Effects of Traffic Signal Retiming on Safety

Peter J. Yauch, P.E., PTOE

Program Manager, TSM&O – Albeck Gerken, Inc.

Introduction

It has long been recognized that traffic signal timing can have an impact on the safety of the
intersection. Some parameters are obvious – Yellow Change and Red Clearance interval timings are
critical to each phase, as are the Walk and Flashing Don’t Walk pedestrian intervals. Preemption times
are also crucial in some situations, as was demonstrated in the 1995 Fox River Grove, Illinois, crash
between a school bus and a commuter train. And, other controller related parameters can also have an
impact on safety, especially if the expectations of the drivers using the intersection are not met.

What about the coordination parameters for progression along a roadway; how could they impact the
crash history at an intersection? It is intuitive that the smoother the flow of traffic along an arterial, the
“safer” the arterial should be. NCHRP’s Guide for Reducing Collisions at Signalized Intersections 1 defines
one strategy as “Employ Signal Coordination”. Some of the key factors for that strategy in the report
included:

• Reducing the number and frequency of required stops and maintaining constant speeds for all
vehicles can help to reduce rear-end conflicts.
• Increased platooning can create more gaps of increased length for permitted vehicle
movements (such as permitted left turns and right turns on red) at intersections and result in
improved intersection operation.
• Platooning will contribute to consistent vehicle speeds along a corridor, which will help reduce
rear-end type crashes.

The Institute of Transportation Engineers’ Traffic Safety Toolbox 2 references two Atlanta area studies of
signal coordination implementation where intersection crash frequencies dropped by 25 and 38
percent. On one of the two studies, statistically significant reductions were noted for total collisions,
property damage only collisions, personal injury collisions, and right angle, rear end, left turn, and
sideswipe crashes. However, both of these studies were for new system implementations, i.e.,
comparing new system timings to isolated, non-interconnected operation.

How much of an impact on safety would a signal “retiming” project make? This performance measure is
not thoroughly addressed for most before-and-after analyses of system retiming projects. Travel time
runs can develop a good indication of project related Measures of Effectiveness, documenting
improvements such as fuel consumption, delay and travel time, and pollutant emissions, but adequate

1
NCHRP Report 500 – Guidance for Implementation of the AASHTO Strategic Highway Safety Plan, Volume 12: A
Guide for Reducing Collisions at Signalized Intersections, Transportation Research Board, 2004
2
Traffic Safety Toolbox, Institute of Transportation Engineers, 1999

1
amounts of crash data require significant “after” periods of evaluation. The original timing engineer or
consultant may no longer be available or under contract by the end of that time period, field conditions
may have changed along the corridor, or the agency may have moved its attention on to the latest
project.

Even the new Highway Safety Manual does not discuss the potential safety benefits of signal
coordination; there is no Crash Modification Factor (CMF) for signalized intersections or arterial
segments that would address the absence or presence of coordination and platooning on an arterial.
There are numerous CMFs related to change and clearance intervals, and there is one CMF included in
the Crash Modifications Factors Clearinghouse 3 that addresses a change in cycle length in a coordinated
system.

The informational gap between the intuitive reasoning that signal coordination will improve traffic
safety and the void of statistical data to support that reasoning is the basis for this paper. Traffic signal
retiming could be a significant safety improvement, but without the supporting cost-effectiveness
information used in accepting and prioritizing safety projects, may be overlooked as a low cost, easy to
implement option.

Methodology

Several suburban corridors in Hillsborough County (Tampa) Florida, that had been retimed by Albeck
Gerken, Inc., in the 2010 time frame, were selected for study. Crash data were obtained from the
County’s crash records system for a minimum of two years before timing implementation and as close
to two years as possible after timing implementation. These were evaluated through the Empirical Bayes
process described in the Highway Safety Manual to eliminate external factors in the before-and-after
analyses; the differences in crash trends before and after implementation of the timing patterns were
quantified, and estimates of societal cost savings made for use in a benefit/cost analysis.

Retiming Efforts

The scope of the signal retiming efforts for each corridor included the collection of seven day system
counts and 8-hour turning movement counts, updating of local timing parameters to meet current
standards (yellow change, red clearance, and pedestrian clearance), timing plan modeling and
development using Synchro for targeted times of the day / days of the week based on the seven day
counts, implementation of controller data, and fine tuning in the field using TruTraffic for conducting
floating car runs.

If appropriate, some phasing sequence changes may have been implemented on a per pattern basis,
switching the main street movements between the various combinations of leading and lagging left turn
phases. However, no display changes were made, and left turn phases were not changed from
protected/ permitted to protected only operation as part of these assignments.

3
www.cmfclearinghouse.org

2
Crash Data Analysis

Crash data were collected for the two year periods before and after the implementation of the new
timing patterns. As the better defined platooning from the signal retiming process might have had
effects on the safety and operations of the entire corridor, all reported crashes from the corridor were
included in the analyses – not just those at the signalized intersections.

In order to isolate the effects of the signal retiming on safety, by making it the lone variable in the
analysis process, each corridor was reviewed to ensure that there were no physical improvements
(widening, median modifications, lane reconfigurations, new signals, etc.) implemented during the
before or after data periods that could also have had an impact on the crash history. Had there been a
major change, the data from that section would not have been used.

A comparison of the crash history before and after timing implementation was made, and an annualized
change (improvement) was determined for each corridor. Most before and after crash analyses typically
stop at this point. However, safety analysts have become aware of some potential biases in using this
approach, and therefore recommend a more rigorous analysis, such as the Empirical Bayes approach as
defined in the new Highway Safety Manual.

For example, there is a tendency for timing engineers to implement new signal timings along rapidly
growing corridors that are congested and therefore probably have higher crash rates. As crashes do
have a degree of randomness, there is a strong statistical possibility that a higher crash rate will give way
to a lower rate without consideration to any roadway improvements. This is the regression-to-the-mean
concept, and it is often a concern when there is limited data availability.

The Empirical Bayes approach provides a means of eliminating any potential regression-to-the-mean
errors by using crash prediction models developed for a reference group (consisting of a much larger
data set), and then adjusting those models with modification factors (representing the local conditions
that are different from the assumptions in the models) to estimate the expected number of crashes for
the arterial being analyzed. The models are used to estimate the expected number of crashes with and
without a safety improvement, and can be used in the planning for roadway projects to determine the
most cost effective components of a project to be implemented.

However, as the safety effects of signal retiming have not been fully investigated in order to develop a
modification factor, the Empirical Bayes approach was used to establish the base conditions for the time
periods of the before and after conditions. This addresses any traffic volume changes along the
corridors. Then, the reported crashes, before and after the retiming implementation, were compared to
the expected number of crashes, before and after, from the base conditions.

Cost Benefit Analysis

Roadway safety projects generally need to be supported by an economic benefit analysis that shows
that the crash countermeasure to be implemented is cost effective in reducing the cost of crashes. In
the case of signal retiming, the cost of implementation is easily determined; its life span can be

3
estimated at three to five years before the timing patterns have degraded to a point where the benefits
may no longer exist.

One method of determining the cost of the crashes involved is the use of the Historical Crash Method,
described in the Florida Department of Transportation’s Plans Preparation Manual (2013 edition) which
allocates costs of average crashes by facility type based on the Highway Safety Improvement Program
Guideline. For the corridors used in these analyses, the following values from the manual are
appropriate for suburban locations:

4 – 5 Lane Divided: $183,372 per crash

6+ Lane Divided: $130,645 per crash

The manual advises to use half these values on non-state highway corridors.

These values can be used to determine the change in annual crash costs based upon the changes in
crash histories as determined previously. Then, the Present Value of the change in annual crash costs
can be estimated. Accepted economic practice for safety analyses in Florida use a discount rate of 4%
(i=0.04) and, for signal timing, the expected life of the improvement in relation to impacts on crashes
could be estimated at a conservative 1.5 years (y=1.5). Comparing the Present Value costs to the cost of
implementing the retiming projects leads to a Benefit / Cost ratio. Benefit / Cost ratios in excess of 1.0
generally are effective projects and may be acceptable for safety improvement funding.

Site Analyses

Six arterial corridors, with Average Annual Daily Traffic (AADT) volumes ranging from 33,000 to 64,000
and totaling approximately 45 miles in length, were selected as having been retimed by Albeck Gerken,
Inc., in the 2009 – 2010 time frame. These corridors are shown in Figure 1 and are tabulated with
various characteristic data in Table 1. Of the six corridors, only two had a significant change (more than
10% increase) in traffic volumes between the before and after periods. All had arterial speed limits
ranging from 40 to 50 miles per hour.

Crash records for 24-month periods before and after the implementation of the fine-tuned timing plans
were obtained and reviewed, and are included in Table 2. Note that all six corridors showed a reduction
in the number of reported crashes in the two years after implementation; summed over the six
corridors, total crashes were reduced 17%.

4
 Figure 1 – Arterials Analyzed in Hillsborough County, Florida

Figure 2 shows the total crashes by corridor, broken down into major types of crashes. The most
significant change in the type of crash was the rear-end crash; there were reductions of rear end crashes
in five of the six corridors. For the most part, there were minimal differences in angle, sideswipe, and
head-on crashes, and generally some reductions in the “other” type of crashes. Angle crashes address
both failure to yield or red light violations at signalized intersections, or right of way violations at non-
signalized locations. Other crashes include pedestrian and bicycle crashes, single vehicle and roadway
departure crashes, and a variety of other types of incidents.

5
 Table 1 – Arterials Analyzed in Hillsborough County, Florida

Speed
Cross Length Timing
ID Corridor AADTbefore AADTafter Signals Limit
Section (miles) Installation
(mph)
Dale Mabry Highway –
February
HC-1 Lambright to Lutz Lake 6LD 52,875 60,000 10.8 20 45
2011
Fern (State Road)
Hillsborough Avenue –
th 6LD – 5.9 mi October
HC-2 Anderson to 40 59,000 59,625 6.4 19 40 & 45
4LD – 0.5 mi 2010
(State Road)
Hillsborough Avenue –
HC-3 Countryway to Hoover 6LD 56,500 57,250 6.0 15 45 & 50 July 2010
(State Road)
Waters Avenue –
6LD – 4.8 mi February
HC-4 Montague to Habana 33,025 32,725 6.7 18 45
4LD – 1.9 mi 2011
(County Road)
Brandon Blvd (SR 60) – 8LD – 4.3 mi
HC-5 Orient to Dover 6LD – 1.0 mi 62,950 64,425 9.7 24 45 June 2009
(State Road) 4LD – 4.4 mi
Bloomingdale Avenue
6LD – 0.5 mi
– US 301 to Lithia September
HC-6 5L – 4.1 mi 41,250 46,000 5.8 14 45
Pinecrest 2010
4LD – 1.2 mi
(County Road)

Table 2 – Before and After Crash Records

Crashes Before Retiming Crashes After Retiming Difference in Crashes

(24 mos.) (24 mos.) (24 mos.)

HC-1 1812 1532 -280

HC-2 1053 915 -138

HC-3 564 509 -55

HC-4 1180 850 -330

HC-5 2623 2151 -472

HC-6 802 723 -79

Total 8034 6680 -1354

6
 3000

2500
Crashes (24 months)

2000
Other
1500 Head On
Angle
1000 Sideswipe
Rear End
500

Figure 2 – Crash Types, Before and After Retiming Implementation

Predictive Crash Analyses

The site analyses thus far looked only at raw numbers of observed crashes; the effects of any regression-
to-the-mean concerns or of any volume changes have not been addressed. To investigate those issues,
an Empirical Bayes analysis, using the predictive models and processes from the Highway Safety Manual,
were conducted. The total reported crashes, shown in Table 2, became the Observed values in the
Empirical Bayes analysis. The N expected values for the corridors were calculated as the summations of the
predictive models for suburban arterials, non-signalized intersections, and signalized intersections, with
adjustment and modification factors to address the specific site characteristics. These were then
compared with the Observed values to determine the estimated change in the number of crashes.

All six of the corridors included six-lane divided sections, and the suburban arterial predictive models
were developed for four-lane divided. On five of the six corridors, the models appeared to predict the
observed crashes closely. However, the model did not accurately track one corridor (HC-5 – Brandon
Boulevard (State Road 60)), significantly underestimating the number of predicted crashes in
comparison to the observed crashes. This is the only corridor that included an eight-lane section. For
that reason, corridor HC-5 was not included in the summary of the Predictive Analyses.

Table 3 summarizes the findings. For four of the five evaluated corridors, the predictive process
indicated that an increase in the overall number of crashes would have been expected, tracking the

7
volume changes in the before and after periods and reflecting the sensitivity of the models to even
minor changes in volumes.

Along all five evaluated corridors, the overall number of crashes was significantly less than what would
have been expected without the retiming efforts. These reductions ranged between 5% and 13% over
the total crashes; values that are consistent in magnitude with published results for other studies
related to signal timing improvements or signal system implementations.

Table 3 – Crash Analysis using Predictive Algorithms

Total Crashes Along Corridor (24-month periods Before and After)

ID Change
Observed before Observed after N expected-before N expected-after (N expected-after) ̶
(Observed after)
-138
HC-1 1812 1532 1437 1670
-8.2%
-128
HC-2 1053 915 1029 1043
-12.2%
-76
HC-3 564 509 575 585
-13.0%
-46
HC-4 1180 850 906 896
-5.1%
HC-5 2623 2151 NA NA NA
-111
HC-6 802 723 726 834
-13.3%

Cost Benefit Analysis

A Cost Benefit Analysis of the retiming process, looking solely at crashes, was performed using the
Historical Crash Method, which allocates costs of average crashes by facility type based on the Highway
Safety Improvement Program Guideline. For the corridors used in these analyses, the following values
from the manual are appropriate for suburban locations:

4 – 5 Lane Divided: $183,372 per crash

6+ Lane Divided: $130,645 per crash

The manual advised to use half these values on non-state highway corridors.

These values were used to determine the change in annual crash costs based upon the changes in crash
histories as determined previously. The Present Value of the change in annual crash costs was

8
estimated using a discount rate of 4% (i=0.04) and an expected life of the improvement of 1.5 years
(y=1.5). This information is displayed in Table 4.

Table 4 – Benefit Cost Analysis

Annual Crash Cost - Present Value of Cost of Benefit /

Annual Crash
ID Crash Weighted by Crash Reduction Retiming Cost
Cost Reduction
Reduction Facility Type at i=4%, y=1.5 Effort Ratio

HC-1 69 $ 130,645 $ 9,014,505 $ 12,875,868 $ 129,956 99

HC-2 64 $ 130,645 $ 8,361,280 $ 11,942,834 $ 133,696 89

HC-3 38 $ 130,645 $ 4,964,510 $ 7,091,058 $ 103,628 68

HC-4 23 $ 72,470 $ 1,666,810 $ 2,380,788 $ 143,774 17

HC-5 NA NA NA NA NA NA

HC-6 55 $ 89,065 $ 4,898,575 $ 6,996,880 $ 96,612 72

Total 249 $ 28,905,680 $ 41,287,428 $ 748,443 55

This analysis shows that a high benefit to cost ratio was realized for all five of the evaluated corridors
and for the sum of the projects as a whole. These five corridors would be eligible for safety funding for
the retiming efforts.

Conclusions

The predictive approach established in the Highway Safety Manual is an effective tool in isolating and
identifying the effects of traffic operational improvements such as signal retiming. The process
addresses the issues of regression-to-the-mean, common in traffic safety project analyses, and also can
counter the effects of changes of traffic volumes to the different periods being analyzed.

The development and implementation of new timing plans can have a very positive effect on the
reduction of overall crashes; looking solely at an average cost per crash system wide, a benefit/cost ratio
of 55:1 was established for the five evaluated corridors. Add to that the benefits of reduced travel time,
fuel consumption, and pollutant emissions, and signal timing becomes one of the most cost effective
traffic operational improvements available.

Acknowledgment

Jacob Mirabella, a graduate student at the University of South Florida and Intern with Albeck Gerken,
Inc., assisted with the analyses conducted for this study.