44
and .118, respectively).
On the other hand, the data indicate increases in the number of robberies for both the STD
and STD+ components. The increase in robberies during the STD only period was significantly
different from the pre-STD time period (57% increase, p = .018). However, the increase in
robberies during the SDT+ intervention period did not differ significantly from the pre-program
period.
To further assess the potential impact of the STD strategy, we conducted a time series
analysis for shots fired in the STD target areas only. Specifically, monthly shots fired and
robberies were assessed for sectors 11, 16 and 22. Figure 6 indicates that the average number of
monthly shots fired increased by 18% when comparing the pre-STD to the STD time period and
20% when comparing the pre to the STD+ period.
Figure 6
Confirmed Shots Fired for STD Areas (Sectors 11, 16, 22),
January 2005 – September 2012
45
The data in Table 6 indicate that both the STD only and STD+ program components were
associated with decreases in the monthly number of shots fired. According to the incidence rate
ratios, the STD only component had a 22% percent decrease in the monthly number of shots
fired events and the STD+ component had a 27% reduction. However, these reductions were not
significantly different from the pre-intervention period. The data also indicate that the monthly
counts of robberies increased during the intervention period. Although not statistically
significant, robberies increase 59% and 97% during the STD only and STD+ program periods,
respectively. It is important to note that the overall model for monthly robberies was not
significant, thus indicating a poor model fit. Most likely, the poor model fit was due to low
monthly frequencies and too many zero counts.
Table 6
Poisson Regression Results for Shots Fired and Robbery
STD Areas (Sectors 11, 16, 22)
Shots Fired
Robbery
variable
B (SE)
IRR
p-value
B (SE)
IRR
p-value
trend
.003
(.007)
1.00 .659 -.008
(.013)
.992 .529
unemployment
.027 (.023)
1.03
.244
.013 (.045)
1.01
.766
% minority pop.
1.39 (1.88)
4.02
.459
.933 (3.04)
2.54
.759
# police officers
.000 (.003)
1.00
.993
.002 (.005)
1.00
.703
Monthly # drug
arrests
.006
(.007)
1.01 .358 -.006
(.011)
.994 .602
STD only
-.248 (.211)
.780
.238
.464 (.366)
1.59
.205
STD +
probation/parole
-.321
(.360)
.725 .372 .678
(.629)
1.97 .281
intercept
1.70 (.334)
---
.000
.539 (.489)
---
.270
46
Shots fired:
Deviance = 203.27, Pearson Chi-square = 191.24 Value/df = 2.39/2.25
Likelihood ratio Chi-Square = 14.09, df = 7, p = .050
Robbery:
Deviance = 169.56, Pearson Chi-square = 159.75 Value/df = 2.00/1.88
Likelihood ratio Chi-Square = 7.35, df = 7, p = .394
The same analysis was conducted for shots fired in the non-STD target areas only. Figure 7
indicates that the average number of monthly shots fired increased by 15% when comparing the
pre-STD to the STD time period and 4% when comparing the pre to the STD+ period.
Figure 7
Confirmed Shots Fired for Non-STD Areas, January 2005 – September 2012
The data in Table 7 show that neither the STD only nor the STD+ program components
were associated significantly with reductions in the monthly number of shots fired. The data do
indicate reductions in monthly shots fired, but these decreases were not significantly related to
the intervention. This is not surprising since these sectors were not the primary focus of the STD
intervention. In addition, there was a significant inverse relationship between the number of
Joliet police officers and shots fired in the non-STD areas.
47
Table 7
Poisson Regression Results for Shots Fired and Robbery, Non-STD Areas
Shots
Fired
Robbery
variable
B (SE)
IRR
p-value
B (SE)
IRR
p-value
trend
-.005 (.007)
.995
.529
-.003 (.007)
.997
.633
unemployment
.002
(.025)
1.00 .939 -.037
(.023)
.963 .103
% minority pop.
2.89 (1.87)
17.99
.122
1.29 (1.63)
3.64
.429
# police officers
-.005 (.002)
.995
.053
.001 (.002)
1.00
.765
Monthly # drug
arrests
-.004
(.004)
.996 .299 .006
(.004)
1.01 .107
STD only
-.084 (.217)
.919
.699
.557 (.216)
1.75
.010
STD +
probation/parole
-.047
(.379)
.954 .901 .518
(.387)
1.68 .181
intercept
3.13 (.231)
---
.000
1.33 (.304)
---
.000
Shots fired: Deviance = 177.40, Pearson Chi-square = 171.35 Value/df = 2.334/2.255
Likelihood ratio Chi-Square = 26.946, df = 7, p = .011
Robbery:
Deviance = 137.11, Pearson Chi-square = 129.20 Value/df = 1.61/1.52
Likelihood ratio Chi-Square = 38.12, df = 7, p = .000
The data in Table 4 also indicate increases in robberies for both the STD and STD+
components. There was a statistically significant increase (75%) in robberies during the STD-
only component as compared to the pre-STD time period. Similarly, though not statistically
significant, there was a 68% increase in monthly robbery counts during the STD+ period.
Difference-in-Difference Poisson Regression Models
The impact of program outcomes can be estimated by computing a double difference, one
over time (before-after) and one across subjects (between beneficiaries and non beneficiaries).
The basic logic behind this
difference-in-differences (
DiD) estimation is one where outcomes are
Dostları ilə paylaş: |