48
observed for two groups for two time periods. One of the groups is exposed to a treatment in the
second period but not in the first period. The second group is not exposed to the treatment during
either period. In the case where the same units within a group are observed in each time period,
the average gain in the second (control) group is subtracted from the average gain in the first
(treatment) group. This removes biases in second period comparisons between the treatment and
control group that could result from permanent differences between those groups, as well as
biases from comparisons over time in the treatment group that could be the result of trends.
The outcome Yi is modeled by the following equation:
Yi = α + βTi + γti + δ (Ti · ti) + εi
where the coefficients given by the Greek letters α, β, γ, δ , are all unknown parameters and εi is
a random, unobserved "error" term which contains all determinants of Yi which our model omits.
The equation coefficients have the following interpretation:
α = constant term
β = treatment group specific effect (to account for average permanent differences between
treatment and control)
γ = time trend common to control and treatment groups
δ = true effect of treatment
In the current evaluation, we used a DiD panel regression design to assess whether the rate
of change in shots fired in the STD target areas (Sectors 11, 16, 22) was significantly different
than the non-target sectors between pre-intervention and post-intervention periods. Similar to the
models presented above, the DiD regression model of shots fired counts was estimated by
general linear modeling using the Poisson distribution. The analysis included a dummy variable
for STD vs. non-STD sectors and a dummy variable for pre and post intervention time periods.
49
The model also included controls for percent minority population, unemployment rates and
average monthly drug arrests.
Table 8 contains the post-intervention by target area interaction estimate (δ = true effect of
treatment) for 7.75 years (January 2005 – September 2012) of observations (n = 3,134). The
interaction estimate (δ = -.004, p = .910) indicates that there was no significant difference in the
relative change between pre-intervention and post-intervention periods in confirmed shots fired
between targeted and non-targeted sectors.
Table 8
DiD Poisson regression Results of Confirmed Shots Fired
Variable Coefficient
SE
IRR
p-value
Targeted Sectors
-.004 .035 .996 .910
Intercept
4.08 .001 --- .000
Table 9 contains the post-intervention by target area interaction estimate of 1, 242 robbery
observations. The interaction estimate (δ = -.013, p = .852) indicates that there was no significant
difference in the relative change between pre-intervention and post-intervention periods in
robberies between targeted and non-targeted sectors.
Table 9
DiD Poisson regression Results of Robberies
Variable Coefficient
SE
IRR
p-value
Targeted
Sectors
-.013 .071 .987 .852
Intercept
4.06 .021 --- .000
50
Conclusions
A number of studies have examined the application of problem oriented policing to hot
spots (Baker & Wolfer
2003
; Braga et al.
1999
; Braga & Bond
2008
; Mazerolle et al.
2000;
Weisburd & Green
1995
), and most of these have shown that the efforts reduced some forms of
crime and disorder. Results, however, have been more mixed with respect to violent crime
.
Studies by
Cohen and Ludwig (2003) and McGarrell, et al. (2001) found that directed police
patrol aimed at violent-crime hotspots help to reduce gun-related crime. On the other hand,
interventions studied by Mazerolle et al. (
2000
), Sherman, Gartin and Buerger (
1989)
and
Weisburd and Green (
1995
) did not reduce violence. Additionally, Sherman and Weisburd
(1995) and Taylor et al. (2011) found that the deterrent effects of police patrol in crime hot spots
for violent crimes were generally non-significant, though in the expected direction. Braga’s
(2007
) meta-analysis of results from five randomized experiments suggests that the effects of
hot-spot policing are most pronounced on disorderly behaviors; although violent and property
crimes declined on average across the studies, these effects were not statistically significant
overall. The lack of significant effects could reflect the impulsive, expressive nature of many
violent crimes (which may make them harder to prevent) and the rarity of violent crime in very
small locations. Alternatively, the particularly high concentration of violence in a relatively small
number of places would seem to weigh in favor of using hot spots strategies to curb violence.
With regard to the current evaluation, the results are consistent with those found in the
studies cited above (
Braga,
2007;
Mazerolle et al.,
2000;
Sherman, Gartin & Buerger,
1989;
Sherman & Weisburd, 1995; Taylor, et al., 2011; Weisburd & Green,
1995)
; that there were no
statistically significant effects of the intervention on violent crime (i.e., shots fired and
robberies).