Cost Estimating Model for Mode Choice between

Light Rail and Bus Rapid Transit Systems

Lo Rosa Hsu1
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

Abstract: A cost estimating model is developed for light rail transit (LRT) and bus rapid transit (BRT) systems to better aid transportation
planners and decision makers in the selection process. In evaluating transit systems, cost estimating has always been a major consideration.
Without an applicable cost estimating model or methodology, the decision on the choice between LRT and BRT would cause significant
controversy and difficulty in the early corridor planning stage. The model developed in this paper can be applied to estimate costs and
compare the LRT and BRT systems that operate on various right-of-way categories, alignment configurations, and different given transit
demand volumes. An example and comparative analyses illustrate how the model can be applied in real-world situations. DOI: 10.1061/
(ASCE)TE.1943-5436.0000466. © 2013 American Society of Civil Engineers.
CE Database subject headings: Light rail transit; Buses; Costs.
Author keywords: Light rail transit; Bus Rapid Transit; Cost estimating model.

Introduction allocation technique and standard cost components have been

used to construct the cost estimating model to avoid unnecessary
Over the preceding years, transit agencies in the United States complications. Once the cost estimating model is developed, it will
have been building light rail transit (LRT) or bus rapid transit proceed through a validation process by applying it in exactly the
(BRT) systems to provide an effective and reliable high-speed same manner it would be used in the future. This determines
transit service. However, the selection between LRT and BRT if the procedures produce reasonable comparisons between the
has been controversial for many years, with no suitable closure. existing data and the estimating data, with the purpose of illustrat-
Hartford, Connecticut has selected the BRT system to upgrade their ing how the cost estimating model can be applied to real-world
mass transit systems, but people still argue that LRT would be a situations.
better choice. Many still criticize that the BRT system is probably
more valid for the Austin area because of the lower capital cost
relative to LRT. Review
LRT is comprised of electrically propelled vehicles that operate
individually or in trains. BRT uses conventional bus vehicles, Both capital and operating costs, which are expressed as unit costs
restricted lanes and signal prioritization to provide a rail-like ser- ($/sp-km or $/prs-km), are usually considered (Vuchic 2005)
vice. Although these two systems provide comparable services, when evaluating different transit modes. Li and Wachs (2000)
they use different technology and exhibit dissimilar operating char- presented a widely used cost estimation method in the transit indus-
acteristics. The capital cost of the LRT system is, on average, higher try by featuring the average unit costs in the allocation method. The
than that of the BRT system on a per unit distance basis, because of capital costs are typically one-time costs, including infrastructure
the rails, electrification, signaling, and vehicles needed for con- cost with the construction of right-of-way (ROW) and facilities,
struction. However, LRT systems are less expensive to operate be- in addition to the acquisition of equipment and vehicles (Bichler
cause of the higher passenger capacity of light rail vehicles and the 1989). The Federal Transit Administration (FTA 1992) provided
ratio of passengers to the driver. The decision between LRT and capital cost information for five selected LRT systems, and recom-
BRT focuses primarily on the project cost with no proper costing mended the capital cost components of LRT systems be classified
method integrated to estimate the overall system cost. Hence, the into eight cost components. Diaz et al. (2004) classified the capital
choice has caused significant controversy and difficulty in the early cost components of BRT systems into five cost components. In 2005
corridor planning stage. the FTA issued an implementation guideline workbook, the Standard
In this paper, an integrated cost estimating model is developed Cost Categories (SCC), with a consistent format for reporting,
to provide a comprehensive cost estimating procedure for LRT estimating, and managing capital costs for New Starts projects
and BRT systems to improve the selection process. The cost (FTA 2007). The transit agencies follow the criteria of Reporting
Instructions for the Section 5309 New Starts Criteria, May 2007
1
Dept. of Environmental Technology and Management, Taoyuan to report their project costs for federal funding eligibility (FTA
Innovation Institute of Technology, 414 Section 3, Jhongshan East Rd., 2008). When the project estimation, project schedule and the imple-
Jhong-Li 32091, Taiwan. E-mail: [email protected] mentation schedule are determined, the escalation rates are input
Note. This manuscript was submitted on April 26, 2010; approved on
to the FTA workbook. Thus, the expenditure costs estimated
June 4, 2012; published online on August 20, 2012. Discussion period open
until June 1, 2013; separate discussions must be submitted for individual for the year will then be automatically calculated in the FTA work-
papers. This paper is part of the Journal of Transportation Engineering, book. However, the differences of the predicted capital costs or
Vol. 139, No. 1, January 1, 2013. © ASCE, ISSN 0733-947X/2013/1-20- operating expenses between the actual capital costs or operating
29/$25.00. expenses are unclear.

20 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013

J. Transp. Eng. 2013.139:20-29.

 The capital costs vary considerably with construction complex- estimating model to properly estimate LRT and BRT systems’ costs
ity and various ROW types. The average capital cost of LRT for evaluation and comparison. Therefore, the problems addressed in
systems is generally higher than that of BRT systems, on a per unit this paper are to (1) obtain the cost components of capital and OM
distance basis, because of the need for rails, electrification and costs for both LRT and BRT systems, (2) examine the unit costs of
signaling, and higher vehicle costs. The capital costs for LRT the cost components of capital and OM costs from existing LRT
systems feature a wide range of variability, ranging from a low and BRT systems, (3) develop a cost estimating model to estimate
of $16.5 million per mile ($10.3 million per km) for the the capital cost, annualized capital cost, OM cost, and cost per pas-
Orange Line in San Diego to a high of $106.0 million per mile senger mile for both LRTand BRT systems operating on various ROW
($66.0 million per km) for the 6.4 mile (10.3 km) LRT system categories and alignment configurations, and (4) find a preferable sys-
in Buffalo (77% tunneling). However, the BRT tunnels in Seattle tem between LRT and BRT systems by comparing both capital and
and Boston (a section of the Silver Line) exhibit a cost as high as OM costs in terms of cost per passenger mile. In other words, the sys-
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

the most expensive rail tunnels, such as the Silver Line tunnel sec- tem with the lower cost per passenger mile is the preferable and more
tion, with a cost estimated at approximately $500 million per mile cost-effective system.
($310.8 million per km) (Vuchic 2007).
Operating and maintenance (OM) cost is incurred by regular
Development of Cost Estimating Model
operation of the system. Typically, regular costs involved in daily
operations include operating and maintenance costs. The National In this paper, a cost estimating model is developed to estimate
Transit Database (NTD) provides historical data of the OM cost for capital cost, annualized capital cost (depreciation cost), OM cost
public transportation systems. However, most transit agencies nei- and cost per passenger mile to help transportation planners and de-
ther collect nor maintain OM cost for BRT systems or their indi- cision makers select the better mode between LRT and BRT sys-
vidual lines. OM costs of the BRT systems were calculated or tems, in the early corridor planning stage or system analysis. An
estimated by the transit agencies according to the average OM cost accurate cost estimation is difficult to obtain; therefore, the cost
of regular bus service. Transit OM cost is primarily influenced by estimating model is established on a unit cost basis and the standard
fleet size, frequency, passenger capacity and route structure. cost components of both systems will be used to avoid unnecessary
Khistry and Lall (2003) found a faster method to determine transit complications. The unit costs allocated to the cost components of
OM cost through application of unit costs to determine an approxi- capital and OM costs in the cost estimating model were derived
mate cost calculation. It was based on (1) distance-related costs, from high-low unit costs of the historical data of existing systems.
i.e., those related to energy, maintenance, and servicing of Such historical data were obtained from multiple resources, includ-
vehicles, (2) time-related costs, i.e., those related to operating staff ing the FTA, Urban Mass Transportation Administration (UMTA),
wages, and (3) route-related costs, i.e., those related to maintenance Transit Cooperative Research Program (TCRP), General Account-
of roadway, track, signals and stations. ing Office (GAO), and the National Transit Database (NTD), in
The fleet size of the system exerts a significant impact on capital addition to local transit agencies and published papers.
and OM costs. Cost estimations can only be conducted after the
fleet sizes of both LRT and BRT have been calculated. Hsu and
Wu (2008) developed a fleet size model, as shown in Eq. (1), Capital Cost
for determining the number of vehicles required in maximum ser-
vice for both LRT and BRT systems operating under various ROW In accordance with the estimating methodology of the FTA, capital
categories. Factors taken into consideration in the fleet size model cost components of the LRT system are classified into eight cost
include vehicle characteristics, passenger demand volume, ROW categories: guideway, yards and shops, systems, stations, ROW, spe-
configurations, numbers of stations, headway, size of vehicle, cial conditions, vehicles, and soft cost (Table 1). The capital cost
and traffic signal timing at the intersections. structure of the BRT system is dissimilar to that of the LRT because
they utilize different technology. Capital cost components of
tt the BRT system are classified into five cost categories: running
pf½T A þ T B þ T c
þ ½2ð2ðmp Þhtp þ ð60 ÞÞ
g
N¼ ð1Þ way, stops and stations, traffic signal priority, vehicles, and soft cost
Cv α (Table 2). The running way of the BRT system has three types: bus-
way, bus-high occupancy vehicle (HOV) lanes and BRT on arterial
where N = number of vehicles needed in maximum service; T A T B , streets. The BRT system’s construction is somewhat similar in many
and T C = round trip travel time, excluding the dwell time at the stop aspects to freeway or highway construction. The route (guideway for
or station of the vehicle running on the exclusive, reserved, and LRT and running way for BRT) of both systems might operate on
mixed traffic ROW (ROW A, B, and C), respectively, in units of various alignments, which are at-grade, elevated or tunnel structure.
hours; p = passengers per hour per direction (pphpd); tt = layover Tables 1 and 2 contain a list of various cost components, high-
and recovery time (min); tp = passenger boarding/alighting time low ranges of the unit costs for each cost component, and their re-
(h/passenger); h = headway (in units of hours); m = number of spective economic lifetime. By using the cost allocation technique,
loading areas per stop for BRT (m ¼ 1 for an LRT system); the LRT and BRT capital cost can be calculated through Eqs. (2)
Cv = seats per vehicle; and α = loading factor. and (3), as follows:
1
CLRT ¼ ½ðG1 U g1 þG2 U g2 þG3 U g3 þU s þU r þU sp ÞL
Problem 1−U so
þðU y þU v ÞN L þðG1 U stL1 þG2 U stL2 þG3 U stL3 ÞS
ð2Þ
Most previous cost models and system comparisons do not take both
capital and OM costs into consideration. However, cost comparisons where CLRT = LRT capital cost in millions of dollars; Gi = structure
among different modes should be examined on the basis of total costs, percentage (grade, elevated, and tunnel at i ¼ 1, 2, and 3,
including capital and OM costs, because the provision of transit respectively); L = route length (in units of miles); S = number
service incurs both types of costs. Because of different planning of stops and stations); N L = number of LRT vehicles; U so = soft
and complicated operation scenarios, it is necessary to develop a cost cost (assuming 10% of the total project cost); U gi = LRT guideway

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013 / 21

J. Transp. Eng. 2013.139:20-29.

 cost (in millions of dollars per mile, with previously noted i where CBRT = BRT capital cost (in millions of dollars);
descriptors); U s = LRT system cost (in millions of dollars N B = number of BRT vehicles; Gij = structure percentage
per mile); U r = LRT right-of-way cost (in millions of dollars (grade, elevated, and tunnel at i ¼ 1, 2, and 3, respectively, and
per mile); U y = LRT yards and shops cost (in millions of arterial, busway, and HOV lane at j ¼ 1, 2, and 3, respectively);
dollars per vehicle); U sp = LRT special condition cost (in millions U Rij = BRT running way cost (in millions of dollars per lane mile,
of dollars per mile); U vL = LRT vehicle cost (in millions of dollars with previously noted i and j descriptors); U stBi = BRT stop or
per vehicle), and U stLi = LRT stop or station cost (in millions of station cost (in millions of dollars per stop or station); U t =
dollars per stop or station, with previously noted i descriptors). BRT traffic signal priority cost (in millions of dollars per intersec-
tion); and U vB = BRT vehicle cost (in millions of dollars per
1 vehicle).
CBRT ¼ ½2LðG11 U R11 þ G12 U R12 þ G13 U R13 þ G2 U R2
1 − U so The author input the high-low unit costs in Tables 1 and 2
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

þ G3 U R3 Þ þ ðU t S þ U vB N B Þ þ SðG1 U stB1 into Eqs. (2) and (3), respectively. Therefore, the capital cost of
LRT and BRT can be estimated through Eqs. (4) and (5), as
þ G2 U stB2 þ G3 U stB3 Þ
ð3Þ follows:

CLRT=high ¼ 1.1½ð8.6G1 þ 21G2 þ 60G3 þ 17.8ÞL þ ð3.3ÞN L þ ð3.6G1 þ 3.6G2 þ 33G3 ÞS

ð4Þ
CLRT=low ¼ 1.1½ð2.9G1 þ 2.9G2 þ 45G3 þ 3.7ÞL þ ð1.1ÞN L þ ð0.2G1 þ 3.6G2 þ 8G3 ÞS



CBRT=high ¼ 1.1½2Lð2.7G11 þ 2.7G12 þ 9.4G13 þ 27.6G2 þ 96.6G3 Þ þ ð0.04S þ 1.5N B Þ þ Sð3.6G1 þ 3.6G2 þ 33G3 Þ

ð5Þ
CBRT=low ¼ 1.1½2Lð0.01G11 þ 2.3G12 þ 6G13 þ 11G2 þ 55.2G3 Þ þ ð0.01S þ 0.3N B Þ þ Sð0.2G1 þ 3.6G2 þ 8G3 Þ

For the purpose of making a cost comparison between LRT and recovery factor is applied for each cost component in the cost esti-
BRT systems, the capital costs should be annualized to calculate mating model. The annual capital recovery factors have been deter-
the total annual costs, which are the sum of annualized capital cost mined based on an FTA-prescribed 7% interest rate (PBQD 2001) to
(depreciation cost) and annual OM cost. The capital costs are annual- construct the annualized capital cost. The LRT and BRT annualized
ized to represent depreciation and interest changes. An annual capital capital costs can be calculated through Eqs. (6) and (7), respectively,
as follows:
Table 1. Cost Components of LRT Capital Cost
Table 2. Cost Components of BRT Capital Cost
Unit cost
(millions Lifetime Unit cost
Components Unit of dollars) (years) (millions Lifetime
Components Unit of dollars) (years)
Guideway
At-grade $/Mile High 8.6 30 Running way
Low 2.9 At-grade
Elevated $/Mile High 21.0 30 Arterial $/Lane Mile High 0.27
Low 2.9 Low 0.009
Subway $/Mile High 60.0 100 Busway $/Lane Mile High 2.7 30
Low 45.0 Low 2.3
Yards and shops $Shop/Vehicle High 0.9 30 HOV $/Lane Mile High 9.4
Low 0.2 Low 6.0
System $/Mile High 6.2 30 Elevateda $/Lane Mile High 27.6 30
Low 1.3 (Exclusive busway) Low 11.0
Stationa $/Station High 3.6 30 Subway $/Lane Mile High 96.6 100
Low 0.2 (Exclusive busway) Low 55.2
Vehicle $/Vehicle High 2.4 30 Stop, stationb $/Stop High 3.6 30
Low 0.9 Low 0.2
Right-of-way $/Mile High 2.7 100 Traffic signal priority $/Intersection High 0.04 10
Low 1.1 Low 0.008
Special condition $/Mile High 8.9 0 Vehicle $/Vehicle High 1.5 12
Low 1.3 Low 0.3
Project soft cost 10–40% of total Project soft cost 15% of total
project cost (miscellaneous) project cost
Note: Data sources are the following: (1) cost of guideway, system, special Note: Data sources are the following: running way cost (Diaz et al. 2004;
conditions (Schneck 1991). Year 1990 costs will be escalated to year 2000 Levinson et al. 2003); traffic signal priority cost (Diaz et al. 2004; de Garmo
using 1.32 escalation rate. 1 mile = 1.609 kms. (2) cost of yards and shops, 2000), vehicle cost (Levinson et al. 2003). 1 mile = 1.609 km. Costs are in
station, systems, vehicle (FTA 1992). Year 1994 costs will be escalated to year 2000 U.S. dollars.
a
year 2000 using 1.16 escalation rate. (3) project soft cost (PFQD 2001; These facilities can be a major highway (Diaz et al. 2004).
b
Schneck 1991). The stop and station cost of BRT assumes the same as LRT system. Each
a
Each elevated station costs $3.6 million; tunnel station costs between elevated station costs $3.6 million; tunnel station costs between $8 million
$8 million to $33 million, average $21 million (Schneck 1991). Year 1990 and $33 million, average $21 million (Schneck 1991). Year 1990 costs are
costs are escalated to year 2000 using 1.32 escalation rate. escalated to year 2000 using 1.45 escalation rate.

22 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013

J. Transp. Eng. 2013.139:20-29.

 1 operation, general administrative, vehicle maintenance, and non-
CPLRT ¼ fA ½LðG1 U g1 þ G2 U g2 þ U s Þ
1 − U SO 30 vehicle maintenance costs.
The major component of OM cost is labor expenses, which are
þ N L ðU y þ U vL Þ þ SðG1 U stL1 þ G2 U stL2 Þ

vehicle operation and administrative costs. Vehicle maintenance
þ A100 ðU r LÞ þ A100 ðG3 ÞðU g3 L þ U stL3 SÞg ð6Þ costs (fuel, tires, vehicle servicing) and vehicle operation costs (op-
erating staff wages, administration, ticketing and fare collection,
1 system security) are variable. Non-vehicle maintenance costs (road-
CPBRT ¼ fA ½2LðG11 U R11 þ G12 U R12 þ G13 U R13 way, track, signals, and stations) are fixed costs. Vehicle mainte-
1 − U SO 30
nance and administration, vehicle operation, and non-vehicle
þ G2 U R2 Þ þ SðG1 U stB1 þ G2 U stB2 Þ
þ A10 ðSU t Þ maintenance costs are distance-, time-, and route-distance related
costs, respectively. Therefore, three units used to express OM
þ A12 ðN B U vB Þ þ A100 G3 ð2LU R3 þ SU stB3 Þg ð7Þ
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

cost are (1) distance-related costs (dollars per vehicle-mile),

where CPLRT ¼ LRT annualized capital cost (in millions of (2) time-related costs (dollars per vehicle-hour), and (3) route-
dollars); CPBRT ¼ BRT annualized capital cost (in millions of related costs (dollars per route unit distance of LRT; dollars per
dollars); and Ai = annual capital recovery factor (A10 , A20 , A30 , vehicle of BRT). The OM cost can be expressed as shown in
and A100 ¼ 0.142, 0.126, 0.081, and 0.07, respectively, and i is Eq. (8), as follows:
in years).    
2 × x1 × x6 × x7 × x8 x2 × x4 × x6 × x7
OM ¼ þ
x5 x5
OM Cost þ ð2 × x3 × x8 Þ ð8Þ

The NTD provides historical data of the OM cost of public trans- where OM = operating and maintenance cost (in millions of dol-
portation systems. Fourteen LRT systems were examined and the lars); x1 = dollars per vehicle-mile; x2 = dollars per vehicle-hour;
data were collected. The existing data of BRT systems are limited. x3 = dollars per route-mile for LRT or dollars per vehicle for
For this reason, BRT systems examined and selected in this re- BRT; x4 = roundtrip travel time (in units of hours); x5 = peak head-
search were based on available data. Data from most existing sys- way (in units of hours); x6 = operating time (in units of hours);
tems were applied to develop the cost estimating model; the data of x7 = cars per train for LRT or buses per stop or station for
the remaining systems proceeded through a validation process to BRT; and x8 = route length (in units of miles) for LRT or vehicles
illustrate how the developed cost estimating model could be applied required in maximum service (in units of number of vehicles)
in the future. for BRT.
Table 3 lists the breakdown of the OM costs of nine existing The roundtrip travel time, x4 , in Eq. (8) can be replaced by the
LRT systems and seven existing bus systems. Most of the data re- product of the number of vehicles required in maximum service (N)
ported by transit agencies to the NTD do not separate BRT buses and peak headway (h) divided by cars per train of LRT or buses per
from regular buses in the system. Consequently, the BRT OM costs stop of BRT (n). Three hundred equivalent working days in a year
are derived from the operating expenses of regular bus systems. The and 20 operating hours per day of transit systems are used as a
cities selected for the seven existing bus systems in Table 3 feature default value. Three peak hours each in the day and evening are
both regular bus and BRT services, because accounting of the BRT applied. Off-peak headway is twice the time of peak headway.
OM cost is more consistent. The components of LRT and BRT/bus Therefore, the annual OM cost for LRT and BRT can be expressed
OM costs are similar, which can be broken down into vehicle as in Eq. (9), as follows:

Table 3. Breakdown of the OM Costs of the Selected LRT and Bus Systems
System Vehicle General Vehicle Non-vehicle
System City operationa operation administration maintenance maintenance Total
LRT Sacramento 40.7 8,091,248 4,306,463 4,819,471 2,111,784 19,328,966
Salt Lake City 29.6 3,834,812 458,701 1,589,763 1,476,417 7,359,693
San Diegob 96.6 14,220,561 7,454,215 6,098,982 4,380,931 32,154,689
Dallas 40.8 12,314,721 7,945,759 5,482,530 7,112,018 32,855,028
Baltimore 57.6 13,750,555 2,706,672 5,315,633 6,962,551 28,735,411
Portland 64.9 11,949,492 9,495,758 9,836,077 10,059,334 41,340,661
St. Louis 34.0 8,701,097 2,479,112 3,762,139 4,648,018 19,590,366
Denver 28.0 4,369,476 2,217,512 2,536,271 2,056,867 11,180,126
Buffalo 12.4 5,963,005 1,756,469 3,247,109 3,549,432 14,516,015
Average percentage 40 21 19 20 100
BRT Seattle 892 151,902,827 26,470,045 46,341,860 19,268,839 243,983,571
San Jose 427 85,661,821 53,304,424 36,029,440 7,300,665 182,296,350
Los Angeles 1,888 346,252,299 112,586,578 157,160,824 21,680,487 637,680,188
Pittsburgh 848 102,988,809 23,749,931 43,436,807 11,219,354 181,394,901
Dallas 441 64,701,387 37,390,858 26,760,162 7,713,013 136,565,420
Denver 639 83,047,498 33,865,201 40,694,718 12,751,070 170,358,487
Miami 530 92,197,192 18,479,944 32,122,633 8,055,418 150,855,187
Average percentage 53 22 19 6 100
Note: Compiled year 2000 data from NTD. 1 mile = 1.609 km. Costs are in year 2000 U.S. dollars.
a
System operation for LRT = rail-total directional route miles, system operation for BRT = vehicles operated in maximum service.
b
Many of the administration costs are not included in the San Diego LRT system (Watkins 2003).

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013 / 23

J. Transp. Eng. 2013.139:20-29.

  
26nL can then be conducted. The comparison of LRT and BRT
OM L=B ¼ 300 × × x1 þ ð13N × x2 Þ
h systems is evaluated by expressing total costs, which include
 capital and OM costs in terms of cost per passenger mile. The
þ CL ð2L × x3 Þ þ CB ðN × x3 Þ ð9Þ cost per passenger mile can be estimated using Eq. (12), as
follows:
where OM L=B ¼ LRT or BRT operating and maintenance cost CP þ ðOMÞ
(in millions of dollars); n = cars per train for LRT or buses per Cost per passenger mile ¼ ð12Þ
p × T adj × 300
stop or station for BRT; h = peak headway (in units of hours);
L = route length (in units of miles); N = number of vehicles;
CL ¼ 1 for LRT and 0 for BRT; CB ¼ 1 for BRT and 0 for ðT þ 6Þ
T adj ¼
LRT; x1 = dollars per vehicle-mile; x2 = dollars per vehicle-hour; 2
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

and x3 = dollars per route-mile of LRT or dollars per vehicle where CP = annualized capital cost (in millions of dollars); OM =
of BRT. operating and maintenance cost (in millions of dollars); P =
x1 , x2 , and x3 can be calculated by applying the historical data in passengers per peak hour per direction (pphpd); T adj = adjusted
Table 3. The calculation results of these three unit costs of the OM operating time (in units of hours); and T = operating time (in units
cost are presented in Table 4. The high and low values of x1 , x2 , and of hours).
x3 in Table 4 are input to Eq. (9). Therefore, the high-low annual
OM costs of LRT and BRT systems can be estimated as follows in
Eqs. (10) and (11): Model Verification

) Finally, the cost estimating model will proceed through a validation

OM LRT=high ¼ 0.043nL
h þ 0.47L þ 0.31N process by applying it exactly the same manner it would be used in
ð10Þ
OM LRT=low ¼ 0.019nL
h þ 0.075L þ 0.138N the future. The cost estimating model will be validated by compar-
ing the cost estimates derived from the model with existing costs of
) the selected systems.
OM BRT=high ¼ 0.032nL
h þ 0.25N
ð11Þ
OM BRT=low ¼ 0.015nL
h þ 0.15N Capital Cost

It is necessary to investigate the ROW categories, alignment con-

Cost Per Passenger Unit Distance figurations, number of vehicles and number of stops and stations of
the selected LRT and BRT systems to estimate the capital cost. Data
After the capital cost, annualized capital cost and OM cost collected from the selected systems were input into the model.
are calculated; the estimation of the cost per passenger mile Comparisons of the capital costs estimated from the cost estimating

Table 4. Three Unit Costs of OM Cost

Dollars per route-mile
Dollars per Dollars per for LRT, dollars per
System City Vehicle-mile Vehicle-hour vehicle-mile (x1 )a vehicle-hour (x2 )b vehicle for BRT (x3 )c
LRT Sacramento 2,267,721 111,752 4.02 72.40 142.16
Salt Lake City 1,508,956 75,464 1.36 50.82 136.65
San Diego 7,166,547 338,801 1.89 41.97 124.25
Dallas 2,451,300 155,624 5.48 79.13 477.57
Baltimore 2,770,769 174,408 2.90 78.84 331.17
Portland 5,079,456 293,828 3.81 40.67 424.65
St. Louis 2,550,783 126,749 2.45 68.65 374.54
Denver 1,565,100 123,367 3.04 35.42 201.26
Buffalo 907,063 76,849 5.52 77.59 784.23
High 5.52 79.13 784.23
Low 1.36 35.42 124.25
BRT Seattle 40,040,176 2,701,471 1.82 56.23 72
San Jose 22,649,071 1,604,158 3.94 53.40 57
Los Angeles 92,451,378 6,978,567 2.92 49.62 38
Pittsburgh 36,422,988 2,578,768 1.84 39.93 44
Dallas 22,291,782 1,599,967 2.88 40.44 58
Denver 33,875,388 2,110,567 2.20 39.35 67
Miami 27,871,134 2,070,989 1.82 44.52 51
High 3.94 56.23 72
Low 1.82 33.18 38
Note: Compiled year 2000 data of vehicle hour and vehicle mile from NTD. 1 mile = 1.609 km. Costs are in year 2000 U.S. dollars.
a
x1 = the sum of vehicle maintenance and general administrative costs (data in Table 3) divided by the vehicle miles (data in Table 4).
b
x2 = vehicle operations costs (data in Table 3) divided by the vehicle hour (data in Table 4).
c
x3 = non-vehicle maintenance costs divided by route mile ($/Route-Mile) for LRT system or divided by the number of vehicles operated in maximum service
(dollars per vehicle) for BRT system (data in Table 3).

24 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013

J. Transp. Eng. 2013.139:20-29.

 Table 5. Existing and Estimating Capital Costs of Selected LRT and BRT Systems
Existing data Estimating data
Capital cost Millions of Capital cost Millions of
(millions of dollars (millions dollars
System City dollars) per mile of dollars) per mile Comments Difference (%)
LRT Buffalo 718 112 High 976 152.6 ROW A: 77% −3
Low 407 63.6 ROW B: 23%
Average 692 108.1 77% tunnel; 14 stations;
27 vehicles;
route mile = 6.4
21.3
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

Salt Lake City 315 High 628 40.5 ROW B: 86% 15

Low 148 9.4 ROW C: 14%
Average 388 24.6 100% At-grade; 20 stations;
23 vehicles
route mile = 14.8
Denver 320 23.0 High 653 46.0 ROW A: 83% 25
Low 164 11.5 ROW B: 17%
Average 408 28.8 25% elevated; 20 stations;
31 vehicles
route mile = 14.0
Portland 1,366 41.0 High 1,844 55.9 ROW A: 36%; −11
Low 560 17.0 ROW B: 63%
Average 1,202 36.3 ROW C: 1%
9.4% tunnel; 17% elevated;
50 stations; 72 vehicles
route mile = 33.0
Dallas 1,770 38.0 High 2,630 56.3 ROW A: 35% −0.5
Low 899 19.3 ROW B: 59%
Average 1,765 37.8 ROW C: 6%
16% tunnel; 12.3% elevated;
44 stations; 95 vehicles
route mile = 46.7
BRT Seattle 593 282 High 671 319.7 ROW A: 100% −17
(tunnel busway) Low 308 146.7 99% tunnel; 5 stations/stops
Average 490 233.1 route mile = 2.1
Miamia 63.1 7.4 High 121 15 ROW B:100% 3
(busway) Low 4 0.5 100% at-grade;
Average 62 7.6 16 stations/stops
route mile = 8.2
Los Angelesb 127.3 11.6 High 184 17.7 ROW A:100% 3
Low 78 7.1 100% at-grade;
Average 131 11.9 3 stations/stops
route mile = 11.0
Pittsburgh 175.5 25.7 High 310 45.7 ROW A: 100% at-grade; 7
(east busway) Low 65 9.5 6 stations/stops
Average 188 27.6 route mile = 6.8
Pittsburgh 63.34 14.7 High 119 27.7 ROW A: 88% 21
(south busway) Low 33 7.8 ROW C: 12%
Average 76 17.8 8 stations/stops
route mile = 4.3
Note: Data sources are the following: capital cost data of the LRT systems (Watkins 2003; Hsu 2003), NTD 2000, capital cost of the BRT systems (Levinson
et al. 2003; de Garmo 2000; USGAO 2001). The soft cost in the developed cost model is 10% of the total projected cost intended to estimate capital cost of the
LRT or BRT system. 1 mile = 1.609 km. Costs are in year 2000 U.S. dollars.
a
Existing capital cost data in Miami do not include the cost of additional buses (de Garmo 2000).
b
The BRT system in Los Angeles is the San Bernardino HOV lane (El Monte busway).

model with the existing capital costs of the selected LRT and BRT The difference between existing and estimated capital costs is
systems are listed in Table 5. In Table 5, the existing capital cost of relatively minor.
each selected system falls between the high and low cost ranges
of the estimated costs. The average capital cost estimated from
the cost estimating model is close to the existing capital cost of OM Cost
each selected LRT or BRT system. The differences between the
existing capital cost and the average estimated capital cost of each Route mile (L), number of vehicles (N), peak headway (h), cars per
system are from 0.5–25%. The average difference is 15.5%. train (n) of the LRT system, and buses per stop or per station (n)

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013 / 25

J. Transp. Eng. 2013.139:20-29.

 Table 6. Existing and Estimating OM Costs of Selected LRT and BRT Systems
Existing data Estimating data
System City OM cost (millions of dollars) OM cost (millions of dollars) Comments Difference (%)
LRT Pittsburgh 27.8 High 37.6 n ¼ 1.38
Low 14.4 h ¼ 4 min −7
Average 26.0 N ¼ 47
route mile ¼ 35
vehicle-mile ¼ 1; 893; 842
vehicle-hour ¼ 130; 778
Newark 9.1 High 11.7 n ¼ 1.0
h ¼ 2.6 min −10
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

Low 4.6
Average 8.2 N ¼ 16
route mile ¼ 8.3
vehicle-mile ¼ 540; 518
vehicle-hour ¼ 45; 312
Boston 89.1 High 111.3 n ¼ 1.9
Low 46.0 h ¼ 2.6 min −12
Average 78.6 N ¼ 154
route mile ¼ 51
vehicle-mile ¼ 6; 334; 450
vehicle-hour ¼ 422; 297
Baltimore 28.7 High 40.0 n ¼ 2.45
Low 13.9 h ¼ 13 min −6
Average 27.0 N ¼ 40
route mile ¼ 57.6
vehicle-mile ¼ 2; 770; 769
vehicle-hour ¼ 174; 408
Cleveland 15.9 High 23.6 n ¼ 1.9 1
Low 8.4 h ¼ 9.2 min
Average 16.0 N ¼ 25
route mile ¼ 30.8
vehicle-mile ¼ 1; 158; 015
vehicle-hour ¼ 76; 173
Los Angeles 61.4 High 77.9 n ¼ 1.65
Low 29.0 h ¼ 5.0 min −13
Average 53.4 N ¼ 51
route mile ¼ 82.4
vehicle-mile ¼ 4; 709; 915
vehicle-hour ¼ 205; 207
BRT Seattle 6.6 High 8.6 n ¼ 2.5 2
Low 4.8 h ¼ 3 min
Average 6.7 N ¼ 25b
route mile ¼ 2.1
Los Angeles 32.7 High 47.4 n ¼ 1.0 11
(Wilshire-Whittier) Low 25.1 h ¼ 2 min
Average 36.3 N ¼ 90b
route mile ¼ 26
Los Angeles 12.5 High 15.0 n ¼ 1.0
(Ventura metro bus) Low 8.0 h ¼ 4 min −8
Average 11.5 N ¼ 29b
route mile ¼ 16
Pittsburgh (east) High 44.7 n ¼ 3.0
Low 23.8 h ¼ 1.7 min
Average 34.2 N ¼ 86b
route mile ¼ 6.8
43.0a 8c
Pittsburgh (south) High 16.1 n ¼ 2.0
Low 8.8 h ¼ 4 min
Average 12.4 N ¼ 37b
route mile ¼ 4.3

Note: Data of the selected BRT systems is from Levinson et al. (2003), n = cars per train of LRT or buses per stop or station of BRT, h = headway, and
N = number of vehicles required. 1 mile = 1.609 km. Costs are in year 2000 U.S. dollars.
a
OM cost in Pittsburgh is estimated as 43,000 average weekday trips × $0.15 per passenger mile × 11.1 route length × 2 × 300 day per year = $43 million
(de Garmo 2000).
b
Applying equation (1) to estimate N, number of vehicles required.
c
8% ¼ ðð34.2 þ 12.4Þ − 43Þ=43.

26 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013

J. Transp. Eng. 2013.139:20-29.

 of the BRT system are collected from the selected LRT and above-grade construction. The passenger demand is 3,400 passen-
BRT systems. Most local transit agencies neither collect nor main- gers per peak hour per direction. Many still criticize that the BRT
tain OM costs for BRT systems or their individual lines. Therefore, system is probably valid for the Austin area with a view to the lower
in Table 6, the BRT systems selected for validating the cost esti- capital cost relative to the LRT system. (de Garmo 2000). The
mating model are based on the available data. The data was input assumption is that the service frequency of LRT and BRT is 10
into the model to estimate the OM costs. Comparisons of the results and 3 min, respectively, during peak hours. The comparison between
between OM costs estimated from the developed model and LRT and BRT is at the same corridor and features the same environ-
existing OM costs obtained from the selected LRT and BRT sys- ment to yield a reasonable result.
tems are listed in Table 6. In Table 6, the existing OM cost of each
selected system falls between the high and low cost ranges of the
estimated OM cost obtained from applying the cost estimating Comparative Analyses
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

model. For each LRT or BRT system, the existing OM cost of

the selected system is close to the estimated OM cost obtained The systems used to conduct comparisons and analyses are LRT-1,
from the cost estimating model. The differences between the LRT-2, BRT-1, and BRT-2 (190, 250, 45, and 67 passengers per
existing O&M cost and the average estimated OM cost of each vehicle, respectively). The model results in Table 7 show that LRT
system are from 1–13%. The average difference is 7.8%. The dif- systems feature higher capital costs than BRT systems; however,
ference between the existing and estimated OM costs is rela- LRT systems exhibit much lower OM costs than BRT systems.
tively minor. The comparisons of feasibility and cost effectiveness between
With the verification results, the reliability of the cost estimating LRT and BRT systems are conducted by expressing total annual
model is well-illustrated by the fact that the differences between the costs in terms of cost per passenger mile. Total annual costs include
existing costs of the selected systems and the costs estimated from capital and OM costs. The higher passenger loads per vehicle, the
the developed model are relatively minor. They indicate that the lower cost per passenger mile. The system with the lower cost
cost estimating model can solve the problems in estimating capital per passenger mile would be the preferable and more cost-effective
and OM costs in real-world situations. system in the respective areas.
As the comparative analyses in Fig. 1, when the passenger de-
mand volume exceeds 2,000 passengers per hour per direction
Application (pphpd), the LRT systems feature less cost per passenger mile than
the BRT systems. In other words, the LRT systems are more cost-
Transportation planners had considered LRT systems to be one of the effective than the BRT systems when the transit volume exceeds
alternatives to transit service in the Austin area. The LRT proposal in 2,000 pphpd. With a view to its higher passenger loads per vehicle,
Austin could be applied as an example to illustrate the applicability of LRT-2 is more cost-effective than LRT-1. In this example, the
the methodology presented in this paper. A demonstration of apply- prediction of passenger demand is approximately 3,400 pphpd.
ing the cost estimating model for systems’ comparison and data Hence, the LRT (LRT-2) system is a suitable transit system for
analysis is presented as follows. Capital Metro was proposing a the 14.6-mile (23.5-km) transit plan in Austin, Texas.
phased implementation of the Austin, Texas area LRT system with
the development of a 14.6 mile (23.5 km), 16 station minimum op-
erable segment (MOS) of the LRT system from McNeil Road in Sensitivity Analyses
north Austin to downtown. The line includes approximately 9 miles
(14.5 km) and 5.6 miles (9.0 km) of exclusive ROW and mixed Given that LRT and BRT operations have many possible situations,
traffic, respectively, and involves approximately one-half mile of sensitivity analyses of the capital and OM costs with respect to

Table 7. Cost Estimating Model Results of the LRT and BRT Systems
LRT-1 LRT-2 BRT-1 BRT-2 (articulated bus)
Input Route length (mi) 14.6
Passenger demand (pphpd) 3,400
Number of stops and stations 16
Operating hour (h) 20
ROW (%) ROW A: 68; ROW B: 0; ROW C: 32
Alignment (%) at grade: 97; elevated:3
Vehicle passenger capacity
Passengers per vehicle 190 250 45 67
Seats per vehicle 68 190 50 74
Peak headway (min) 10 10 3 3
Output Number of vehicles requireda 42 31 118 74
Capital cost (millions of dollars) 509.1 477.0 311.0 257.0
Millions of dollars per mile 34.8 32.7 21.3 17.6
OM cost (millions of dollars) 27.0 21.6 62.4 38.4
Depreciation cost (millions of dollars)b 29.6 27.1 31.2 24.1
Total annual costs (millions of dollars)c 56.5 48.6 93.6 62.5
Dollars per passenger mile 0.146 0.125 0.242 0.162
Note: Year 2000 costs are escalated to year 2010 using 1.26 escalation rate. 1 mile = 1.609 km. Costs are in year 2010 U.S. dollars.
a
Number of vehicle required is estimated by applying Eq. (1).
b
Depreciation cost = annualized capital cost.
c
Total annual costs = depreciation cost + OM cost.

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013 / 27

J. Transp. Eng. 2013.139:20-29.

 Break-even point
0.35 Table 9. Capital and OM Costs with Respect to Variables in the Cost
Estimating Model: Alignment Configuration
0.30
Alignment (%) Capital cost OM cost
$ /Passenger-mile

0.25 (millions of (millions

BRT-1
0.20 System At grade Elevated Tunnel dollars per mile) of dollars)
BRT-2 LRT 100 0 0 32.0 24.9
0.15
LRT-1 50 50 0 37.2 24.9
0.10 LRT-2
40 50 10 45.9 24.9
0.05 40 40 20 53.5 24.9
40 30 30 61.2 24.9
0.00
1000 2000 3000 3400 4000 5000
BRT 100 0 0 17.4 57.6
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

Passenger demand volumes (pphpd) 50 50 0 40.3 57.6

40 50 10 62.2 57.6
LRT-1 LRT-2 BRT-1 BRT-2 40 40 20 78.9 57.6
40 30 30 95.9 57.6
Fig. 1. Comparative analyses
Note: Passenger demand volume = 3,400 pphpd; route length = 14.6 mi;
number of stations and stops = 16; LRT ¼ 68 seats/vehicle, 2.8 passengers/
seat; BRT ¼ 50 seats/vehicle, 0.9 passengers/seat; 1 mile = 1.609 km; costs
certain key variables in the model are conducted to obtain a further are in year 2009 U.S. dollars.
understanding of how the cost estimating model works.

Table 10. Capital and OM Costs with Respect to Variables in the Cost
Capital and OM costs with Respect to Route Length Estimating Model: Row Configuration
The capital and OM costs with respect to route length in Table 8 ROW (%) Capital cost OM cost
show that the capital cost of BRT is less expensive than that of LRT Fleet size (millions of (millions of
at the same route length. In other words, the cost estimated from the System A B C (vehicles) dollars per mile) dollars)
developed model indicates that the construction of BRT is less ex- LRT 100 0 0 37 31.5 23.8
pensive than that of LRT. On a per unit distance basis, the longer the 40 50 10 40 32.0 24.6
route length, the lower the cost for both LRT and BRT systems; 20 60 20 41 32.4 25.0
however, as the route length increases, the OM cost increases. With 5 65 30 43 32.7 25.4
longer route length, the O&M cost of the BRT is much higher than 5 95 0 40 32.1 24.6
that of the LRT. BRT 100 0 0 94 18.0 52.6
40 50 10 108 19.1 55.8
20 60 20 116 19.8 57.7
Capital and OM costs with Respect to Alignment 5 65 30 124 20.4 59.5
5 95 0 108 19.1 55.7
Configuration
Note: Passenger demand volume = 3,400 pphpd; route length = 14.6 mi;
With respect to alignment configuration of LRT guideway or BRT number of stations and stops = 16; LRT ¼ 68 seats/vehicle, 2.8 passengers/
running way in Table 9, the capital costs of both LRT and BRT seat; BRT ¼ 50 seats/vehicle, 0.9 passengers/seat; 1 mile = 1.609 km. costs
systems will increase greatly if the system is to be elevated or are in year 2009 U.S. dollars.
put underground, especially in the case of tunneling. The capital
cost of the BRT is more expensive than that of the LRT if there
is a portion of elevated or tunnel alignment. The OM costs of Capital and OM costs with Respect to ROW
BRT systems are more expensive than those of LRT systems. How- Configuration
ever, the OM cost will not change no matter how the alignment More or less exclusive ROW will significantly affect the number of
configuration may change. vehicles required and costs. With respect to ROW configuration,
the capital and OM costs of LRT and BRT in Table 10 operating
Table 8. Capital and OM Costs with Respect to Variables in the Cost on a non-exclusive ROW feature higher capital and OM costs than
Estimating Model: Route Length those operating on an exclusive ROW. LRT vehicles exhibit larger
size vehicles and can carry more passengers than bus vehicles.
Route Capital cost (millions of OM Cost (millions
System length (mi) dollars per mile) of dollars)
With the same passenger demand volume and route length, the
differences of capital or OM costs of the BRT that operating on
LRT 5 47.1 12.9 a non-exclusive ROW and an exclusive ROW are greater than those
10 35.8 19.1 of the LRT. The OM cost of BRT depends primarily on ROW
15 32.1 25.4
configuration.
20 30.2 31.6
25 29.1 37.6
BRT 5 32.0 26.8
10 22.7 42.8 Conclusions
15 19.6 58.9
20 18.0 74.9 Choosing between LRT and BRT systems is often controversial and
25 17.1 91.0 not an easy task for transportation planners. In evaluating transit
Note: Passenger demand volume = 3,400 pphpd; number of stations and systems, cost estimating has always been a major consideration.
stops = 16; LRT ¼ 68 seats/vehicle, 2.8 passengers/seat; BRT ¼ 50 seats/ In this paper, a cost estimating model is developed to estimate
vehicle, 0.9 passengers/seat; 1 mile = 1.609 km; costs are in year 2009 capital cost, annualized capital cost (depreciation cost), OM cost
U.S. dollars. and cost per passenger mile for LRT and BRT systems, to aid

28 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013

J. Transp. Eng. 2013.139:20-29.

 transportation planners and decision makers in selection of the Federal Transit Administration (FTA). (1992). “Characteristics of urban
better mode between LRT and BRT systems in the early corridor transportation systems-revised version.” U.S. Dept. of Transportation
planning stage or system analysis. Only after those costs are calcu- DOT-T-93-07, Washington, DC.
lated can the systems’ comparative analyses be conducted. The Federal Transit Administration (FTA). (2007). “Reporting instructions for
model estimates costs for LRT and BRT systems that operate on the section 5309 new starts criteria.” Report, Office of Planning and
various ROW categories, alignment configurations, and different Environment, Washington, DC.
Federal Transit Administration (FTA). (2008) “Proposed guidance on new
given transit demand volumes.
starts/small starts policies and procedures.” Report, Office of Planning
The model is established through application of unit costs
and Environment, Washington, DC.
available in the historical data of existing systems. However, unit Hsu, J. J. (2003). “Development of a medium-capacity guideway transit
costs might vary to a larger extent with a system’s characteristics, planning guide.” Ph.D. dissertation, Dept. of Civil and Environmental
locations, and various uncertainties, such as possible design Engineering, Florida Int. Univ., Greater Miami, FL.
Downloaded from ascelibrary.org by PRINCETON UNIVERSITY LIBRARY on 10/09/13. Copyright ASCE. For personal use only; all rights reserved.

modifications, unknown environmental and engineering condi- Hsu, L. R., and Wu, J. W. (2008). “A fleet size model for light rail and
tions, construction schedule, and price inflation. Given that accu- bus rapid transit systems.” Transportation Research Record 2063,
rate values are difficult to obtain, the high and low cost ranges of Transportation Research Board, Washington, DC, 34–42.
cost components are applied to provide more reasonable estimates. Khisty, C. J., and Lall, B. K. (2003). Transportation engineering, Prentice-
The average cost derived from calculation of the cost estimating Hall, Englewood Cliffs, NJ.
model can be used for the systems’ comparison process. The sys- Levinson, H., et al. (2003). TCRP Report 90, National Research
tem with the lower cost per passenger mile is the preferable and Council, Transportation Research Board, Washington, DC.
more cost-effective system. Li, J., and Wachs, M. (2000). “A test of inter-modal performance measures
As a simplified planning tool for evaluation and comparison, the for transit investment decisions.” Transportation, 27(3), 243–267.
cost estimating model can be used for evaluation and comparison to Parsons Brinckerhoff Quade & Douglas (PBQD), Inc. (2001). “Technical
distinguish which system would be a better mode for the respective memorandum: Capital cost estimating methodology-north/southeast
area, so long as the passenger demand volume is given. Beyond corridor study alternative analysis/draft environmental impact state-
ment.” Rep. Prepared for Jacksonville Transportation Authority and
LRT and BRT systems, the cost estimating model could be applied
Florida Department of Transportation, Jacksonville, FL.
extensively to various public transportation systems by changing
Schneck, D. C., Amodei, R. M., and Ferreri, M. G. (1991). “Light rail
the variables and unit costs in the process of cost estimating. transit capacity cost study.” Report, U.S. Department of Transportation,
Urban Mass Transportation Administration, Washington, DC.
United States General Accounting Office (USGAO). (2001). “Mass
References
transit-bus rapid transit shows promise.” Report No. GAO-01-984,
Biehler, A. D. (1989). “Exclusive busways versus light rail transit-a com- Washington, DC.
parison of new fixed- guideway systems.” Transportation Research Vuchic, V. R. (2005). Urban transit: Operations, planning, and economics,
Board Special Report, Washington, DC, 89–97. Wiley, Hoboken, NJ.
de Garmo, F. A. (2000). “Bus rapid transit-a feasibility study for Austin, Vuchic, V. R. (2007). Urban transit systems and technology, Wiley,
Texas.” Thesis, Univ. of Texas at Austin, Austin, TX. Hoboken, NJ.
Diaz, R. B., et al. (2004). “Characteristics of bus rapid transit for Watkins, K. E. (2003). “A comparison of bus rapid transit and light rail
decision-making.” Federal Transit Administration Project No. FTA- transit in ten north American cities.” Thesis, Dept. of Civil Engineering,
VA-26-7222-2004.1, Washington, DC. Univ. of Connecticut, Storrs, CT.

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / JANUARY 2013 / 29

J. Transp. Eng. 2013.139:20-29.