SlideShare a Scribd company logo
1 of 8
Download to read offline
Automated Speed Enforcement Cameras Effect on Traffic
Speed: a Chicago Case Study
Andrew W. Szmurlo
December 15, 2016
Abstract
Using public data from the City of Chicago, I analyze the effect of installing automated
speed enforcement cameras on traffic speed.
1 Introduction
The City of Chicago’s Children’s Safety Zone Program is intended to protect children and other
pedestrians by slowing motorists in established school and park safety zones. The City has installed
Speed Enforcement Cameras in these areas to encourage drivers to obey speed laws. The objective
of reducing traffic speeds is to reduce death and disability resulting from high speed crashes. When
the number of speeding motorists decreases, then the probability and severity of accidents will
decrease. A multitude of studies have been conducted to measure the effectiveness of enforcement
cameras in both reducing traffic speed and the resulting collisions from speeding motorists (Wilson
et al.).There has been rather intense public discourse over the effectiveness of Chicago’s program
and the Chicago Tribune has undertaken several studies to measure different outcomes (Bordens et
al.). I develop a difference-in-differences model to study the effect of installing a speed enforcement
camera (along a road segment) on a measure of traffic speed: Chicago Transit Authority’s bus
data. This paper contributes to the literature by using a novel dataset of bus speeds along road
segments; it measures the effect of cameras installed on traffic speed – comparing both before/after
installation, before/after activation, and road segments that receive the camera against those that
do not (constituting a difference-in-differences model). The different estimation results stem from
different cameras analyzed (through zone types or through the number of violations recorded).
2 Data
2.1 Traffic Speed
The speed variable analyzed is a daily average of all traffic speeds recorded along a specific road
segment (in one direction of travel) for that day. There is public data for two time periods (Chicago
Traffic Tracker):
1. January 15th through February 28th of 2013 (45 Days)
2. November 25th through December 30th of 2014 (36 Days)
The availability of only four months of data is not ideal: only two periods separated by one year,
but the majority of cameras, for which we have data, were installed in-between these two time pe-
riods (84 cameras). One camera was installed before traffic data was recorded, and seven cameras
were installed after both time periods (Speed Camera Locations).
The traffic measurements are recorded by the City of Chicago, which monitors GPS traces re-
ceived from the Chicago Transit Authority (CTA) buses.The data-set used is the "City of Chicago:
Chicago Traffic Tracker - Historical Congestion Estimates by Segments." It contains the estimated
speeds of 1026 unique half mile segments of streets in one direction of traffic. This covers 300
miles of arterial roads (non-freeway streets). Note: the City of Chicago considers 0-9 miles per
hour (mph) as heavy traffic, 10-20 mph as medium, and 21 mph or higher to be free flow traffic
1
conditions. All bus estimates with less than five message readings (of the GPS) have been dropped
from the data for improved accuracy, which is consistent with previous studies (Lin and Pu).
There is large volatility in traffic segment speed due to multiple factors: frequent intersections,
traffic signals, transit movements, availability of alternative routes, crashes, or short length of the
segments are examples. Therefore, I have chosen to take the average (and median) daily speed
as the dependent variable in these regressions. While, this estimation does not reflect well the
reduction in speeding motorists, it does provide an analysis of whether overall traffic is slower due
to automated speed cameras.
2.2 Automated Speed Enforcement Cameras
The location and activation date of the Automated Speed Enforcement Cameras was pulled from
the "City of Chicago: Speed Camera Locations" dataset. It contains the address, direction of
traffic, activation date and GPS coordinates for 150 speed cameras.This paper matches the speed
cameras to their corresponding road segments by address and direction of traffic (92 cameras were
matched to corresponding road segments).
In the City of Chicago, Automated Speed Enforcement Cameras are only installed in Children’s
Safety Zones. These zones comprise a 1/8th-mile perimeter around every school and park. Safety
Zone times and speed limits vary based on type:
1. School Zones are regulated on school days, Monday through Friday. From 7 a.m. to 4 p.m.
the speed limit is 20 miles per hour when children are present and 30 miles per hour when
absent. From 4 p.m. to 7 p.m. the speed returns to the normal posted speed limit.
2. Park Zones are regulated every day. Times for these zones vary, but are normally enforced 6
a.m. to 11 p.m. at a 30 miles per hour speed limit.
For the first thirty days after a camera’s installation, or if it is their first infraction, vehicles that
exceed the regulated limit are provided with a warning notice. The fine for a speeding violation of
six mph over is 35 dollars, but the City of Chicago only issues a ticket if the vehicle reaches ten
mph over the posted speed limit. If the vehicle exceeds eleven mph, then the fee is increased to
100 dollars.
The tracking systems used by the City of Chicago have a high-resolution digital camera, a
high-definition video camera, and a 3D tracking radar to identify vehicles traveling faster than
the posted speed limit. The cameras record the event. Data captured by this system are the
following: license plate photo, HD video providing evidence, time, date, posted speed limit, vehicle
speed, location, land, and the direction of travel. The cameras are equipped with higher-resolution
imaging compared to other traffic management programs serviced by the vendor, American Traffic
Solutions, to allow for the ticket-reviewing technicians to zoom and search for present children;
this is a requirement for the reduced speed limit in School Zones.
The public was initially informed of the speed camera implementation with a 90-day public
awareness campaign including public service announcements, advertisements, and radar signs in
safety zones to indicate driving speeds. The public is informed of new camera locations through
posted signage in each Children’s Safety Zone when cameras are deployed, the City of Chicago
website will be updated weekly, and the aforementioned 30-day warning period before cameras
begin ticketing.
Illinois state law allots only 20 percent of the 1,500 Safety Zones for potential speed camera
locations. Therefore, only (approximately) 300 locations can be in effect. Furthermore, The City
has partitioned Chicago into six geographical regions where no fewer than 10 percent of speed
enforcement cameras will be located in each region. This ensures an even spread of Speed Enforce-
ment Cameras around the City of Chicago. The first cameras were installed in August 2013 with
the first violations being issued in September.
The locations are chosen based on traffic, speeding, and accident data. There is an advisory
committee on this decision, which includes both the Chairman and Vice-Chairman of the Pedestrian
2
and Traffic Safety Committee along with leadership from the religious, academic, public safety, and
medical communities.
3 Analysis
3.1 Model
Yit = βTit + γAit + αRi + ηDt + it
The dependent variable: Y is the average (or median) daily speed across all bus readings for each
road segment measured by the CTA.
The indicator variable T describes a treatment where T=1 if an automated speed enforcement
camera has been installed along the segment and T=0 otherwise. β is the corresponding estimator.
The indicator variable A describes a treatment where A=1 if an automated speed enforcement
camera has been activated on that road segment and A=0 otherwise. Note that a camera is only
active after it has been installed. Park cameras are active every day after they are installed, but
school cameras are only active on weekdays. γ is the corresponding estimator.
R is a vector of dummy variables corresponding to each individual road segment, whereas α is
the corresponding vector of estimators for each road segment. Therefore, α can be considered the
fixed effect estimator of the model. It accounts for unobserved time-invariant effects in this model.
The subscript i denotes an individual road segment.
D is a vector of dummy variables corresponding to each day and η is the corresponding vector
of estimators for each individual day. This time fixed effect accounts for variation specific to each
day. The subscript t denotes an individual day.
The last term: considers accounts for the residual effects that impact traffic speed.
The results for this initial model are within figures 4 and 5. In the OLS regression (1), R and
D are left out of the model. This is because ordinary least squares does not include fixed nor time
effects. However, the difference-in-differences models (2), (3), (4), (5), and (6) all include fixed and
time effects. Therefore, R and D are included. Note that all 6 models are based on this model.
The changes across the last 5 columns result from the different cameras used in estimation. (2)
includes all cameras, (3) includes only cameras installed in park zones (38 cameras), (4) includes
cameras only installed in school zones (46 cameras), (5) includes cameras that record the highest
number of violations (24 cameras in the top third of recorded violations have corresponding traffic
data), and (6) includes only the cameras that record the least number of violations (22 cameras
in the lowest third of recorded violations have corresponding traffic data). The camera activation
variable (A) is omitted for the third column because, once installed, they are active everyday.
Standard errors are clustered by road segment because it is likely that unobserved error terms are
correlated across time for each road segment (road speeds will likely be similar across time due to
posted speed limits and road qualities inherent to specific segments).
The OLS model describes the highest decrease in traffic speed due to camera installations
and camera activation, but note the installation coefficient is insignificant. The more accurate
difference-in-differences regression describes a significant reduction in speed of approximately one
mile per hour due to the combined effect of camera installation and activation; both are significant.
Note the small changes between (3) and (4) when accounting for the combined effect of activation
and instillation. This indicates that reductions in speed are not different dependent on whether
the camera is installed in a park zone or a school zone, but that the activation of a camera does
contribute to a reduction in traffic speed. Median daily speed regressions discount the extreme
values. In this paper my focus is on speeding motorists (extreme values). Therefore, I believe
figure 4 more relevant.
3
3.2 Potential Sources of Bias
A major issue stems from the potential endogeneity between the installment of Automated Speed
Enforcement Cameras and other "tools" the City of Chicago has used in its Children’s Safety Zone
Program. As these cameras are only installed in the Safety Zones, this paper cannot separate
the speed camera’s effect on traffic speed from the other methods of safety enhancement, such as:
pedestrian refuge islands, safety zone signage and street stencils, high-visibility crosswalk mark-
ings, speed feedback signs, speed humps, traffic signal improvements, curb and ramp improvements,
pedestrian countdown timers, lead pedestrian intervals, and in-street "Stop for Pedestrians" signs.
While it seems unlikely that the City of Chicago would install a speed hump and speed camera
in the same road segment, other road improvements (e.g. pedestrian refuge islands and street
stencils) could slow traffic speeds creating a downward bias in this model. To remedy this bias, I
would need to lobby the City for detailed data on road improvements or visit each road segment
where a traffic camera has been installed and record other changes that had been made to the road
segment. However, these records are not available for the current draft of this paper.
Another source of bias results from the traffic data from the Chicago Transit Authority. This
data only measures the speed of Chicago City buses. Note: while bus probe data is a good reflec-
tion of car speed, buses are at least 1.6 miles per hour slower than cars on arterial streets (Lin and
Pu). This indicates that this model has an upward bias (meaning the true effect of cameras on
traffic speed has a larger negative magnitude) because, assuming car drivers fear receiving a ticket,
they will have to slow a greater amount from their average speed to achieve the same speed limit
that bus drivers reach to not be ticketed.
Another issue with the bus data is novel to the City of Chicago: bus drivers working for the
CTA are not required to pay the corresponding ticket from the speeding violation resulting from
Automated Speed Enforcement Cameras. While bus drivers originally had to pay these fines, their
union sued the CTA on "doubly-jeopardy grounds" because the drivers are also disciplined inter-
nally. CTA drivers can receive up to four safety violations before being fired, but other infractions
are considered safety violations. Not forcing bus drivers to pay their tickets would reduce the
negative coefficient pertaining to installing a speed camera, but the fear of losing their job could
be a greater incentive than a 35 (or 100) dollar ticket. In their analysis, the Chicago Tribune found
714 speed camera violations among CTA bus and minibus drivers.
The Speed Enforcement Cameras also have their issues. When analyzing the public data on
speed violations, the Chicago Tribune found 62,000 tickets issued during summer months in School
Zones. School is not in session during the summer months. While the public traffic data used in
this analysis only pertains to November, December, January, and February; this issue of inaccurate
ticketing could pertain to weekends as well. Mayor of Chicago Rahm Emanuel’s administration
officials have defined schooldays (for ticketing purposes) to mean any day when a child is present
for class, even if not a scheduled school day. If schools hold attendance on weekends (e.g. extracur-
ricular activities or private schools holding different calendars), then A, the indicator describing
camera activation may not accurately reflect whether a camera is active.
Also, for the lower speed limit to be enforced in School Safety Zones, schoolchildren must be
present. The camera does not recognize whether children are present. Three camera technicians
must review the video and photographs of the incident before ticketing the registered vehicle owner,
yet Chicago hearing officers regularly note a lack of children when tossing out speeding tickets.
Therefore, total camera violations may overstate the true numbers of people receiving tickets.
However, this uncertainty in ticketing policy could also engender fear in drivers and overall force
them to drive slowly in road segments with speed cameras.
4 Conclusion
The City of Chicago’s Automated Speed Enforcement Cameras play a role in reducing traffic
speeds. The data estimated in this paper stems from Chicago Transit Authority bus estimates;
therefore, the true reduction in speeds are likely greater. The installation of an active camera
along a specific road segment slows average daily bus speeds by approximately one mile per hour.
4
Figure 1: Automated Speed Enforcement Signage
Figure 2: Automated Speed Enforcement Signage
It is unlikely that automated speed enforcement cameras slow motorists already driving below the
speed limit; therefore, it is likely that cameras slow speeding drivers to a much greater extent. This
analysis does not portray a reduction in the number of speeders, but instead overall daily traffic.
There are not significant differences between speed reductions in park and school zones. Overall,
automated speed enforcement cameras slow traffic.
5
Figure 3: Camera Process
(1) (2) (3) (4) (5) (6)
VARIABLES OLS FE & TE Park School High Low
Camera Installation -0.761 -0.492*** -1.032*** -0.500*** -0.655 -0.647**
(0.464) (0.152) (0.201) (0.166) (0.500) (0.273)
Camera Activation -1.731*** -0.553*** -0.563*** -0.530 -0.388**
(0.335) (0.131) (0.130) (0.473) (0.162)
Constant 23.67*** 24.03*** 24.09*** 24.10*** 24.12*** 24.10***
(0.181) (0.0493) (0.0513) (0.0508) (0.0519) (0.0521)
Number of Observations 78,277 78,277 74,318 74,831 72,722 72,563
R-squared 0.006 0.088 0.080 0.083 0.078 0.078
Road Segment FE No Yes Yes Yes Yes Yes
Daily TE No Yes Yes Yes Yes Yes
Number of Road Segments 1,027 978 984 958 956
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Figure 4: Results – Mean Daily Speed
(1) (2) (3) (4) (5) (6)
VARIABLES OLS FE & TE Park School High Low
Camera Installation -0.825* -0.611*** -1.081*** -0.630*** -0.482 -0.776**
(0.488) (0.177) (0.220) (0.195) (0.584) (0.354)
Camera Activation -1.792*** -0.516*** -0.549*** -0.750 -0.456**
(0.355) (0.151) (0.153) (0.545) (0.195)
Constant 23.59*** 23.86*** 23.91*** 23.93*** 23.95*** 23.93***
(0.190) (0.0541) (0.0562) (0.0558) (0.0569) (0.0570)
Observations 78,277 78,277 74,318 74,831 72,722 72,563
R-squared 0.006 0.084 0.076 0.079 0.074 0.074
Road Segment FE No Yes Yes Yes Yes Yes
Daily TE No Yes Yes Yes Yes Yes
Number of Road Segments 1,027 978 984 958 956
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Figure 5: Results – Median Daily Speed
6
Works Cited
Bordens, Alex, Abraham Epton, Geoff Hing, and David Kidwell. "Speed Cameras." Speeding
Tickets Questioned. Chicago Tribune, n.d. Web. 12 Dec. 2016.
<http://apps.chicagotribune.com/news/local/chicago-speed-camera-tickets/>.
"Chicago Traffic Tracker - Congestion Estimates by Segments." Chicago Traffic Tracker -
Congestion Estimates by Segments. City of Chicago, 08 Dec. 2016. Web. 12 Dec. 2016.
<https://catalog.data.gov/dataset/chicago-traffic-tracker-congestion-estimates-by-
segments-0b9d7>.
Kidwell, David, and Abraham Epton. "Chicago Speed Cameras Catch School Bus Drivers,
Police Cars, CTA Operators." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016.
<http://www.chicagotribune.com/news/watchdog/redlight/ct-met-chicago-speed-cameras-
government-met-20151228-story.html>.
Kidwell, David, and Abraham Epton. "Emanuel's Speed Cameras Issue $2.4 Million in Bad
Tickets." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016.
<http://www.chicagotribune.com/news/watchdog/ct-speed-camera-bad-tickets-met-
20151117-story.html>.
Kidwell, David, and Abraham Epton. "Many School-zone Speed Camera Tickets Issued without
Danger to Kids, Tribune Finds." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016.
<http://www.chicagotribune.com/news/watchdog/ct-speed-camera-school-tickets-met-
20151118-story.html>.
Kidwell, David, and Abraham Epton. "Speed Cameras near Tiny Playlots, Closed Parks Lead to
Big Payout." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016.
<http://www.chicagotribune.com/news/watchdog/ct-speed-camera-park-tickets-met-
20151118-story.html>.
Kidwell, David, and Abraham Epton. "Top-ticketing Speed Cameras Not Where Most Kids Have
Been Hit, Tribune Finds." Chicago Tribune, 05 Jan. 2016. Web. 12 Dec. 2016.
<http://www.chicagotribune.com/news/watchdog/ct-speed-camera-tickets-safety-met-
20151120-story.html>.
Pu, Wenjing, and Jie Lin. "Urban Travel Time Estimation Using Real Time Bus Tracking Data."
Transport Chicago (n.d.): n. pag. Web. 12 Dec. 2016.
<http://www.transportchicago.org/uploads/5/7/2/0/5720074/intelligentbus-pulin.pdf>.
"Speed Camera Locations." City of Chicago, n.d. Web. 12 Dec. 2016.
<https://data.cityofchicago.org/Transportation/Speed-Camera-Locations/4i42-
qv3h/about>.
"Speed Camera Violations." Speed Camera Violations. City of Chicago, 08 Dec. 2016. Web. 12
Dec. 2016. <https://catalog.data.gov/dataset/speed-camera-violations-997eb>.
Wilson, C., C. Willis, JK Hendrikz, R. Le Brocque, and N. Bellamy. "Do Speed Cameras Reduce
Road Traffic Crashes, Injuries and Deaths?" National Center for Biotechnology
Information. U.S. National Library of Medicine, 01 Jan. 1970. Web. 12 Dec. 2016.
<https://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0012902/>.

More Related Content

Viewers also liked

Presentación computación
Presentación computaciónPresentación computación
Presentación computaciónNathys-15
 
Hintergrundpapier zu verringerten ausbauzielen für erneuerbare energien
Hintergrundpapier zu verringerten ausbauzielen für erneuerbare energienHintergrundpapier zu verringerten ausbauzielen für erneuerbare energien
Hintergrundpapier zu verringerten ausbauzielen für erneuerbare energienmetropolsolar
 
Exe learning arbol de contenidos-uch
Exe learning   arbol de contenidos-uchExe learning   arbol de contenidos-uch
Exe learning arbol de contenidos-uchGladys Flores Hurtado
 
I sistemi di Web Content Filtering
I sistemi di Web Content FilteringI sistemi di Web Content Filtering
I sistemi di Web Content Filteringmmarcuzzi
 

Viewers also liked (8)

Presentación computación
Presentación computaciónPresentación computación
Presentación computación
 
Hintergrundpapier zu verringerten ausbauzielen für erneuerbare energien
Hintergrundpapier zu verringerten ausbauzielen für erneuerbare energienHintergrundpapier zu verringerten ausbauzielen für erneuerbare energien
Hintergrundpapier zu verringerten ausbauzielen für erneuerbare energien
 
Exe learning arbol de contenidos-uch
Exe learning   arbol de contenidos-uchExe learning   arbol de contenidos-uch
Exe learning arbol de contenidos-uch
 
Charla Debian en Servidores
Charla Debian en ServidoresCharla Debian en Servidores
Charla Debian en Servidores
 
I sistemi di Web Content Filtering
I sistemi di Web Content FilteringI sistemi di Web Content Filtering
I sistemi di Web Content Filtering
 
MongoDB for Developers
MongoDB for DevelopersMongoDB for Developers
MongoDB for Developers
 
Jenkins в docker in mesos in ...
Jenkins в docker in mesos in ...Jenkins в docker in mesos in ...
Jenkins в docker in mesos in ...
 
1.2 terbitan berseri
1.2 terbitan berseri1.2 terbitan berseri
1.2 terbitan berseri
 

Similar to Master's Project -- Andrew W. Szmurlo

A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...
A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...
A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...ITIIIndustries
 
Control of Traffic Signals by AI based Image Processing
Control of Traffic Signals by AI based Image ProcessingControl of Traffic Signals by AI based Image Processing
Control of Traffic Signals by AI based Image ProcessingIRJET Journal
 
Traffic Light Controller System using Optical Flow Estimation
Traffic Light Controller System using Optical Flow EstimationTraffic Light Controller System using Optical Flow Estimation
Traffic Light Controller System using Optical Flow EstimationEditor IJCATR
 
JiaxuZhou_GRAPoster
JiaxuZhou_GRAPosterJiaxuZhou_GRAPoster
JiaxuZhou_GRAPosterJiaxu Zhou
 
Real time deep-learning based traffic volume count for high-traffic urban art...
Real time deep-learning based traffic volume count for high-traffic urban art...Real time deep-learning based traffic volume count for high-traffic urban art...
Real time deep-learning based traffic volume count for high-traffic urban art...Conference Papers
 
Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...
Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...
Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...Biplav Srivastava
 
Vehicle detection and tracking techniques a concise review
Vehicle detection and tracking techniques  a concise reviewVehicle detection and tracking techniques  a concise review
Vehicle detection and tracking techniques a concise reviewsipij
 
LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...
LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...
LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...cscpconf
 
A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...
A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...
A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...JANAK TRIVEDI
 
IRJET- Simulation based Automatic Traffic Controlling System
IRJET- Simulation based Automatic Traffic Controlling SystemIRJET- Simulation based Automatic Traffic Controlling System
IRJET- Simulation based Automatic Traffic Controlling SystemIRJET Journal
 
P ERFORMANCE M EASUREMENTS OF F EATURE T RACKING AND H ISTOGRAM BASED T ...
P ERFORMANCE  M EASUREMENTS OF  F EATURE  T RACKING AND  H ISTOGRAM BASED  T ...P ERFORMANCE  M EASUREMENTS OF  F EATURE  T RACKING AND  H ISTOGRAM BASED  T ...
P ERFORMANCE M EASUREMENTS OF F EATURE T RACKING AND H ISTOGRAM BASED T ...ijcsit
 
Traffic Light Detection for Red Light Violation System
Traffic Light Detection for Red Light Violation SystemTraffic Light Detection for Red Light Violation System
Traffic Light Detection for Red Light Violation Systemijtsrd
 
Vehicle Traffic Analysis using CNN Algorithm
Vehicle Traffic Analysis using CNN AlgorithmVehicle Traffic Analysis using CNN Algorithm
Vehicle Traffic Analysis using CNN AlgorithmIRJET Journal
 
Vehicle tracking system using python project.pptx
Vehicle tracking system using python project.pptxVehicle tracking system using python project.pptx
Vehicle tracking system using python project.pptxparashuram430
 
IRJET- Intelligent Traffic Management System
IRJET- Intelligent Traffic Management SystemIRJET- Intelligent Traffic Management System
IRJET- Intelligent Traffic Management SystemIRJET Journal
 
Smart Traffic Monitoring System based on vehicle counts.
Smart Traffic Monitoring System based on vehicle counts.Smart Traffic Monitoring System based on vehicle counts.
Smart Traffic Monitoring System based on vehicle counts.IRJET Journal
 
Iain Peat-06012564-main report
Iain Peat-06012564-main reportIain Peat-06012564-main report
Iain Peat-06012564-main reportIain Peat
 
Route optimization using network analyst tools of arcgis(mid term evaluation)...
Route optimization using network analyst tools of arcgis(mid term evaluation)...Route optimization using network analyst tools of arcgis(mid term evaluation)...
Route optimization using network analyst tools of arcgis(mid term evaluation)...PRABHATKUMAR751
 

Similar to Master's Project -- Andrew W. Szmurlo (20)

A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...
A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...
A Method for Predicting Vehicles Motion Based on Road Scene Reconstruction an...
 
Control of Traffic Signals by AI based Image Processing
Control of Traffic Signals by AI based Image ProcessingControl of Traffic Signals by AI based Image Processing
Control of Traffic Signals by AI based Image Processing
 
A real-time system for vehicle detection with shadow removal and vehicle clas...
A real-time system for vehicle detection with shadow removal and vehicle clas...A real-time system for vehicle detection with shadow removal and vehicle clas...
A real-time system for vehicle detection with shadow removal and vehicle clas...
 
Traffic Light Controller System using Optical Flow Estimation
Traffic Light Controller System using Optical Flow EstimationTraffic Light Controller System using Optical Flow Estimation
Traffic Light Controller System using Optical Flow Estimation
 
JiaxuZhou_GRAPoster
JiaxuZhou_GRAPosterJiaxuZhou_GRAPoster
JiaxuZhou_GRAPoster
 
Real time deep-learning based traffic volume count for high-traffic urban art...
Real time deep-learning based traffic volume count for high-traffic urban art...Real time deep-learning based traffic volume count for high-traffic urban art...
Real time deep-learning based traffic volume count for high-traffic urban art...
 
Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...
Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...
Case Studies in Managing Traffic in a Developing Country with Privacy-Preserv...
 
proceedings of PSG NCIICT
proceedings of PSG NCIICTproceedings of PSG NCIICT
proceedings of PSG NCIICT
 
Vehicle detection and tracking techniques a concise review
Vehicle detection and tracking techniques  a concise reviewVehicle detection and tracking techniques  a concise review
Vehicle detection and tracking techniques a concise review
 
LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...
LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...
LANE CHANGE DETECTION AND TRACKING FOR A SAFE-LANE APPROACH IN REAL TIME VISI...
 
A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...
A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...
A VISION-BASED REAL-TIME ADAPTIVE TRAFFIC LIGHT CONTROL SYSTEM USING VEHICULA...
 
IRJET- Simulation based Automatic Traffic Controlling System
IRJET- Simulation based Automatic Traffic Controlling SystemIRJET- Simulation based Automatic Traffic Controlling System
IRJET- Simulation based Automatic Traffic Controlling System
 
P ERFORMANCE M EASUREMENTS OF F EATURE T RACKING AND H ISTOGRAM BASED T ...
P ERFORMANCE  M EASUREMENTS OF  F EATURE  T RACKING AND  H ISTOGRAM BASED  T ...P ERFORMANCE  M EASUREMENTS OF  F EATURE  T RACKING AND  H ISTOGRAM BASED  T ...
P ERFORMANCE M EASUREMENTS OF F EATURE T RACKING AND H ISTOGRAM BASED T ...
 
Traffic Light Detection for Red Light Violation System
Traffic Light Detection for Red Light Violation SystemTraffic Light Detection for Red Light Violation System
Traffic Light Detection for Red Light Violation System
 
Vehicle Traffic Analysis using CNN Algorithm
Vehicle Traffic Analysis using CNN AlgorithmVehicle Traffic Analysis using CNN Algorithm
Vehicle Traffic Analysis using CNN Algorithm
 
Vehicle tracking system using python project.pptx
Vehicle tracking system using python project.pptxVehicle tracking system using python project.pptx
Vehicle tracking system using python project.pptx
 
IRJET- Intelligent Traffic Management System
IRJET- Intelligent Traffic Management SystemIRJET- Intelligent Traffic Management System
IRJET- Intelligent Traffic Management System
 
Smart Traffic Monitoring System based on vehicle counts.
Smart Traffic Monitoring System based on vehicle counts.Smart Traffic Monitoring System based on vehicle counts.
Smart Traffic Monitoring System based on vehicle counts.
 
Iain Peat-06012564-main report
Iain Peat-06012564-main reportIain Peat-06012564-main report
Iain Peat-06012564-main report
 
Route optimization using network analyst tools of arcgis(mid term evaluation)...
Route optimization using network analyst tools of arcgis(mid term evaluation)...Route optimization using network analyst tools of arcgis(mid term evaluation)...
Route optimization using network analyst tools of arcgis(mid term evaluation)...
 

Master's Project -- Andrew W. Szmurlo

  • 1. Automated Speed Enforcement Cameras Effect on Traffic Speed: a Chicago Case Study Andrew W. Szmurlo December 15, 2016 Abstract Using public data from the City of Chicago, I analyze the effect of installing automated speed enforcement cameras on traffic speed. 1 Introduction The City of Chicago’s Children’s Safety Zone Program is intended to protect children and other pedestrians by slowing motorists in established school and park safety zones. The City has installed Speed Enforcement Cameras in these areas to encourage drivers to obey speed laws. The objective of reducing traffic speeds is to reduce death and disability resulting from high speed crashes. When the number of speeding motorists decreases, then the probability and severity of accidents will decrease. A multitude of studies have been conducted to measure the effectiveness of enforcement cameras in both reducing traffic speed and the resulting collisions from speeding motorists (Wilson et al.).There has been rather intense public discourse over the effectiveness of Chicago’s program and the Chicago Tribune has undertaken several studies to measure different outcomes (Bordens et al.). I develop a difference-in-differences model to study the effect of installing a speed enforcement camera (along a road segment) on a measure of traffic speed: Chicago Transit Authority’s bus data. This paper contributes to the literature by using a novel dataset of bus speeds along road segments; it measures the effect of cameras installed on traffic speed – comparing both before/after installation, before/after activation, and road segments that receive the camera against those that do not (constituting a difference-in-differences model). The different estimation results stem from different cameras analyzed (through zone types or through the number of violations recorded). 2 Data 2.1 Traffic Speed The speed variable analyzed is a daily average of all traffic speeds recorded along a specific road segment (in one direction of travel) for that day. There is public data for two time periods (Chicago Traffic Tracker): 1. January 15th through February 28th of 2013 (45 Days) 2. November 25th through December 30th of 2014 (36 Days) The availability of only four months of data is not ideal: only two periods separated by one year, but the majority of cameras, for which we have data, were installed in-between these two time pe- riods (84 cameras). One camera was installed before traffic data was recorded, and seven cameras were installed after both time periods (Speed Camera Locations). The traffic measurements are recorded by the City of Chicago, which monitors GPS traces re- ceived from the Chicago Transit Authority (CTA) buses.The data-set used is the "City of Chicago: Chicago Traffic Tracker - Historical Congestion Estimates by Segments." It contains the estimated speeds of 1026 unique half mile segments of streets in one direction of traffic. This covers 300 miles of arterial roads (non-freeway streets). Note: the City of Chicago considers 0-9 miles per hour (mph) as heavy traffic, 10-20 mph as medium, and 21 mph or higher to be free flow traffic 1
  • 2. conditions. All bus estimates with less than five message readings (of the GPS) have been dropped from the data for improved accuracy, which is consistent with previous studies (Lin and Pu). There is large volatility in traffic segment speed due to multiple factors: frequent intersections, traffic signals, transit movements, availability of alternative routes, crashes, or short length of the segments are examples. Therefore, I have chosen to take the average (and median) daily speed as the dependent variable in these regressions. While, this estimation does not reflect well the reduction in speeding motorists, it does provide an analysis of whether overall traffic is slower due to automated speed cameras. 2.2 Automated Speed Enforcement Cameras The location and activation date of the Automated Speed Enforcement Cameras was pulled from the "City of Chicago: Speed Camera Locations" dataset. It contains the address, direction of traffic, activation date and GPS coordinates for 150 speed cameras.This paper matches the speed cameras to their corresponding road segments by address and direction of traffic (92 cameras were matched to corresponding road segments). In the City of Chicago, Automated Speed Enforcement Cameras are only installed in Children’s Safety Zones. These zones comprise a 1/8th-mile perimeter around every school and park. Safety Zone times and speed limits vary based on type: 1. School Zones are regulated on school days, Monday through Friday. From 7 a.m. to 4 p.m. the speed limit is 20 miles per hour when children are present and 30 miles per hour when absent. From 4 p.m. to 7 p.m. the speed returns to the normal posted speed limit. 2. Park Zones are regulated every day. Times for these zones vary, but are normally enforced 6 a.m. to 11 p.m. at a 30 miles per hour speed limit. For the first thirty days after a camera’s installation, or if it is their first infraction, vehicles that exceed the regulated limit are provided with a warning notice. The fine for a speeding violation of six mph over is 35 dollars, but the City of Chicago only issues a ticket if the vehicle reaches ten mph over the posted speed limit. If the vehicle exceeds eleven mph, then the fee is increased to 100 dollars. The tracking systems used by the City of Chicago have a high-resolution digital camera, a high-definition video camera, and a 3D tracking radar to identify vehicles traveling faster than the posted speed limit. The cameras record the event. Data captured by this system are the following: license plate photo, HD video providing evidence, time, date, posted speed limit, vehicle speed, location, land, and the direction of travel. The cameras are equipped with higher-resolution imaging compared to other traffic management programs serviced by the vendor, American Traffic Solutions, to allow for the ticket-reviewing technicians to zoom and search for present children; this is a requirement for the reduced speed limit in School Zones. The public was initially informed of the speed camera implementation with a 90-day public awareness campaign including public service announcements, advertisements, and radar signs in safety zones to indicate driving speeds. The public is informed of new camera locations through posted signage in each Children’s Safety Zone when cameras are deployed, the City of Chicago website will be updated weekly, and the aforementioned 30-day warning period before cameras begin ticketing. Illinois state law allots only 20 percent of the 1,500 Safety Zones for potential speed camera locations. Therefore, only (approximately) 300 locations can be in effect. Furthermore, The City has partitioned Chicago into six geographical regions where no fewer than 10 percent of speed enforcement cameras will be located in each region. This ensures an even spread of Speed Enforce- ment Cameras around the City of Chicago. The first cameras were installed in August 2013 with the first violations being issued in September. The locations are chosen based on traffic, speeding, and accident data. There is an advisory committee on this decision, which includes both the Chairman and Vice-Chairman of the Pedestrian 2
  • 3. and Traffic Safety Committee along with leadership from the religious, academic, public safety, and medical communities. 3 Analysis 3.1 Model Yit = βTit + γAit + αRi + ηDt + it The dependent variable: Y is the average (or median) daily speed across all bus readings for each road segment measured by the CTA. The indicator variable T describes a treatment where T=1 if an automated speed enforcement camera has been installed along the segment and T=0 otherwise. β is the corresponding estimator. The indicator variable A describes a treatment where A=1 if an automated speed enforcement camera has been activated on that road segment and A=0 otherwise. Note that a camera is only active after it has been installed. Park cameras are active every day after they are installed, but school cameras are only active on weekdays. γ is the corresponding estimator. R is a vector of dummy variables corresponding to each individual road segment, whereas α is the corresponding vector of estimators for each road segment. Therefore, α can be considered the fixed effect estimator of the model. It accounts for unobserved time-invariant effects in this model. The subscript i denotes an individual road segment. D is a vector of dummy variables corresponding to each day and η is the corresponding vector of estimators for each individual day. This time fixed effect accounts for variation specific to each day. The subscript t denotes an individual day. The last term: considers accounts for the residual effects that impact traffic speed. The results for this initial model are within figures 4 and 5. In the OLS regression (1), R and D are left out of the model. This is because ordinary least squares does not include fixed nor time effects. However, the difference-in-differences models (2), (3), (4), (5), and (6) all include fixed and time effects. Therefore, R and D are included. Note that all 6 models are based on this model. The changes across the last 5 columns result from the different cameras used in estimation. (2) includes all cameras, (3) includes only cameras installed in park zones (38 cameras), (4) includes cameras only installed in school zones (46 cameras), (5) includes cameras that record the highest number of violations (24 cameras in the top third of recorded violations have corresponding traffic data), and (6) includes only the cameras that record the least number of violations (22 cameras in the lowest third of recorded violations have corresponding traffic data). The camera activation variable (A) is omitted for the third column because, once installed, they are active everyday. Standard errors are clustered by road segment because it is likely that unobserved error terms are correlated across time for each road segment (road speeds will likely be similar across time due to posted speed limits and road qualities inherent to specific segments). The OLS model describes the highest decrease in traffic speed due to camera installations and camera activation, but note the installation coefficient is insignificant. The more accurate difference-in-differences regression describes a significant reduction in speed of approximately one mile per hour due to the combined effect of camera installation and activation; both are significant. Note the small changes between (3) and (4) when accounting for the combined effect of activation and instillation. This indicates that reductions in speed are not different dependent on whether the camera is installed in a park zone or a school zone, but that the activation of a camera does contribute to a reduction in traffic speed. Median daily speed regressions discount the extreme values. In this paper my focus is on speeding motorists (extreme values). Therefore, I believe figure 4 more relevant. 3
  • 4. 3.2 Potential Sources of Bias A major issue stems from the potential endogeneity between the installment of Automated Speed Enforcement Cameras and other "tools" the City of Chicago has used in its Children’s Safety Zone Program. As these cameras are only installed in the Safety Zones, this paper cannot separate the speed camera’s effect on traffic speed from the other methods of safety enhancement, such as: pedestrian refuge islands, safety zone signage and street stencils, high-visibility crosswalk mark- ings, speed feedback signs, speed humps, traffic signal improvements, curb and ramp improvements, pedestrian countdown timers, lead pedestrian intervals, and in-street "Stop for Pedestrians" signs. While it seems unlikely that the City of Chicago would install a speed hump and speed camera in the same road segment, other road improvements (e.g. pedestrian refuge islands and street stencils) could slow traffic speeds creating a downward bias in this model. To remedy this bias, I would need to lobby the City for detailed data on road improvements or visit each road segment where a traffic camera has been installed and record other changes that had been made to the road segment. However, these records are not available for the current draft of this paper. Another source of bias results from the traffic data from the Chicago Transit Authority. This data only measures the speed of Chicago City buses. Note: while bus probe data is a good reflec- tion of car speed, buses are at least 1.6 miles per hour slower than cars on arterial streets (Lin and Pu). This indicates that this model has an upward bias (meaning the true effect of cameras on traffic speed has a larger negative magnitude) because, assuming car drivers fear receiving a ticket, they will have to slow a greater amount from their average speed to achieve the same speed limit that bus drivers reach to not be ticketed. Another issue with the bus data is novel to the City of Chicago: bus drivers working for the CTA are not required to pay the corresponding ticket from the speeding violation resulting from Automated Speed Enforcement Cameras. While bus drivers originally had to pay these fines, their union sued the CTA on "doubly-jeopardy grounds" because the drivers are also disciplined inter- nally. CTA drivers can receive up to four safety violations before being fired, but other infractions are considered safety violations. Not forcing bus drivers to pay their tickets would reduce the negative coefficient pertaining to installing a speed camera, but the fear of losing their job could be a greater incentive than a 35 (or 100) dollar ticket. In their analysis, the Chicago Tribune found 714 speed camera violations among CTA bus and minibus drivers. The Speed Enforcement Cameras also have their issues. When analyzing the public data on speed violations, the Chicago Tribune found 62,000 tickets issued during summer months in School Zones. School is not in session during the summer months. While the public traffic data used in this analysis only pertains to November, December, January, and February; this issue of inaccurate ticketing could pertain to weekends as well. Mayor of Chicago Rahm Emanuel’s administration officials have defined schooldays (for ticketing purposes) to mean any day when a child is present for class, even if not a scheduled school day. If schools hold attendance on weekends (e.g. extracur- ricular activities or private schools holding different calendars), then A, the indicator describing camera activation may not accurately reflect whether a camera is active. Also, for the lower speed limit to be enforced in School Safety Zones, schoolchildren must be present. The camera does not recognize whether children are present. Three camera technicians must review the video and photographs of the incident before ticketing the registered vehicle owner, yet Chicago hearing officers regularly note a lack of children when tossing out speeding tickets. Therefore, total camera violations may overstate the true numbers of people receiving tickets. However, this uncertainty in ticketing policy could also engender fear in drivers and overall force them to drive slowly in road segments with speed cameras. 4 Conclusion The City of Chicago’s Automated Speed Enforcement Cameras play a role in reducing traffic speeds. The data estimated in this paper stems from Chicago Transit Authority bus estimates; therefore, the true reduction in speeds are likely greater. The installation of an active camera along a specific road segment slows average daily bus speeds by approximately one mile per hour. 4
  • 5. Figure 1: Automated Speed Enforcement Signage Figure 2: Automated Speed Enforcement Signage It is unlikely that automated speed enforcement cameras slow motorists already driving below the speed limit; therefore, it is likely that cameras slow speeding drivers to a much greater extent. This analysis does not portray a reduction in the number of speeders, but instead overall daily traffic. There are not significant differences between speed reductions in park and school zones. Overall, automated speed enforcement cameras slow traffic. 5
  • 6. Figure 3: Camera Process (1) (2) (3) (4) (5) (6) VARIABLES OLS FE & TE Park School High Low Camera Installation -0.761 -0.492*** -1.032*** -0.500*** -0.655 -0.647** (0.464) (0.152) (0.201) (0.166) (0.500) (0.273) Camera Activation -1.731*** -0.553*** -0.563*** -0.530 -0.388** (0.335) (0.131) (0.130) (0.473) (0.162) Constant 23.67*** 24.03*** 24.09*** 24.10*** 24.12*** 24.10*** (0.181) (0.0493) (0.0513) (0.0508) (0.0519) (0.0521) Number of Observations 78,277 78,277 74,318 74,831 72,722 72,563 R-squared 0.006 0.088 0.080 0.083 0.078 0.078 Road Segment FE No Yes Yes Yes Yes Yes Daily TE No Yes Yes Yes Yes Yes Number of Road Segments 1,027 978 984 958 956 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Figure 4: Results – Mean Daily Speed (1) (2) (3) (4) (5) (6) VARIABLES OLS FE & TE Park School High Low Camera Installation -0.825* -0.611*** -1.081*** -0.630*** -0.482 -0.776** (0.488) (0.177) (0.220) (0.195) (0.584) (0.354) Camera Activation -1.792*** -0.516*** -0.549*** -0.750 -0.456** (0.355) (0.151) (0.153) (0.545) (0.195) Constant 23.59*** 23.86*** 23.91*** 23.93*** 23.95*** 23.93*** (0.190) (0.0541) (0.0562) (0.0558) (0.0569) (0.0570) Observations 78,277 78,277 74,318 74,831 72,722 72,563 R-squared 0.006 0.084 0.076 0.079 0.074 0.074 Road Segment FE No Yes Yes Yes Yes Yes Daily TE No Yes Yes Yes Yes Yes Number of Road Segments 1,027 978 984 958 956 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Figure 5: Results – Median Daily Speed 6
  • 7. Works Cited Bordens, Alex, Abraham Epton, Geoff Hing, and David Kidwell. "Speed Cameras." Speeding Tickets Questioned. Chicago Tribune, n.d. Web. 12 Dec. 2016. <http://apps.chicagotribune.com/news/local/chicago-speed-camera-tickets/>. "Chicago Traffic Tracker - Congestion Estimates by Segments." Chicago Traffic Tracker - Congestion Estimates by Segments. City of Chicago, 08 Dec. 2016. Web. 12 Dec. 2016. <https://catalog.data.gov/dataset/chicago-traffic-tracker-congestion-estimates-by- segments-0b9d7>. Kidwell, David, and Abraham Epton. "Chicago Speed Cameras Catch School Bus Drivers, Police Cars, CTA Operators." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016. <http://www.chicagotribune.com/news/watchdog/redlight/ct-met-chicago-speed-cameras- government-met-20151228-story.html>. Kidwell, David, and Abraham Epton. "Emanuel's Speed Cameras Issue $2.4 Million in Bad Tickets." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016. <http://www.chicagotribune.com/news/watchdog/ct-speed-camera-bad-tickets-met- 20151117-story.html>. Kidwell, David, and Abraham Epton. "Many School-zone Speed Camera Tickets Issued without Danger to Kids, Tribune Finds." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016. <http://www.chicagotribune.com/news/watchdog/ct-speed-camera-school-tickets-met- 20151118-story.html>. Kidwell, David, and Abraham Epton. "Speed Cameras near Tiny Playlots, Closed Parks Lead to Big Payout." Chicago Tribune, 11 Jan. 2016. Web. 12 Dec. 2016.
  • 8. <http://www.chicagotribune.com/news/watchdog/ct-speed-camera-park-tickets-met- 20151118-story.html>. Kidwell, David, and Abraham Epton. "Top-ticketing Speed Cameras Not Where Most Kids Have Been Hit, Tribune Finds." Chicago Tribune, 05 Jan. 2016. Web. 12 Dec. 2016. <http://www.chicagotribune.com/news/watchdog/ct-speed-camera-tickets-safety-met- 20151120-story.html>. Pu, Wenjing, and Jie Lin. "Urban Travel Time Estimation Using Real Time Bus Tracking Data." Transport Chicago (n.d.): n. pag. Web. 12 Dec. 2016. <http://www.transportchicago.org/uploads/5/7/2/0/5720074/intelligentbus-pulin.pdf>. "Speed Camera Locations." City of Chicago, n.d. Web. 12 Dec. 2016. <https://data.cityofchicago.org/Transportation/Speed-Camera-Locations/4i42- qv3h/about>. "Speed Camera Violations." Speed Camera Violations. City of Chicago, 08 Dec. 2016. Web. 12 Dec. 2016. <https://catalog.data.gov/dataset/speed-camera-violations-997eb>. Wilson, C., C. Willis, JK Hendrikz, R. Le Brocque, and N. Bellamy. "Do Speed Cameras Reduce Road Traffic Crashes, Injuries and Deaths?" National Center for Biotechnology Information. U.S. National Library of Medicine, 01 Jan. 1970. Web. 12 Dec. 2016. <https://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0012902/>.