, Lee Y (2015) Early warning system for temporary crime hot spots. J Quant Criminol 31(1):25–47

    Article  Google Scholar 

  13. Granger CW (1969) Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37(3):424–438

    Article  Google Scholar 

  14. Hansen GR, Leflang WG (1976) Application of automatic vehicle location in law enforcement: An introductory planning guide. US Department of Justice, Washington, DC

    Google Scholar 

  15. Hutt O, Bowers K, Johnson S, Davies T (2018) Data and evidence challenges facing place-based policing. Policing Int J Police Strat Manag 41(3):339–351

    Article  Google Scholar 

  16. Judson RA, Owen AL (1999) Estimating dynamic panel data models: a guide for macroeconomists. Econ Lett 65(1):9–15

    Article  Google Scholar 

  17. Kelling GL, Pate T, Dieckman D, Brown CE (1974) The Kansas City preventive patrol experiment. Police Foundation, Washington, DC

    Google Scholar 

  18. Koper CS (1995) Just enough police presence: reducing crime and disorderly behavior by optimizing patrol time in crime hot spots. Justice Q 12(4):649–672

    Article  Google Scholar 

  19. Koper CS, Lum C, Wu X, Fritz N (2020) Proactive policing in the United States: a national survey. Policing 43(5):861–876

    Article  Google Scholar 

  20. Koper C, Taylor BG, Woods D (2013) A Randomized test of initial and residual deterrence from directed patrols and use of license plate readers at crime hot spots. J Exp Criminol 9(2):213–244

    Article  Google Scholar 

  21. Law Enforcement Assistance Administration (1969) Annual report of the Law Enforcement Assistance Administration (LEAA) for the fiscal year 1969. U.S. Department of Justice, Washington, DC

  22. Lum C, Koper CS (2017) Evidence-based policing: translating research into practice. Oxford University Press, Oxford, UK

    Google Scholar 

  23. Lum C, Koper CS, Telep CW (2011) The evidence-based policing matrix. J Exp Criminol 7(1):3–26

    Article  Google Scholar 

  24. Lum C, Koper CS, Wu X, Johnson W, Stoltz M (2018) The proactive policing lab. Final Report to the Laura and John Arnold Foundation. George Mason University, Fairfax, VA

  25. Lum C, Koper CS, Wu X, Johnson W, Stoltz M (2020) Examining the empirical realities of proactive policing through systematic observations and computer-aided dispatch data. Police Q 23(3):283–310

    Article  Google Scholar 

  26. Mastrofski SD, Parks RB, Reiss AJ Jr, Worden RE, DeJong C, Snipes JB, Terrill W (1998) Systematic observation of public police: applying field research methods to policy issues. National Institute of Justice, Washington, DC

    Google Scholar 

  27. Mitchell RJ (2017) Frequency versus duration of police patrol visits for reducing crime in hot spots: non-experimental findings from the sacramento hot spots experiment. Camb J Evid Based Polic 1(1):22–37

    Article  Google Scholar 

  28. National Academies of Sciences Engineering and Medicine (2018) Proactive policing: effects on crime and communities. National Academies Press, Washington

    Google Scholar 

  29. National Research Council (2004) Fairness and effectiveness in policing: the evidence. National Academies Press, Washington

    Google Scholar 

  30. Nickell S (1981) Biases in dynamic models with fixed effects. Econom J Econom Soc 49(6):1417–1426

    Google Scholar 

  31. Roodman D (2007) A short note on the theme of too many instruments. Center for Global Development Working Paper, 125.

  32. Roodman D (2009) How to do xtabond2: an introduction to difference and system GMM in Stata. Stata J 9(1):86–136

    Article  Google Scholar 

  33. Santos RG, Santos RB (2015) An ex post facto evaluation of tactical police response in residential theft from vehicle micro-time hot spots. J Quant Criminol 31(4):679–698

    Article  Google Scholar 

  34. Sherman LW, Eck JE (2002) Policing for crime prevention: evidence-based crime prevention. Taylor and Francis, Abingdon, p 295

    Google Scholar 

  35. Sherman LW, Gartin PR, Buerger ME (1989) Hot spots of predatory crime: routine activities and the criminology of place. Criminology 27(1):27–56

    Article  Google Scholar 

  36. Sherman LW, Rogan DP (1995) Effects of gun seizures on gun violence: “Hot spots” patrol in Kansas City. Justice Q 12(4):673–693

    Article  Google Scholar 

  37. Sherman LW, Weisburd D (1995) General deterrent effects of police patrol in crime “hot spots”: a randomized, controlled trial. Justice Q 12(4):625–648

    Article  Google Scholar 

  38. Simpson R, Hipp JR (2019) What came first: the police or the incident? Bidirectional relationships between police actions and police incidents. Polic Soc 29(7):783–801

    Article  Google Scholar 

  39. Telep CW, Mitchell RJ, Weisburd D (2014) How much time should the police spend at crime hot spots? Answers from a police agency directed randomized field trial in Sacramento, California. Justice Q 31(5):905–933

    Article  Google Scholar 

  40. Weisburd D, Bruinsma GJ, Bernasco W (2009) Units of analysis in geographic criminology: Historical development, critical issues, and open questions. In Putting crime in its place (pp. 3–31). Springer, New York, NY

  41. Weisburd D, Bushway S, Lum C, Yang SM (2004) Trajectories of crime at places: a longitudinal study of street segments in the city of Seattle. Criminology 42(2):283–322

    Article  Google Scholar 

  42. Weisburd D, Eck JE (2004) What can police do to reduce crime, disorder, and fear? Ann Am Acad Polit Social Sci 593(1):42–65

    Article  Google Scholar 

  43. Weisburd D, Gill C, Wooditch A, Barritt W, Murphy J (2020) Building collective action at crime hot spots: Findings from a randomized field experiment. J Exp Criminol 2020:1–31

    Google Scholar 

  44. Weisburd D, Groff ER, Jones G, Cave B, Amendola KL, Yang SM, Emison RF (2015) The Dallas patrol management experiment: can AVL technologies be used to harness unallocated patrol time for crime prevention? J Exp Criminol 11(3):367–391

    Article  Google Scholar 

  45. Weisburd D, Wooditch A, Weisburd S, Yang SM (2016) Do stop, question, and frisk practices deter crime? Evidence at microunits of space and time. Criminol Public Policy 15(1):31–56

    Article  Google Scholar 

  46. Williams S, Coupe T (2017) Frequency vs length of hot spots patrols: a randomised controlled trial. Camb J Evid Based Polic 1(1):5–21

    Article  Google Scholar 

  47. Wu X, Lum C (2017) Measuring the spatial and temporal patterns of police proactivity. J Quant Criminol 33(4):915–934

    Article  Google Scholar 

  48. Wu X, Lum C (2019) The practice of proactive traffic stops. Policing 43(2):229–246

    Article  Google Scholar 

Download references

Author information

Affiliations

Authors

Corresponding author

Correspondence to Xiaoyun Wu.

Additional information

Publisher's Note

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

Appendices

Appendix A

See Table 6.

Table 6 Types of calls included in the measure of crime and disorder, serious crime, and proactivity

Appendix B

See Table 7.

Table 7 Unit root test for crime and proactivity series

Appendix C

See Table 8.

Table 8 Granger causality tests for serious crime and proactivity (adjusted z values reported)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Wu, X., Koper, C. & Lum, C. Measuring the Impacts of Everyday Police Proactive Activities: Tackling the Endogeneity Problem. J Quant Criminol (2021). https://doi.org/10.1007/s10940-021-09496-8

Download citation

Keywords

  • Police proactivity
  • Police everyday practice
  • Deterrence
  • GMM
  • Reverse causality