G Model
CSTP 83 No. of Pages 10
Case Studies on Transport Policy xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Case Studies on Transport Policy
journa l homepage: www.elsevier.com/locate/cstp
Costs and benefits of a road diet conversion
a, a b a
Robert B. Noland *, Dong Gao , Eric J. Gonzales , Charles Brown
a
Alan M. Voorhees Transportation Center, Edward J. Bloustein School of Planning and Policy, Rutgers University, 33 Livingston Ave, New Brunswick, NJ 08901,
United States
b
Department of Civil and Environmental Engineering, University of Massachusetts Amherst, Amherst, MA 01003, United States
A R T I C L E I N F O A B S T R A C T
Article history: The analysis presented here is of a cost benefit evaluation of a road diet conversion, reducing an urban
Received 30 July 2014
arterial street from two-lanes in each direction to one-lane in each direction with a middle turning lane.
Accepted 14 September 2015
Livingston Avenue is located in New Brunswick, New Jersey, and is an oversized arterial corridor
Available online xxx
accessing the center of the city. The costs include increases in travel time from the capacity reduction.
These are modeled with a VISSIM micro-simulation of the street. Safety benefits of road diets are one of
Keywords:
the main reasons that these are pursued. Evidence on the safety effect suggests that road diet conversions
Road diet
of arterial streets in urban areas will achieve about a 19% reduction in crashes. The benefits of crash
Cost-benefit analysis
reductions and the costs of increased delay are evaluated based on the value of statistical lives saved
Safety
versus the cost of travel time. This is done for various different scenarios and includes robustness checks.
Results overwhelmingly find benefits exceed costs over a 20 year period.
ã 2015 World Conference on Transport Research Society. Published by Elsevier Ltd. All rights reserved.
1. Introduction legislation and increased traffic enforcement in an attempt to
eradicate preventable vehicular-pedestrian crashes. (Bradshaw,
Complete Streets policies are aimed at balancing the needs of all 2013a,b) This study, thus, is an instrumental part of this plan.
roadway users, encouraging and allowing safe travel by bicyclists, However, at initial stages of this work, it was clear that the City was
pedestrians, transit users, and freight, in addition to existing car concerned about public opposition to any reductions in road
1
traffic. This is made possible through the transformation of the capacity that might slow traffic entering the City.
built environment and may include the installation of bicycle This analysis is intended to evaluate the feasibility of a road diet
lanes, crosswalks, sidewalks, pedestrian signals, and transit stops. on Livingston Avenue to understand how this may affect traffic
It may also include the addition of median islands, curb extensions, flow and also to evaluate the benefits and costs of the conversion.
or the occasional road diet. The latter is the focus of this analysis for There is evidence that road diets are effective at reducing crashes
Livingston Avenue in New Brunswick, New Jersey, a major arterial and reducing speeds (Gates et al., 2007; Pawlovich et al., 2006;
road that cuts through a residential area and feeds the center of the Thomas, 2013) and some evidence that there is no effect (Huang
city. There is substantial pedestrian traffic that crosses the street at et al., 2002). A synthesis of the literature concluded that road diets
both signalized and unsignalized crosswalks. This study was a are effective at reducing crashes (Thomas, 2013). Specifically, the
collaboration funded by a Rutgers University Community–Univer- costs are primarily associated with any traffic delay that might
sity partnership grant. The focus was on determining the costs and occur, while the benefits are based on potential reductions in traffic
benefits of implementing a road diet along Livingston Avenue. crashes, including those with pedestrians.
The City of New Brunswick adopted a Complete Streets policy in The approach taken here is to investigate the feasibility of a road
May 2009. The policy reinforced the city’s commitment to diet by analyzing the results of a micro-simulation of the traffic
“creating a comprehensive, integrated, connected street network network using VISSIM software. VISSIM is a microscopic, time step
that safely accommodates all road users of all abilities and for all and behavior-based simulation model developed to model urban
trips.” (City of New Brunswick, 2013a) The city has since made traffic and public transport operations (Planung Transport Verkehr
plans to increase investment in bicycle and pedestrian infrastruc-
ture(Barna, 2013; Bradshaw, 2013a,b) and has also passed
1
This was expressed in an early meeting with city staff. There was a desire to
* Corresponding author. complete the work prior to the start of a mayoral election campaign, since the plan
E-mail address: [email protected] (R.B. Noland). was seen as controversial and would likely be opposed by voters.
http://dx.doi.org/10.1016/j.cstp.2015.09.002
2213-624X/ã 2015 World Conference on Transport Research Society. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
2 R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx
(PTV), 2012). VISSIM is particularly useful for examining different downtown New Brunswick with US route 1 in the south. It
scenarios with altered lane configurations and signal timings. The stretches about 3.5 miles (5.6 km) and provides access to the many
key output provided by the model is the total travel time within the neighborhoods within the city. Livingston Avenue has two lanes in
network as well as the time each vehicle is delayed. This provides a each direction with a 25 mph speed limit for almost its entire
means of assessing how the relative level of service of Livingston length in New Brunswick. The roadway has an AM peak traffic
Avenue and the key intersections are affected by various road diet volume of about 14,000 vehicles and a PM peak volume of about
configurations. Estimates of crash and injury reduction are based 17,500–18,000 vehicles, over three hours for each peak. Most
on recent research suggesting that road diets can reduce crashes by vehicles exceed the speed limit of 25 mph (40 km/h). Land uses
19% (Thomas, 2013), and additional scenario analysis is conducted along the street are a mix of residential, retail, and small offices. A
to investigate the robustness of the cost/benefit analysis. number of uses fronting on Livingston Avenue attract vulnerable
road users, including three elementary schools, a rehabilitation
2. Background center for the blind, the public library, a 50-unit senior citizen
apartment building, multiple churches and a Rutgers University
New Brunswick is located in central New Jersey and sits along academic building. Total pedestrian activity tabulated during this
the Raritan River. It is the county seat of Middlesex County. The study amounted to over 9000 street crossings for both the morning
population is about 50% Hispanic, 16% African–American and totals and evening peak periods (over three hours in each peak), for
about 55,000 residents. Over one-third of the population is foreign counts at nine intersections along Livingston Avenue.
born. About 25% of the population is below the poverty line and The street enters North Brunswick to the south where it reduces
median household income is $40,280 based on the American to three lanes, one of which is a central turn lane and has an
Community Survey 5-year estimate, and average household increased speed limit of 35 mph (56 km/h). Along this final stretch
income is $53,854. New Brunswick is home to the main campus in North Brunswick the environment becomes more suburban in
of Rutgers University with an enrollment of over 40,000 students. character and the sidewalks disappear.
The city is relatively densely populated at about 10,500 residents Livingston Avenue has been identified as having on over-
per square mile (4075 per sq km) (City of New Brunswick, 2013b). representation of pedestrian crashes among county roads in
Employment in the city has grown to about 27,000 jobs from Middlesex county (Kaplan et al., 2012). Classified as an urban
20,000 in 1990. Rutgers University and Johnson & Johnson (which minor arterial, it is meant to interconnect with and augment the
is headquartered in New Brunswick) are the two major employers principal highway system. According to a road safety audit
in the city. conducted in 2011 there were a total of 113 vehicle crashes
The northeast corridor offers direct train service to New York between 2007 and 2009. Of these,17% involved a pedestrian and 6%
City, which has led to major redevelopment of the city as a transit- involved a bicyclist (Kaplan et al., 2012). Over two-thirds of the
oriented development. About 10% of the population commutes to pedestrian crashes occurred at night and most were at major
work by public transportation and slightly over 3% walk to work. intersections along Livingston Avenue. According to the safety
Much of the city is very walkable, with adequate sidewalks along audit, 49 of the 113 crashes (43%) resulted in an injury, with 54
almost all major and minor streets. people having injuries; there were no fatal crashes during this time
Livingston Avenue (also known as county route 691), the period. One issue noted in the audit is that the police department
subject of this study, is a major local street corridor that connects reported that many pedestrian crashes occur when vehicles that
Fig. 1. Road diet cross-section showing a typical configuration.
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx 3
are stopped for the pedestrian block the view of other vehicles that duration for both AM peak hour (7AM–10AM) and PM peak hour
are passing. The road safety audit recommended consideration of a (3PM–6PM). Two additional counts were performed during the AM
road diet as one way to mitigate these occurrences. and PM peak period to confirm the proportion of truck and bus
A road diet generally consists of converting a four lane road to a traffic on Livingston Avenue.
three lane road where the middle lane is a two-way turning lane; In setting up the traffic assignment in VISSIM, we used the
Fig. 1 displays a typical cross-section. Bicycle lanes (or wider average counts for each signalized intersection. The traffic flow at
shoulders) are also often part of the redesign of the road as there is each unsignalized intersection was scaled proportionally based on
now adequate space for them. Figs. 2 and 3 show before and after the counts from one unsignalized intersection. Thus, the turning
photos of a road diet on Livingston Avenue (the latter being a movements for each unsignalized intersection were proportionally
visualization of what it would look like). The study area for this equivalent, but scaled up or down based on the Livingston Avenue
project includes only the 1.5 mile segment of Livingston Avenue traffic flow passing each unsignalized intersection. Pedestrian
(shown in Fig. 4) that lies fully within the City of New Brunswick. counts were also obtained at these intersections, plus three
There are a number of key pedestrian attractors shown in Fig. 4, in additional locations with large pedestrian attractors. The total
addition to bus stops that line the street. number of pedestrians who crossed Livingston Avenue for the six
Recent research has demonstrated that road diets can hours in which counts were made is 3393 pedestrians (the balance
significantly reduce crashes (Thomas, 2013). A review commis- of almost 6000 crossings were along side streets), indicating a high
sioned by the FHWA synthesized the available research that has degree of exposure and potential risk. Our simulation did not
evaluated crash rates before and after implementation of a road include vehicle-pedestrian interactions.
diet. While not many projects have been evaluated, the results of
this synthesis suggest that in large urban areas crashes can be 3.2. Road geometry
reduced by about 19%. For rural highways that pass through
smaller urban areas this can be as much as a 47% reduction in The road geometry data are derived from the 2012 New Jersey
crashes. No studies have evaluated the impact on traffic fatalities High Resolution Orthophotography, downloaded from the New
and more severe injuries, but one would expect these to also be Jersey Geographic Information Network (New Jersey Geographic
reduced. Information Network, 2013). The images were combined together
in ArcGIS from which an image was exported to serve as the base
3. Data map for the simulation. The map provides basic road geometry
data, such as the edges of the roads, the centerline of the roads and
3.1. Traffic counts other geometry. The basic road network was then plotted in VISSIM
using this image as background.
To estimate travel times, a VISSIM model was calibrated for
Livingston Avenue. In order to calibrate the VISSIM model to actual 3.3. Signal timings
traffic conditions, traffic counts were collected at various sites
along the corridor. Physical turning movement counts at each of The existing traffic signals are timed with actuated signals at
the five major signalized intersections were also collected, and one the intersections of Livingston Avenue with New Street, Suydam
representative unsignalized intersection location. These data were Street, Handy Street, Sandford Street, and Nassau. There are
collected on weekdays in November and December, 2012 for 3 h separate timing plans for the morning and evening. The signals
Fig. 2. Livingston Avenue as it is currently configured.
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
4 R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx
Fig. 3. A visualization of Livingston Avenue with a road diet.
operate on simple two- or three-phase plans allowing for extended actuated for pedestrians or how often the vehicle extension was
green time when the pedestrian request is called in order to allow actuated. The analysis was conducted treating the signal timings as
sufficient time for pedestrians to safely cross Livingston Avenue. fixed as though there were high traffic and pedestrian demand in
The cycle length is fixed at 90 s for New Street and Nassau Street. every cycle. The fixed timing plans used as the existing case for the
Variable cycle lengths are allowed at Suydam Street (52–80 s), study are summarized in Table 1. This is consistent with operation
Handy Street (56–80 s), and Sandford Street (54–80 s). of the corridor at its capacity when all of the actuated phases
For the study of the Livingston Avenue corridor, we did not would be at their maximum length. In the interest of studying the
collect sufficiently detailed data to record how often the signal is performance of the corridor during the height of the AM and PM
Fig. 4. Livingston Avenue Study Area, with key pedestrian attractors.
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx 5
Table 2
peaks, we are able to provide an estimate of the worst-case level of
Re-optimized timing plan for PM peak (s).
service for car traffic during the busiest time of the day. Therefore
predictions of future performance are conservative and robust if Intersection Green Yellow All red Phase total
pedestrian traffic in the corridor were to increase substantially in
New St. (offset = 0)
the future. Livingston Ave all movements) 32 3 3 38
New St. (WB lead) 10 3 13
The simulation model was calibrated to recreate existing traffic
New St. (all movements) 42 3 4 49
conditions with the current signal timings and traffic flows. The
road diet on Livingston Avenue was then simulated with these
Suydam St. (offset = 61)
existing signal timings to estimate the effect on intersection level Livingston Ave (all movements) 48 4 2 54
of service from changing the street geometry without adjusting the Suydam St. (all movements) 40 4 2 46
traffic signals. The simulations resulted in excess congestion
Handy St. (offset = 52)
during the evening peak period. A set of re-optimized signal
Livingston Ave (NB lead) 6 3 9
timings were developed specifically for the evening peak period to
Livingston Ave (all movements) 49 3 2 54
fl fi
better re ect the changed traf c conditions. Handy St. (all movements) 31 4 2 37
Synchro 8 was used to re-optimize the cycle length, phase
Sandford St. (offset = 98)
lengths, and offsets in order to provide coordination for an efficient
Livingston Ave (all movements) 51 4 2 57
green wave for the peak travel direction (i.e., outbound from
Sandford St. (all movements) 37 4 2 43
downtown New Bunswick), and minimize the effects of queue
spillbacks blocking other intersections. Since the proposed road Nassau St. (offset = 0)
Livingston Ave (all movements) 51 4 2 62
diet reduces the capacity at intersections, a longer cycle length is
Nassau St. (all movements) 32 4 2 38
needed in order to reduce the effect of the lost time between
Note: All offsets are for the beginning of the first listed phase relative to the
phases on the intersection's capacity. The optimized signal timing
intersection at New St.
plan for the evening peak, accounting for the changed geometry of
the road diet, is summarized in Table 2.
from collecting actual speed data for Livingston Avenue. Average
The cycle length is set to 100 s at all intersections so that the
speed along Livingston Ave. within the simulation was 15.35 mph
signal phases are coordinated in every signal cycle. All phases are
(24.7 km/h) during the AM peak period and 14.71 mph (23.7 km/h)
greater in length than the minimum time required for pedestrians
during the PM peak period. This takes into account any vehicle
to cross the intersection as estimated from existing pedestrian
stops at signalized intersections and other vehicles in the roadway.
signal phases along Livingston Avenue.
Data on vehicle mix was collected. Buses and trucks are a very
small proportion of the traffic flow during peak periods. We set
4. Simulation assumptions
both as Heavy Goods Vehicles (HGV) in VISSIM. During the AM
peak the percent HGV was 5%, while during the PM peak the
Various assumptions and settings must be made in VISSIM prior
percent HGV was 3%.
to running a simulation. These include speed distribution
The settings for driver behavior were assumed to be the urban
assumptions, vehicle composition, and driver and pedestrian
fi
behavior. (motorized) environment default parameters as speci ed by
VISSIM. These are based on Wiedeman’s car following model
For the speed distribution we use an “S” curve based on a speed
(Planung Transport Verkehr (PTV), 2012; Wiedemann, 1974). The
limit of 30 mph, despite the posted limit being 25 mph. Observa-
driving behavior parameters as specified in VISSIM are as follows:
tion of the street suggested that only a small fraction of vehicles
look ahead distance, minimum: 0 ft, maximum: 820.21 ft (250 m);
observed the posted speed limit. Resource limitations prevented us
look back distance minimum: 0 ft, maximum: 492.13 ft. (150 m);
average standstill distance: 6.56 ft (2 m), additive part of safety
Table 1
distance: 2.00 ft (0.61 m), and multiplicative part of safety
Existing timing plan (s).
distance: 3.00 ft (0.91 m).
Intersection Green Yellow All red Phase total
The road diet was configured in VISSIM by realigning the lanes
–
New St. (6AM 9AM) for vehicular traffic to fit within a narrower right of way that would
Livingston Ave (all movements) 38 3 3 44
allow for installation of bicycle lanes on both sides of Livingston
New St. (WB lead) 10 3 13
Avenue. Since VISSIM does not have a straightforward way to
New St. (all movements) 26 3 4 33
model a shared left turn lane, the lane alignment at each
New St. (all other times) intersection has been carefully constructed so that left turning
Livingston Ave (all movements) 28 3 3 34 vehicles from Livingston Avenue and from side streets interact in a
New St. (WB lead) 10 3 13
realistic way. The simulated network with the road diet has been
New St. (all movements) 36 3 4 43
configured to recreate the vehicular interactions that would result
Suydam St. from a shared left turn lane and, most importantly, to exhibit the
Livingston Ave (all movements) 34 4 2 40 expected capacity for vehicular traffic in the corridor.
Suydam St. (all movements) 34 4 2 40
For turning behavior we assume that if a vehicle is turning right
from a side street to Livingston Ave, it will enter the right side lane
Handy St.
of Livingston Ave. Left turning vehicles will enter the left side lane
Livingston Ave (all movements) 39 3 2 44
Handy St. (all movements) 30 4 2 36 of Livingston Ave. Our road diet scenarios (discussed below)
include a center lane that left turning vehicles will turn into from
Sandford St.
side streets, prior to merging with the flow of traffic.
Livingston Ave (all movements) 36 4 2 42
Sandford St. (all movements) 32 4 2 38
5. Road diet scenarios
Nassau St.
Livingston Ave (all movements) 36 4 2 42
Our base case scenarios included both peak hours, AM and PM,
Nassau St. (all movements) 32 4 2 38
based on our traffic counts and the signal cycles discussed
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
6 R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx
previously. The model was calibrated and adjusted as needed to turning into Livingston Avenue from New Street. This scenario
match these counts and achieve a smooth flow without any major worked well with the re-optimized signal timing.
backups at intersections. Of particular note is the intersection with
New Street at the northern end of Livingston Avenue. This is just 6. Simulation results
prior to the core area of downtown New Brunswick. Most of the
inbound traffic turns right, flowing towards NJ state route 18, a In total seven simulations were run. These included our base
controlled access road that parallels the Raritan River, both case AM and PM calibrations that provide a baseline for existing
bypassing downtown New Brunswick to the north and headed traffic conditions. Road diet scenarios included one AM scenario
towards US route 1 and the New Jersey Turnpike towards the south. with the full road diet. For the PM road diet scenarios several
Our first road diet scenario included a center turning lane for simulations were run, given the spillback problems that occurred.
the entire length of Livingston Avenue, from New Street to Nassau PM road diet scenarios included 8% and 10% traffic reduction, the
Street. The traffic lanes were reduced to one lane in each direction former with the re-optimized signal plan, one with no road diet
from two lanes in each direction in the base case. Bicycle lanes from New St. to Handy St., and one with one northbound and two
were also included, although we did not simulate bicycle traffic. southbound lanes between New St. and Handy St. Each simulation
In our initial road diet simulations, the PM peak hour model was run for three hours of traffic, omitting the first and last five
resulted in significant congestion and spillback at the northern end minutes of the simulation in calculating results. Details of each
of Livingston Avenue. This resulted in blocking side street turns scenario are listed in Table 3.
into Livingston Avenue. We re-optimized the signal timing in an As VISSIM is a stochastic model, each simulation was run ten
attempt to fix this problem. Unfortunately, these changes were times and each measure of effectiveness was evaluated based on its
insufficient to alleviate the spillback problem. mean and standard deviation. The standard deviation in all our
Given the problem with this simulation we tested what would simulations is small, thus random variation in results is not an
happen by incrementally reducing the total traffic volume from 2% issue. The vehicles in the simulations were calibrated to actual
up to 10%, in steps of 2%. The simulation performed adequately traffic count data but vary slightly due to stochasticity. For the AM
with the traffic volume reduced by 10% with the original signal peak the simulation count is generally slightly less than 14,000
timings. With modified signal timings (discussed previously) the vehicles and for the PM peak about 18,500–19,000 vehicles (except
simulation was adequate with an 8% reduction in traffic. for those scenarios where the vehicles are intentionally reduced).
We feel that an 8–10% reduction in volume may be a realistic These results are shown in Table 4.
outcome when the road diet is implemented. First, if the road were We measured the total travel time per vehicle along the
to congest to the extent that our simulation suggested with a major corridor, the delay time per vehicle, and the delay time per vehicle
spillback, then some drivers would opt to find different routes. In for each of the signalized intersections, allowing the Level of
theory, the increase in congestion should lead to some suppression Service (LOS) to be evaluated. Results for total travel along the
of demand. The opposite effect, induced demand, is a documented corridor and corresponding delay times per vehicle are shown in
response to increased road capacity (Noland and Lem, 2002). As we Table 4. These are evaluated relative to the AM and PM base case
are reducing road capacity by up to 50% for the 1.5 mile stretch of scenarios.
Livingston Avenue, applying a lower bound lane-mile elasticity of The base case scenarios (1 and 2) both have an average delay of
0.2 would result in a 20% reduction in vehicle-miles of travel along slightly over 30 s. For our road diet simulations this increases to
the road (Noland and Lem, 2002). Thus, we feel an 8–10% reduction about 47 s for the AM period. Of the various PM road diet scenarios
is at the low range of probable responses. tested, the lowest delay occurs in scenario 6, when no road diet is
An alternative scenario was also tested that would not require implemented for the northern blocks of Livingston Avenue,
any assumptions on reducing total traffic. Since most of the avoiding the spillbacks in the simulation. If we assume a 10%
spillback problem is between New Street and Handy Street we reduction in traffic (scenario 5), the average delay is similar. When
tested alternative modifications to the road network between only two southbound lanes are maintained (scenario 7) the
these streets. increase in delay is larger than implementing the entire road diet
One scenario involved not implementing the road diet between (with a reduction in traffic). The last column shows the increase in
these two intersections. Thus, this part of Livingston Avenue would total delay relative to the base case (AM and PM, respectively).
continue to be a four-lane road. This is not an optimal design from a Using the road diet results with the lowest level of delay for both
walkability perspective as it is precisely this end of Livingston AM and PM periods we have a total increase calculated for all
Avenue that is closest to downtown New Brunswick and has the vehicles of 61.89 and 56.15 h of delay per workday, respectively, or
most pedestrian traffic. This scenario worked well with the original 118.04 h. This does not account for any alternative routes that the
signal timings. 10% reduction in PM traffic may have taken to avoid the more
Our other scenario to alleviate this problem was to keep two congested road.
lanes for the southbound travel direction from New Street to An alternative measure is to examine the change in travel times
Handy Street while the northbound flow would remain one lane. for those vehicles traversing the length of Livingston Avenue. These
There were no spillback issues with the northbound flow, so this would likely be vehicles making non-local trips and commute
solution was aimed at allowing better flow for those vehicles trips. Table 4 also displays the result for both northbound and
Table 3
Scenarios simulated with VISSIM.
Scenario number Scenario Scenario description
1 Base case AM
2 PM
3 Road diet AM
4 PM–8% traffic reduction, re-optimized signal timings
5 PM–10% traffic reduction, original signal timings
6 PM–4 lanes maintained from New St. to Handy St.
7 PM–2 southbound lanes maintained from New St. to Handy St., and 1 northbound lane
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx 7
Table 4
Performance measures for simulated results.
Scenario Scenario Average Average Average Total Average Average travel Average Average delay Increase Increase Total increase
Number delay speed Number increase travel time— time—Nassau delay time— time—Nassau in average in average in average
per (mph) of in delay New St. to St. to New St. New St. to St. to New St. delay delay delay, both
vehicle vehicles (hrs) Nassau (north- Nassau (north- (south- (north- directions (s)
(s) (south- bound) (s) (south- bound) (s) bound) (s) bound) (s)
bound) (s) bound) (s)
1 Base 31.10 15.35 13964 – 297.33 319.78 66.86 89.83 – – –
2 case 32.44 14.71 18963 – 316.98 304.56 85.21 74.48 – – –
3 Road 47.03 13.32 13984 61.89 333.29 406.58 101.94 176.03 35.08 86.20 121.27
4 diet 49.31 12.57 17798 83.36 385.64 368.59 153.75 138.72 68.53 64.24 132.77
5 44.00 13.11 17499 56.15 381.52 360.96 150.27 129.61 65.06 55.13 120.19
6 43.47 13.25 18975 58.14 375.45 348.08 143.41 119.30 58.20 44.82 103.02
7 52.66 12.30 18982 106.61 390.08 391.21 157.50 159.69 72.29 85.21 157.50
southbound average travel times and average delay as well as the value is $9,295,782 with the corresponding low and high values
increase in average delay time (relative to the base case AM and PM being $5,311,875 and $13,177,537, respectively.
scenarios). VISSIM does not provide data on the number of vehicles While these values apply to fatal crashes, most crashes do not
traversing the length of Livingston Avenue, thus we cannot result in a fatality. Thus, guidance is also provided on crashes that
calculate a total aggregate delay. result in a range of different levels of severity. These are valued as a
Base case average travel times are about 300 s, for both the AM fraction of the value of a statistical life as shown in Table 6.
and PM peaks in both directions. The road diet increases these As previously mentioned, the road audit of Livingston Avenue
average travel times to no more than about 400 s, with most reported 113 crashes over 3 years, so this is slightly less than 38
2
scenarios less than 400 s. crashes per year, on average. Some 17% involved pedestrians and
The final column summarizes the average increase in delay time 6% involved bicyclists, and 43% involved an injury, or 49 crashes
for each scenario. This is the sum of the northbound and with 54 injuries in total (Kaplan et al., 2012). Assuming an
southbound increase in average delay. The road diet scenarios estimated reduction in crashes of 19%, this would be 7.16 fewer
add about 120 s of additional delay; the lowest increase in average crashes per year, on average (Thomas, 2013) and assuming an
delay is when no road diet is implemented between Handy St. and equivalent reduction in injuries, this would be 3.42 fewer injuries
New St. (scenario 6). per year. Slightly more than half the crashes were property-
Delay at each signalized intersection is also analyzed. Results damage only crashes; these have additional costs that are not
are shown in Table 5 for the five signalized intersections along considered in this analysis.
Livingston Avenue that were simulated. Both average delay time The road safety audit of Livingston Avenue does not provide full
for all vehicles entering each intersection as well as the Level of information on the severity of crashes. About 6–7% of all pedestrian
Service (LOS) are reported. For New Street and Nassau Street the casualties are fatal in the state of New Jersey, but we do not expect
LOS does not change for any of the scenarios, remaining at C for this high a fraction on roads in urban areas, so we assume 1% is not
New Street and B for Nassau Street. The other three intersections survivable. For vehicle only crashes we assume 0.01% are not
are all at LOS B in the base case analysis; some deteriorate to LOS C, survivable. Based on this information we assume the distribution
but this is still considered an adequate LOS. Those intersections of severity as displayed in Table 7.
that deteriorate from LOS B to LOS C show minor changes in total The assumptions in Table 7 can be combined to develop a
average delay. weighted value of statistical life based on this distribution of
crashes and the split of vehicle versus non-motorized crashes on
7. Benefit-cost analysis of road diet implementation Livingston Ave. This is shown for low, mean, and high estimates of
the value of a statistical life in Table 8 for annual crashes on
Implementation of a road diet results in benefits to a Livingston Avenue.
community from the reduction of traffic crashes that would Assuming the road diet results in a 19% reduction in crashes,
otherwise occur. The trade-off is that this also involves an increase and that the distribution of severity levels stay the same, the cost
in travel time for those using the road, as one objective of a road savings from the road diet in 2014 would be as shown in Table 9.
diet is to slow traffic down. The benefits and costs of implementa- The cost of the increase in travel time associated with the road
tion can be compared by conducting a benefit-cost analysis of the diet can be estimated based on the value of travel time
impacts of the road diet. This is based on analyzing the estimated (Trottenberg and Belenky, 2011). Guidance from the US DOT
dollar value of crash reduction and the dollar value of increases in recommends that for personal travel this is 50% of the nationwide
travel time. The US Department of Transportation provides median household income, with plausible ranges varying from 35
guidance on how to conduct these analyses (Trottenberg and to 60%. For business travel, the value of travel time is 100% of the
Rivkin, 2013; Trottenberg and Belenky, 2011). nationwide median household income. Both are inflated by 1.6%
To analyze the benefits of crash reduction, US DOT guidance per year, reflecting estimates of increases in the valuation of time.
provides estimates of the value of a statistical life. This is defined as The nationwide median household income was $53,046 in 2012.
“the additional cost that individuals would be willing to bear for New Jersey has a higher median household income of $71,637,
improvements in safety . . . that, in the aggregate, reduce the while Middlesex County's median household income is even
expected number of fatalities by one.” (Trottenberg and Rivkin, higher at $79,442. New Brunswick however has a median
2013). The value estimated for the year 2011 is $9.1 million with the household income of only $40,280. This suggests a range of values
recommendation that low and high values of $5.2 million to $12.9
million be evaluated. This value is indexed by 1.07% per year, under
the assumption that increases in median real wages over time will 2
The road safety audit only measured crashes between Suydam and Nassau St.,
result in higher valuations of statistical life. Therefore, for 2014 the
about a one-mile stretch of the full 1.5 mile road. Thus, this is a low estimate.
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
8 R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx
Table 5
Average delay and level of service (LOS) at each signalized intersection.
Scenario Scenario New St. Suydam street Handy street Sanford street Nassau street
number
Average delay time LOS Average delay time LOS Average delay time LOS Average delay time LOS Average delay time LOS
(s) (s) (s) (s) (s)
1 Base 22.32 C 17.81 B 16.60 B 16.22 B 15.77 B
2 case 23.11 C 19.59 B 17.95 B 17.61 B 16.23 B
3 Road 21.84 C 18.68 B 19.58 B 18.15 B 16.41 B
4 diet 26.11 C 25.57 C 25.14 C 21.64 C 17.85 B
5 22.08 C 23.54 C 19.22 B 18.08 B 16.63 B
6 23.15 C 19.76 B 18.31 B 21.02 C 17.28 B
7 25.27 C 23.58 C 26.39 C 22.49 C 18.78 B
Table 6 in Table 10. These are applied for 260 work days per year based on
Relative disutility factors by injury severity level (AIS).
US DOT guidance (Trottenberg and Belenky, 2011).
AIS level Severity Fraction of VSL Travel for business must also be taken into account. While our
traffic counts could not discern all business travel, we did count
AIS 1 Minor 0.003
AIS 2 Moderate 0.047 heavy-goods vehicles and buses. These were 5% of the count during
AIS 3 Serious 0.105 the PM peak and 3% during the AM peak.
AIS 4 Severe 0.266
We also assume that average occupancy of each vehicle is 1.2
AIS 5 Critical 0.593
people. We did not count occupancy rates, but this is a reasonable
AIS 6 Unsurvivable 1.000
assumption as most vehicles are single-occupant.
Source: (Trottenberg and Rivkin, 2013).
Based on these assumptions, we can estimate how each road
Note: AIS is abbreviated injury scale.
diet scenario will increase the cost of travel time in 2014, including
both AM and PM peak periods when delay occurs. Results for each
Table 7 scenario are presented, summing both the AM and PM analysis of
Assumptions on severity of crashes on Livingston Avenue. delay. These are shown in Table 11.
The construction costs of converting the road must also be
AIS Severity Distribution of crashes Distribution of non-
taken into account. While we don't have any specific values, road
Level vehicle crashes motorized crashes
diet conversions can be relatively cheap and mainly involve
AIS 1 Minor 80.00% 20.00%
restriping of the road. Removal of existing stripes would also be
AIS 2 Moderate 10.00% 40.00%
AIS 3 Serious 6.99% 20.00% necessary. Based on contractor estimates from New Jersey bid
AIS 4 Severe 2.00% 12.00%
sheet data, we assume that the cost for both striping and restriping
AIS 5 Critical 1.00% 7.00%
is about $2000 per mile. This would be done for a 1.5 mile length;
AIS 6 Unsurvivable 0.01% 1.00%
four existing stripes would be removed while an additional eight
new stripes would be painted. A bicycle lane would also be
required for each side of the road; we assume the bicycle lanes cost
Table 8
as much as putting down eight stripes. This gives a cost estimate of
Cost of all annual crashes on livingston avenue based on statistical value of life and
$60,000.
weighted severity assumptions.
The total cost and benefits of the annual travel time costs and
AIS Level Severity Low valuation Mean valuation High valuation
safety benefits are evaluated over a 20 year project lifetime.
AIS 1 Minor $144,720 $253,259 $359,016
Construction costs occur in the first year and the stream of costs
AIS 2 Moderate $1,094,486 $1,915,351 $2,715,167
and benefits are discounted to the present. We assume a 4%
AIS 3 Serious $1,324,522 $2,317,913 $3,285,833
discount rate. This provides a net present value for the project.
AIS 4 Severe $1,728,502 $3,024,879 $4,288,015
AIS 5 Critical $2,199,585 $3,849,274 $5,456,663 Results are presented for all combinations of estimates for each
AIS 6 Unsurvivable $465,065 $813,864 $1,153,719 road diet scenario. In Table 12 we present the results using US DOT
Total $6,956,880 $12,174,540 $17,258,413
recommended valuations of travel time (based on median US
household income), for low, middle, and high valuations of
statistical lives saved. In Table 13 we include the highest value
Table 9
of time estimates, based on Middlesex County median incomes. All
Cost savings based on value of a statistical life and severity weighting assumptions,
scenarios show a large positive net present value over 20 years.
for year 2014.
These results assume a 19% reduction in crashes and that the
AIS Level Severity Low valuation Mean valuation High valuation
cost of restriping is $60,000. Both these assumptions could be off.
AIS 1 Minor $27,497 $48,119 $68,213
To evaluate this, the break-even crash reduction needed to have a
AIS 2 Moderate $207,952 $363,917 $515,882
benefit is calculated by varying these assumptions for the scenario
AIS 3 Serious $251,659 $440,403 $624,308
with the highest travel time cost and the lowest VSL. Assuming a
AIS 4 Severe $328,415 $574,727 $814,723
AIS 5 Critical $417,921 $731,362 $1,036,766 $60,000 construction cost, a crash reduction of only 16.46% is
AIS 6 Unsurvivable $88,362 $154,634 $219,207 needed to break even. If the cost were $6,000,000, then the crash
Total $1,321,807 $2,313,163 $3,279,099
reduction needed would be 22.19%. Both of these are within the
margin of error one would expect and given that this is the most
conservative scenario, it is clear that overall there would be net
to use for estimating travel time costs for this analysis; we analyze
benefits from a road diet conversion.
results assuming New Brunswick household income (low),
Our analysis did not account for growth in traffic over time. While
Middlesex County household income (high), as well as US
there is evidence of a saturation in growth of travel (otherwise
household income. Hourly value of time assumptions are shown
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx 9
of the benefits, we would not expect these to change the finding
Table 10
Value of travel time assumptions. that the road diet provides a positive net present value.
Value of time per hour
8. Conclusions
Personal travel (50%) Business travel (100%)
New Brunswick $9.68 $19.37 Our results have shown that implementation of a road diet,
Nationwide $12.75 $25.50
consistent with a complete streets policy along Livingston Avenue
New Jersey $17.22 $34.44
will result in some extra delay to traffic both along and within the
Middlesex County $19.10 $38.19
corridor. However, the costs of the delay to traffic are less than the
large benefits associated with the reduction in traffic crashes,
based on a cost/benefit analysis. Furthermore, the implementation
Table 11
of a road diet with modifications to the signal timing to improve
Estimated cost of delay for one year for different value of travel time assumptions
traffic flow will result in no or little change in the Level of Service
(2012 valuations).
(LOS) along Livingston Avenue. Where the LOS is degraded, the
Scenario, Total cost of delay Total cost of Total cost of delay
new LOS is still within acceptable service levels. Additionally, as a
combining AM (New Brunswick delay (US (Middlesex County
large majority of vehicles on the roadway have been identified as
and PM delay household income) household household income)
income) traveling in excess of the speed limit, the minor increases in travel
time on the road should result in lower overall speeds. Lower
4 $396,001 $593,849 $883,474
fi
5 $320,650 $481,987 $717,276 vehicle speeds correlate with reduced traf c crashes. Research
6 $326,161 $490,168 $729,431 suggests that road diets may decrease crashes by about 19% on
7 $460,386 $689,431 $1,025,485
urban arterials such as Livingston Avenue.
The AM peak period is not affected as much as the PM peak
period. Various adjustments to signal timings and reducing the
fl fi
Table 12 ow of traf c provide some mitigation to the increase in delay;
Net present value of road diet, assuming US median household income for estimates additional exploration of other mitigation options might be worth
of travel time costs.
exploring, however one of the objectives of a road diet is to reduce
Scenario, combining AM and PM Low VSL Middle VSL High VSL speed, thus one would expect to see delay times increase.
delay The use of micro-simulation software is one approach to
fi
4 $9,789,504 $21,315,067 $38,946,998 assessing changes in road con gurations. This analysis is limited by
5 $11,643,283 $23,168,846 $40,800,777 any inherent assumptions embedded within the model, although
6 $11,507,707 $23,033,270 $40,665,201
our analysis was robust to multiple simulations for each scenario.
7 $8,205,515 $19,731,077 $37,363,009
The standard deviations in travel time and delay were minor for
VSL: valuation of statistical life.
each scenario. Probably the main limitation is how travelers react
to a new configuration. Some may divert to alternative routes,
Table 13 some may opt to bicycle along the corridor if it is safer; while we
Net present value of road diet, assuming Middlesex County median household
estimated results with up to a 10% reduction in traffic during the
income for estimates of travel time costs.
PM period, shifts in traffic behavior may be greater.
Scenario, combining AM and PM Low VSL Middle VSL High VSL The decision to move forward with a road diet plan for
delay
Livingston Avenue depends on a balancing of the costs and benefits
4 $4,989,838 $16,515,401 $34,147,332 by the City of New Brunswick and Middlesex County. Our benefit/
5 $7,744,071 $19,269,634 $36,901,565
cost analysis balances the costs of travel time delay with the
6 $7,542,641 $19,068,204 $36,700,135
benefit of reducing crashes, and shows overwhelmingly positive
7 $2,636,441 $14,162,004 $31,793,935
benefits outweighing the travel time costs associated with the road
VSL: valuation of statistical life.
diet conversion. The net present value over 20 years is estimated to
range from $2.6 million to over $37 million for the road diet
known as “peak car”) (Goodwin and Van Dender, 2013), we might
conversion. This assumes a 19% crash reduction. Robustness checks
still expect some growth in traffic associated with further develop-
that vary the crash reduction rate and the construction costs also
ment in New Brunswick. While we did not run a simulation model
result in benefits for almost every scenario analyzed.
withadditionaltraffic,weadjustedthetravelcostforeachscenarioto
assume a 3% annual growth in traffic over 20 years. This is probably
9. Epilogue
unrealistic, but the bottom line is that it results in only two negative
net present value estimates over a 20 year lifetime (scenarios 4 and 7
This report was completed in February 2014. In March 2014, the
with a low valuation of statistical life). Thus, the results are very
City of New Brunswick announced that it intended to make
robust with respect to growth in traffic.
changes to Livingston Avenue to improve safety for pedestrians
Many other factors not accounted for in this analysis may affect
and bicyclists.
both the benefits and costs of the road diet conversion. On the
New Brunswick Mayor James Cahill stated that “Initiatives such
benefits side, we have not accounted for the benefits that
as a road diet for Livingston Avenue are part of the City’s effort to
pedestrians receive from less time crossing Livingston Avenue
make New Brunswick as ‘walkabl’ and healthy a city as possible”
and how their travel time may be reduced. We have also not
(Amaral, 2014). In May 2014, three children were injured while
accounted for the benefits that bicyclists receive, such as health
crossing Livingston Avenue as they were walking home from
benefits, or the savings associated with using a lower cost mode of
school (Attrino, 2014). One of the children sustained major injuries,
travel. Reducing property damage only crashes also is a benefit, and
while the other two had more minor injuries. In response, there
this has not been included in the valuations. On the cost side, we
was widespread demand in the community that something be
have not accounted for the costs associated with diverting some
done about Livingston Avenue (Cangiano et al., 2014).
traffic or suppressing demand, as well as how travel time and
Within days of the protests, the City worked with Middlesex
crashes may change on parallel routes. Given the large magnitude
County to have two sections of the street restriped. This can be
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002
G Model
CSTP 83 No. of Pages 10
10 R.B. Noland et al. / Case Studies on Transport Policy xxx (2015) xxx–xxx
Fig. 5. Restriping of Livingston Avenue in May 2014.
seen in Fig. 5. As of September 2015, no additional restriping has Barna, M., 2013. City Traffic Commission Approves Nine New Crosswalks on Georges
Road, (July 26th), http://newbrunswick.patch.com/groups/politics-and-
been done, however, the city has developed alternative designs to
elections/p/city-traffic-commission-approves-nine-new-crosswalks-
complete the work. Anecdotally, there does not seem to be any
902562401a (accessed Sept. 2015).
fi
disruption to traf c on the road from the two sections that have Bradshaw, J., 2013a. Bike Lane Project Moves Forward (July 26th), http:// 3 newbrunswick.patch.com/groups/politics-and-elections/p/bike-lane-project-
been changed.
moves-forward (accessed Sept. 2015).
From a policy perspective, one of the interesting conclusions is
Bradshaw, J., 2013b. City Council Votes in New Bicycle Regulations (July 26th),
the reaction of the public. The protests demanded that something http://newbrunswick.patch.com/groups/politics-and-elections/p/city-council-
votes-in-new-bicycle-regulations (accessed Sept. 2015).
be done about unsafe traffic on Livingston Avenue. At the early
Cangiano, A., Rabinowitz, R., Kratovil, C., 2014. Amit Protests Over Pedestrian Crash,
phases of this project, the City appeared concerned about the
County Fast-tracks Proposed Changes to Livingston Avenue, New Brunswick
political consequences of implementing a road diet. Yet the Today, http://newbrunswicktoday.com/article/amid-protests-over-pedestrian-
protests suggested the public was very much in favor of changes. crash-county-fast-tracks-proposed-changes-livingston-ave (accessed Sept.
2015).
Public officials should be less timid in their approach to
City of New Brunswick’, 2013a. Demographics, http://thecityofnewbrunswick.org/
implementing positive changes that improve safety and walk-
planninganddevelopment/demographics/ (accessed Sept. 2015).
ability, even at the expense of potential delays to traffic. City of New Brunswick, 2013b. Complete Streets Policy, http://
thecityofnewbrunswick.org/planninganddevelopment/complete-streets/
(accessed Sept. 2015).
Acknowledgements
Gates, T.J., Noyce, D.A., Talada, V. and Hill, L., 2007. The Safety and Operational Effects
of Road Diet Conversions in Minnesota (07-1918).
Goodwin, P., Van Dender, K., 2013. Peak car—themes and issues. Transp. Rev. 33 (3),
Funding for this project was provided by Rutgers University as a
243–254.
Community-University Research Partnership Grant for New
Huang, H.F., Stewart, J.R., Zegeer, C.V., 2002. Evaluation of lane reduction road diet
Brunswick. These are intended to recognize Rutgers University measures on crashes and injuries. Transport. Res. Record: J. Transport. Res.
Board 1, 80–90.
as a strong partner serving the City of New Brunswick. We also
Kaplan, A., Machado, A., Le, T., 2012. Livingston Avenue Road Safety Audit, http://
thank Glenn Patterson of the City of New Brunswick for comments thecityofnewbrunswick.org/planninganddevelopment/wp-content/uploads/
on an earlier draft and James Van Schoick, formerly of the Voorhees sites/8/2014/05/Livingston-Ave-RSA.pdf (accessed Sept. 2015).
New Jersey Geographic Information Network, 2013, https://njgin.state.nj.us/
Transportation Center who coordinated student researchers to
NJ_NJGINExplorer/index.jsp (accessed Sept. 2015).
conduct traffic counts; we also thank all the students who worked
Noland, R.B., Lem, L.L., 2002. A review of the evidence for induced travel and
on this project. changes in transportation and environmental policy in the US and the UK.
Transp. Res. Part D 7 (1), 1–26.
References Pawlovich, M.D., Li, W., Carriquiry, A., Welch, T., 2006. Iowa's experience with road
diet measures: use of Bayesian approach to assess impacts on crash frequencies
and crash rates. Transp. Res. Record: J. Transp. Res. Board 1953 (1), 163–171.
Amaral, B., 2014. New Brunswick pushing for Livingston Avenue bike lanes (July),
Planung Transport Verkehr (PTV), 2012. VISSIM 5.40 User Manual, PTV, Karlsruhe,
http://blog.nj.com/middlesex_impact/print.html?entry=/2014/03/ Germany.
livingston_avenue_new_brunswick_improvements.html (accessed Sept. 2015).
Thomas, L., 2013. Road Diet Conversions: A Synthesis of Safety Research, for Federal
Attrino, A.G., 2014. New Brunswick Study Suggested Traffic Changes where 3 Kids
Highway Administration, DTFH61-11-H-00024.
were Struck (July), http://www.nj.com/middlesex/index.ssf/2014/05/
Trottenberg, P., Belenky, P., 2011. Revised Departmental Guidance on Valuation of
new_brunswick_road_study_urged_traffic_changes_where_3_kids_were_
Travel Time in Economic Analysis.
struck.html (accessed Sept. 2015).
Trottenberg, P., Rivkin, R.S., 2013. Guidance on Treatment of the Economic Value of a
Statistical Life in U.S. Department of Transportation Analyses.
Wiedemann, R., 1974. Traffic Flow Simulation (in German).
3
The design plans are available at: http://thecityofnewbrunswick.org/blog/2015/
08/20/public-meeting-addresses-proposals-for-livingston-avenue-improvements/
Please cite this article in press as: R.B. Noland, et al., Costs and benefits of a road diet conversion, Case Stud. Transp. Policy (2015), http://dx.doi. org/10.1016/j.cstp.2015.09.002