Content uploaded by Duy Q. Nguyen-Phuoc
Author content
All content in this area was uploaded by Duy Q. Nguyen-Phuoc on May 08, 2019
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=ujst20
International Journal of Sustainable Transportation
ISSN: 1556-8318 (Print) 1556-8334 (Online) Journal homepage: http://www.tandfonline.com/loi/ujst20
Exploring the impact of public transport strikes on
travel behavior and traffic congestion
Duy Q. Nguyen-Phuoc, Graham Currie, Chris De Gruyter & William Young
To cite this article: Duy Q. Nguyen-Phuoc, Graham Currie, Chris De Gruyter & William Young
(2018): Exploring the impact of public transport strikes on travel behavior and traffic congestion,
International Journal of Sustainable Transportation, DOI: 10.1080/15568318.2017.1419322
To link to this article: https://doi.org/10.1080/15568318.2017.1419322
Published online: 10 Jan 2018.
Submit your article to this journal
Article views: 80
View related articles
View Crossmark data
Exploring the impact of public transport strikes on travel behavior and traffic
congestion
Duy Q. Nguyen-Phuoc
a
,
b
, Graham Currie
a
, Chris De Gruyter
a
, and William Young
c
a
Public Transport Research Group, Monash Institute of Transport Studies, Monash University, Melbourne, Victoria, Australia;
b
The University of Danang
–University of Science and Technology, Danang, Vietnam;
c
Department of Civil Engineering, Monash University, Melbourne, Victoria, Australia
ARTICLE HISTORY
Received 5 April 2017
Revised 15 December 2017
Accepted 16 December 2017
ABSTRACT
Public transport (PT) disruption can occur due to various factors such as malfunctions and breakdowns of
vehicles, power outages, and personnel strikes. This paper explores the network-wide impacts of PT strikes
(train, tram, and bus strikes) on traffic congestion in Melbourne, Australia using a network modeling
approach. A primary survey aimed to investigate the mode shift of users when each public transport
mode ceases was conducted with 648 public transport users in May 2016. Findings show that train
withdrawal was expected to result in 43% of users shifting to car. Smaller yet significant shifts to car was
also expected with bus withdrawal (34%) and tram withdrawal (17%). Based on the survey results and the
use of a four-step transport model, train withdrawal was expected to increase the number of severely
congested road links by 130% and reduce the average travel speed from 48 km/h to 39 km/h (20%
decrease). Bus and tram withdrawal was also found to increase congestion although the result was less
severe. Future research should investigate the switching behavior in actual withdrawal events and explore
the long-term effects of public transport withdrawal.
KEYWORDS
Congestion; network-wide;
personnel strikes; public
transport; transport model;
withdrawal
1. Introduction
Traffic congestion is a major issue in the daily lives of com-
muters, especially those living in big cities. As the number of
vehicles on the road network grows, congestion has an
increasing direct effect on commuters. For example it is
expected that congestion costs in Australian cities could be
over $20 billion AUD by 2020 (Garnaut, 2012). In order to
reduce the effect of traffic congestion, public transport (PT)
offers a method of increasing person throughput. In Mel-
bourne, PT carries over 9% of travel and this figure is
expected to increase in the future (BITRE, 2015). The value of
PT in terms of traffic congestion relief is often graphically
demonstrated when strikes causing a withdrawal of PT occur.
The withdrawal of an entire PT system or an individual PT
mode can be expected to have significant effects on traffic
congestion as a share of PT users may shift to car (Blumstein
& Miller, 1983, Exel & Rietveld, 2001, Exel & Rietveld, 2009).
Other users could switch to nonmotorized modes such as
cycling or walking. For those who cannot find an appropriate
alternative mode, they might cancel their trips. In the event of
a PT mode withdrawal, a number of PT users would be
expected to switch to other PT modes that still operate. This
mode shift would put pressure on other PT modes as the
number of passengers suddenly increases during withdrawal.
PT service withdrawal can result from a variety of factors
including malfunctions and breakdowns, power outages, and
labor strikes (Pnevmatikou et al., 2015). Recently, PT strikes
have occurred more frequently in several large cities around
the world. For example, in 2015, Melbourne’s PT system expe-
rienced tram and train strikes, as unions negotiated working
conditions. However, as strikes only occur on an irregular basis,
only limited research has explored the effects of PT strikes on
travellers (Exel & Rietveld, 2001). In addition, since strikes can
demonstrate how PT acts to relieve congestion, we only have a
limited understanding of these benefits due to the limited
research in this area. Thus, it is important to investigate the
behavioral reaction of users to PT withdrawal and their effects
on transit and road traffic congestion. Based on the findings,
appropriate remedial actions can be proposed and imple-
mented to better mitigate the impact on the transport system.
For example, in The Hague, the government allowed travellers
to park on downtown bus lanes and tramways in the event of a
PT strike. In New York, on-street parking was banned to
increase road capacity through the city during a strike (Exel &
Rietveld, 2001).
This paper aims to explore how PT users change their
travel behavior if individual PT modes cease in the short term
(a full-day strike on a weekday). The network-wide impact of
each PT mode withdrawal on traffic congestion is also
investigated.
The paper is organized as follows: a review of available
research regarding the behavioral reactions of PT users when
PT withdrawal occurs. Research on impacts on traffic is then
CONTACT Duy Q. Nguyen-Phuoc nguyen.duy@monash.edu Public Transport Research Group, Monash Institute of Transport Studies, Monash University,
Melbourne, Victoria, Australia 3800.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/ujst.
© 2017 Taylor & Francis Group, LLC
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION
2017, VOL. 0, NO. 0, 1–11
https://doi.org/10.1080/15568318.2017.1419322
presented. This is followed by a description of Melbourne’sPT
system as the context for this research. The research methodol-
ogy is presented. Results are then described. The final section
concludes with major findings, policy implications, and makes
suggestion for further research in this area.
2. Background
Although the withdrawal of PT, particularly PT strikes, has
occurred more frequency in recent years (Exel & Rietveld,
2001), studies in this area are very limited. Exel and
Rietveld (2001) reviewed 13 studies of PT strikes between
1966 and 2000 in Europe and the United States to explore
the behavioral reaction of PT users. The impact of PT
strikes varies depending on the type of strike, usual travel
patterns, and the policy response. They found that when
PT ceases, PT users would switch to car (ranging from
20%–67% of PT users), switch to other modes (23%–51%),
or cancel their trips (15%–67%). Exel and Rietveld (2009)
carried out secondary analysis on data collected from 976
people who had planned to travel by train on the day of a
national rail strike in the Netherlands in 2004. The main
purpose of their study was to explore the behavioral reac-
tions of train travellers to a rail strike and investigate the
characteristics of travellers and trips that affect the chosen
alternative. They found that 24% of train travelers shifted
to car as a driver, 14% shifted to another mode, and 18%
decided to reschedule their trips to another day. Overall,
44% of trips were cancelled on that day. Factors affecting
the behavioral reactions of PT users in the event of a PT
withdrawal have also been explored in a number of studies
(Exel & Rietveld, 2009,Pnevmatikouetal.,2015).
There has been a number of studies assessing the impact of
PT strikes on traffic congestion. Lo and Hall (2006) explored
the impact of PT strikes that took place in Los Angeles over a
35-day period in October and November 2003. Traffic condi-
tions during the strike were measured to understand how PT
affects congestion experienced by car drivers. They measured
the traffic speed on freeways before and after the strike by using
various sensors. They found that there was a traffic speed
decrease of 20% during the strike. Using the same method,
Adler and van Ommeren (2016) assessed the congestion relief
benefit of PT with the use of quasinatural experimental data on
multiple PT strikes which took place over the period from 2001
to 2011 in Rotterdam, Netherlands. They found that car speed
within the city reduced by around 10% while there was a reduc-
tion of approximately 3% for highways during a citywide strike.
Another assessment on the effects of PT strikes was conducted
by Bauernschuster et al. (2017). By analysing 71 one-day strikes
in PT from 2002 to 2011 in five largest cities in Germany, travel
time was found to increase by 9.3%.
Anderson (2013) explored whether PT strikes generate a
much larger congestion impact than earlier estimates using a
choice model and data from a sudden strike in 2003 by Los
Angeles PT workers. A regression discontinuity design was
used to calculate the travel delay if PT is not available. He found
that the average highway delay would increase by 47% during
peak hours when PT ceases, particularly adjacent to rail corri-
dors (Lo and Hall, 2006). This indicates that high quality,
grade-separated services (such as trains) are considerably effec-
tive at reducing congestion. Similarly, the research of Laval
et al. (2004) pointed out that there were severe traffic problems
on roads when a disruption of the Bay Area Rapid Transit
(BART) system in San Francisco occurred. The absence of
BART services on major East Bay corridors would generate
morning traffic queues stretching 26 miles with 9 mi/hr speeds,
and afternoon queues stretching 31 miles with 11 mi/hr speeds.
More recently, Moylan et al. (2016) investigated the impacts of
rapid transit in the San Francisco Bay Area region on roadway
travel demand and travel time when PT services are suspended
during a strike. In order to estimate the lower bound of the
impact, they compared traffic volumes, which were collected
from a system of 2,000 buried-loop-detector stations on free-
ways during the strike, against observations from the same
time and day of week throughout the year. In contrast, the
upper bound of the impact was measured using an experiment.
They assumed that all PT users with access to a car would shift
to driving alone. A nonparametric modeling technique was
then used to compare the travel time distributions associated
with the traffic volume and travel demand. They found that at
the network level, the impact of the BART strike was not signif-
icant. However, on roads running parallel to PT services, there
were significant delay, particularly in the peak periods.
Morning peak conditions on a parallel road (Highway 24) were
nearly at the 80th percentile of annual volume-weighted travel
times.
Aftabuzzaman et al. (2010b)exploredtheimpactsof
individual PT modes (train, tram, and bus) on trafficcon-
gestion relief in Melbourne. They used a four-step transport
model and assumptions relating to the diversion of PT
users to car when PT is removed. From secondary research,
they suggested that on average 32% of PT users would shift
to car if PT was not available. This fixed value was also
applied for individual PT modes if they were separately
removed. The modeling was based on removing the PT sys-
tem from the network and reallocating of transit trips to
car travel. They found that Melbourne’strainoperations
havethegreatestimpactoncongestionreliefacrossallsub-
urbs, reducing the number of congested links by 23%. This
was followed by bus and tram which reduced the number
of congested links by 13% and 9% respectively. As can be
seen from this research, only a negative impact of PT with-
drawals on trafficcongestion(themodeshiftfromPTto
car) was concerned.
It can be seen that there has been a lack of comprehen-
sive method that can be used to estimate the congestion
impact of PT strikes. This paper builds on the analytical
process used by Aftabuzzaman et al. (2010b) in estimating
the impact of individual PT mode withdrawal on traffic
congestion. However, this research incorporates the results
of a field survey of PT users to determine how mode shift
to car would vary in different parts of the city (inner, mid-
dle, and outer). This method is considered to be more pre-
cise as changes in travel behavior are accounted for
spatially (Nguyen et al., 2015). In addition, the positive
impacts of PT withdrawals on trafficcongestionsuchas
the removal of at-grade rail crossings and tram operations
are also considered in this study.
2D. Q. NGUYEN-PHUOC ET AL.
3. Research context
3.1. Melbourne’s PT system
Melbourne has a population of 4.42 million people over nearly
2,000 km
2
. The Central Business District (CBD) plays a domi-
nant role for many forms of retailing, employment, and recrea-
tion. Like other cities in Australia, Melbourne has a high
dependency on the automobile with a total of 3.6 million pri-
vate vehicle trips per day. Melbourne has an integrated public
transport system that extends from the city center in all direc-
tions, with trains, trams, and buses offering comprehensive PT
services. The PT system in Melbourne carries 9% of all trips
within the metropolitan area, or 11% when expressed in terms
of passenger kilometers (Currie & Burke, 2013). As shown in
Figure 1, Melbourne’s PT system consists of train, tram, and
bus services.
Melbourne’s metropolitan train network consists of 16 lines
with a total length of 372 km (track length of 830 km). The net-
work is primarily at-grade, with more than 170 level crossings.
Melbourne’s train network carried more than 230 million pas-
senger trips in 2014 (DOT, 2014).
Tram is a major form of public transport in Melbourne, with
250 km of tram track and 25 routes. It is the largest urban tram-
way network in the world (Currie et al., 2012). Tram is the sec-
ond most used form of public transport in Melbourne after the
commuter railway network, with 183 million passenger trips in
2015 (PTV, 2015). The majority of the tram network is located
in the inner city area.
Melbourne has a total of 346 bus routes carring over 127 mil-
lion passenger trips in 2015 (PTV, 2015). While the city relies
on a radial train network and inner city tram network, the outer
suburbs are primarily serviced by bus. Buses normally operate
in mixed traffic conditions although there are several exclusive
bus lanes provided for premium bus services.
3.2. Spatial unit of analysis
Local Government Areas (LGAs) are the base unit of analysis
used in this study. There are 31 LGAs in Melbourne (VicRoads,
2005) which are grouped into three categories: inner (4 LGAs),
middle (14 LGAs), and outer (13 LGAs). These are shown in
Figure 2.
4. Study methodology
This section describes the methodology developed to estimate
the impact of individual PT mode withdrawal on traffic conges-
tion. Firstly, the method used to estimate mode shift to car is
presented. Secondly, the method used for assessing the net-
work-wide impact of PT withdrawal on traffic congestion is
described.
4.1. Primary research
Strikes do not occur often, a survey was carried out to investi-
gate the behavioral reaction of PT users if PT was not available
for their last PT trip in the weekday morning peak (a hypotheti-
cal situation). This period was chosen as the morning peak has
the highest level of traffic congestion so the effect of PT strikes
is expected to be the highest during this period. The online sur-
vey of PT users across metropolitan Melbourne (inner, middle,
and outer) was conducted in April 2016. A sampling frame
Figure 1. Public transport network in Melbourne.
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 3
targeting spatial spread and demographics characteristics
(gender and age) was used to make sure the representativeness
of the collected sample. The aim of the survey was to under-
stand the behavioral reaction of PT users in the event of a PT
withdrawal. Respondents who used PT in the weekday morning
peak were asked about the impact of individual PT mode with-
drawal and their likely change in travel behavior.
Firstly, an email was sent to all members of a market
research panel inviting them to take part in the study by
answering an online questionnaire. In the email invitation,
each panel member was given a link to access the questionnaire.
A reminder email was sent to those who had not accessed the
questionnaire one week after the initial email was sent. Data
was collected over a three-week period during autumn and
therefore reflected autumn travel behavior. A total of 3,559 peo-
ple accessed the survey in which 648 respondents (18.2%)
passed a screening process. Only regular PT users who used PT
last week in the weekday morning peak (7 am–9 am) were eligi-
ble to access the survey. Out of the 648 respondents, 443 users
confirmed that they used the train, 234 users used the tram,
and 187 users used the bus for their last PT trips (some travel-
lers used multiple PT modes).
There are three main parts in the questionnaire: socioeco-
nomic characteristics, PT trip characteristics, and flexibility in
travel behavior. This questionnaire was designed based on the
results of a qualitative research which explored the mode shift
of PT users in the event of a PT withdrawal (Nguyen-Phuoc
et al., 2016).
1. The socioeconomic part of the survey gathers informa-
tion on gender, age, vehicle ownership, driver’s license
ownership, number of adults with a drivers’license in
the household and weekly income.
2. The second part of the questionnaire was designed to get
information regarding the context of the last PT trips
that respondents undertook such as the locations that
they started and ended their trips, trip purpose, station
accessibility as well as the weather conditions during
their trips. In order to get information about the origin
and destination, PT users were asked to provide the
address and postcode of the location that they started
and ended their trip. With this information, the
researchers could estimate trip distance and identify
whether a trip was to the CBD, the area which is
expected to have a high level of traffic congestion and
high parking costs, or not by using GIS.
3. In the last part of the questionnaire, participants were
asked to imagine that the PT modes that they used in the
last PT trips were not available for the whole day and
they were notified about this disruption. Respondents
were then asked about their likely behavioral reactions
after considering carefully the advantages and disadvan-
tages of each alternative transport modes such as travel
cost and travel time. A choice-set including eight
options: take other PT modes, drive a car, take a lift, take
taxi/Uber, cycle, walk, cancel trip, and other was pro-
vided (Table 1).
These respondents were asked to describe their behavioral
reactions in the event of each PT mode closure. From the
results of the survey, the share of mode shift to other travel
modes for inner, middle, and outer areas could be estimated.
This research has assumed that PT user diversion to car
when each PT mode ceases would have an impact on traffic
congestion. It is clear that the mode shift to a car as a driver
would directly increase the number of car trips on the road
Figure 2. Local Government Areas (LGAs) in Melbourne.
4D. Q. NGUYEN-PHUOC ET AL.
network (diversion to walking or cycling is not considered to
directly influence congestion). However, in the case of switch-
ing to a car as a passenger, this may or may not influence traffic
congestion. For example, Litman (2004) argued that some car
users can spend a significant amount of time driving children
to school, family members to work, and elderly relatives on
errands (chauffeuring trips). These trips can be particularly
inefficient if drivers are required to make an empty return trip
which can contribute to congestion. For the purpose of this
modeling analysis, it is assumed that half of all car passenger
trips involve chauffeuring (Aftabuzzaman et al., 2010a). Thus,
the car mode shift share contributing to traffic congestion if PT
operations cease would be the sum of the share of mode shift to
car as driver and a half of the share of mode shift to car as
passenger.
4.2. Modeling the impact of individual PT mode
withdrawal
A modeling procedure was developed to explore the impact of
individual PT mode withdrawal on traffic congestion. The pro-
cedure adopted an assumption regarding PT user diversion to
car and a conventional four-step transport model (the Victo-
rian Integrated Transport Model, or VITM). The modeling
analysis was carried out for weekday morning peak (7 am –
9 am).
VITM is a conventional four-step transport model used to
estimate travel demand in the Australian state of Victoria. The
model is implemented in a Cube software platform. In VITM,
the road network is represented by a set of links (66,848 links)
and nodes (28,499 nodes), divided into 2,959 zones. Nodes usu-
ally represent an intersection or a change in road characteris-
tics, while links represent the segments of actual roads in the
network. VITM contains a number of submodels which work
together to create the required output for each link such as
speed, volume, and travel time.
In order to simplify the modeling, it is assumed that train
operations and bus operations have no negative impacts on
generating traffic congestion. This assumption is consistent
with the findings of the research conducted by Nguyen-Phuoc
et al. (2017a) which show that the network-wide impact of
at-grade rail crossings on traffic is not significant. Thus, in the
event of a train or bus strike, the mode shift from train or bus
to private car will contribute to the increase in the level of traf-
fic congestion. The removing of at-grade rail crossings and bus
stop operations has no effect on reducing congestion.
Regarding the effect of tram withdrawal on traffic, Nguyen-
Phuoc et al. (2017b) found that tram operations have both posi-
tive and negative effects on traffic congestion. The negative
effects of trams include the impact of low tram speeds, the
impact of curbside tram stops on nonexclusive tram rights-of-
way, and the occupation of priority tram lanes on semiexclusive
tram rights-of-way. This figure is significant (Nguyen-Phuoc
et al., 2017b). If tram withdrawal occurs, the impacts of low
tram speeds and curbside tram stops on trafficflow are
removed. It is assumed that vehicles cannot use priority tram
lanes in the event of tram strikes since the ground level of pri-
ority tram lanes may differ to that of normal traffic roads or
priority tram lanes are separated by barriers. Thus, when
modeling the impact of tram withdrawal, the approach needs
to consider both the positive and negative effects of tram with-
drawal on traffic congestion.
The modeling of PT strike congestion effects undertaken in
this research includes the following major steps (Figure 3):
In the scenario of “Base case”or “With PT”:
Identify the level of congestion on the road network with
the operation of all PT modes. The negative effects of
tram operations on vehicle trafficflow (the impact of low
tram speeds, the impact of curbside tram stops on nonex-
clusive tram rights-of-way) are modeled in this stage by
integrating the results of microsimulation, which models
the impact of trams on a road link, into VITM (Nguyen-
Phuoc et al., 2017b).
Estimate the PT demand for each zone (train trip matrix,
tram trip matrix, and bus trip matrix) by conducting a PT
assignment process.
In the scenario of “Train withdrawal”, buses and trams still
operate:
Determine the additional car trips caused by train service
withdrawal (multiply the train trip matrix by the share of
mode shift to car for inner, middle, and outer areas, which
is calculated from the survey).
Table 1. Questions about mode shift used in the questionnaire.
For train users For tram users For bus users
If only the train system was no longer available for the
whole day of your last public transport trip. How
would you travel to your destination for that trip
(choose only one major transport mode)?
Considering carefully the advantages and
disadvantages of each alternative transport modes
such as travel cost and travel time.
If only the tram system was no longer available for the
whole day of your last public transport trip. How
would you travel to your destination for that trip
(choose only one major transport mode)?
Considering carefully the advantages and
disadvantages of each alternative transport modes
such as travel cost and travel time.
If only the bus system was no longer available for the
whole day of your last public transport trip. How
would you travel to your destination for that trip
(choose only one major transport mode)?
Considering carefully the advantages and
disadvantages of each alternative transport modes
such as travel cost and travel time.
a. take tram a. take train a. take train
b. take bus b. take bus b. take tram
c. drive a car c. drive a car c. drive a car
d. take a lift d. take a lift d. take a lift
e. take taxi/Uber e. take taxi/Uber e. take taxi/Uber
f. cycle f. cycle f. cycle
g. walk g. walk g. walk
h. cancel trip h. cancel trip h. cancel trip
i. other i. other i. other
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 5
Add the additional car trip matrix to the existing car trip
matrix to create a new car trip matrix that represents the
impact of train withdrawal.
Assign this new car matrix to the road network to esti-
mate the level of congestion after train operations are
removed.
In the scenario of “Bus withdrawal,”trains and trams still
operate:
Calculate the additional car trips caused by bus service
withdrawal (multiply the bus trip matrix by the share
of mode shift to car for inner, middle, and outer
areas).
Add the additional car trip matrix to the existing car trip
matrix to create a new car trip matrix that represents the
effect of bus withdrawal.
Assign this new car matrix to the road network to esti-
mate the level of congestion after bus strikes occur.
In the scenario of “Tram withdrawal,”trains and buses still
operate:
The negative impacts of trams on creating traffic conges-
tion are not modeled.
Determine the additional car trips caused by tram service
withdrawal (multiply the tram trip matrix by the share of
mode shift to car for inner, middle, and outer areas).
Add the additional car trip matrix to the existing car trip
matrix to create a new car trip matrix that represents the
impact of tram withdrawal.
Assign this new car matrix to the road network to esti-
mate the level of congestion in the event of a tram strike.
A comparison of the level of congestion between the two
scenarios, “Base case”and “Train withdrawal,”is undertaken to
understand the effect of train withdrawal on traffic congestion.
The impacts of tram and bus withdrawal were also estimated
using a similar process.
Figure 3. The process for estimating travel demand in each scenario.
6D. Q. NGUYEN-PHUOC ET AL.
Figure 3 describes the process to estimate the congestion
impacts of individual PT withdrawals.
5. Results and discussion
The results are presented in two parts. The findings from the
field survey are presented first, followed by the results of the
modeling which detail the impact of PT withdrawal on traffic
congestion.
5.1. Primary research
5.1.1. Respondent characteristics
A total of 648 respondents completed the survey, comprised 323
males (49.8%) and 325 females (50.2%) (Table 2). The highest
proportion (23.1%) of respondents was 30–39 year olds, closely
followed by 18–29 year olds (21.8%), and 40–49 year olds
(20.5%). Users aged from 50 to 59 years accounted for the lowest
proportion of respondents (16.5%). A Chi-squared test was
conducted to compare the gender and age distribution between
the sample and Melbourne’s PT user population from the 2011
Census. The results of the chi-square test showed that there was
no significant difference and that the sample is therefore
representative of the broader PT user population.
Table 3 presents the use of each PT mode among survey
respondents. This indicates that the highest proportion of users
travelled by train (68.1%), followed by tram (36.9%) and bus
(29.3%). These proportions are generally consistent with the
analysis of PT users in the Victorian Integrated Survey of
Travel and Activity (VISTA) and the modeled outcome from
VITM. Given that train is the most utilized PT mode in Mel-
bourne, the withdrawal of train services is expected to generate
the largest effect on traffic congestion of all PT modes.
Figure 4 shows the spatial distribution of the PT trip origins
of survey respondents. Train trips are distributed across all
parts of Melbourne while a high proportion of tram trips are
within inner city. Respondents travelling by bus tend to make
trips from the middle and outer areas rather than in inner city.
5.1.2. Mode shift
Based on the survey results, Table 4 provides information about
the stated behavioral reactions to each PT mode withdrawal
among PT users. In the event of a train withdrawal, a relatively
high proportion of train users would shift to car as a driver
(39.4%), particularly in outer areas where the mode shift is
55.0%. The number of users switching to other PT modes
(tram and bus) accounts for around 40% of train users in total.
Nonmotorized modes were chosen by less than 5% of train
users, while 6.6% said that they would cancel their trips.
In the event of a tram withdrawal, 34% of tram users would
switch to train, while only 12% would shift to bus. In the inner
city, a relatively high proportion of tram users would choose to
walk (25.2%), which is much higher than the proportion which
would choose to walk in the event of a train withdrawal (2.7%).
The number of tram users who would shift to car as a driver
accounted for only 15%.
The highest share of bus users 28.9% would shift to car as a
driver as a result of a bus withdrawal. This is followed by mode
shift to train (23.5%) and tram (11.8%). Only 11% of bus users
would choose to walk while around 9% would cancel their trips.
Table 4 also shows the share of mode shift to car when indi-
vidual PT modes cease. Train withdrawal is expected to gener-
ate the highest mode shift to car (42.7%). This is followed by
bus withdrawal and tram withdrawal with 33.5% and 16.7%
respectively. These figures are substantially different for each
part of metropolitan Melbourne, reflecting the traffic character-
istics of those areas (Nguyen et al., 2015). For example, in the
event of a train withdrawal, mode shift to car in outer areas is
nearly triple that for the inner city. In contrast, mode shift to
car in outer areas is the lowest if tram operations cease, reflect-
ing the predominance of the tram network in the inner and
middle areas. These figures are used in the four-step transport
model (VITM) to examine the expected changes in traffic con-
gestion during PT withdrawal.
5.2. Modeling results
Table 5 shows the increase in the number of car trips caused by
individual PT mode strikes in the AM peak period
(7am–9am). It can be seen that the number of car trips increase
by over 7% in the event of a train strike while this figure is 1%
and 2% for tram and bus strikes respectively.
Table 6 reveals the impact of individual PT mode with-
drawal on Melbourne’s road network. A comparative assess-
ment of all measures indicates the following:
Train withdrawal causes an increase of over 130% in the
number of severely congested links while the withdrawal
of trams and buses results in an increase of only 7.3% and
17.2% respectively.
There is an increase in the number of vehicles experienc-
ing congestion when train, tram, and bus cease
Table 2. Demographic profile of respondents.
Survey Census
*
Characteristic
Number of
respondents (n)
Proportion
(%)
Expected
value (n)
Proportion
(%)
Chi-
squared x
2
Gender Male 323 49.8 322 49.7 0.0031
Female 325 50.2 326 50.3 0.0031
Age 18–29 141 21.8 152 23.5 0.8582
30–39 150 23.1 127 19.6 3.5267
40–49 133 20.5 122 18.8 0.9098
50–59 107 16.5 102 15.7 0.2336
60C117 18.1 145 22.4 6.7009
Total 648 100 648 100
Population with a journey to work by PT in Melbourne (2011 Census).
x2
Gender 0:062ðÞ<x2
Critical(6.635), x2
Age 12:229ðÞ<x2
Critical 13:227ðÞ,p-value D0.01.
Table 3. PT mode distribution of users in Melbourne.
Survey VISTA Modeled in VITM
No. % No. % No. %
Train 441 68.1 1,193 68.7 214,511 76.7
Tram 239 36.9 431 24.8 77,418 27.7
Bus 190 29.3 467 26.9 81,247 29.0
Total
*
648 1,737 279,688
Total is not 100% because a number of users travelled by multiple PT modes.
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 7
(considering the impact of each PT mode separately) of
57.5%, 2.8%, and 8.8% respectively.
Train withdrawal increases total delay on the road net-
work by 87.1%, whereas tram and bus withdrawal
increases delay by only 2% and 8.2% respectively.
Average travel speed decreases from 47.9 km/h to 38.6
km/h as a result of train withdrawal. Similarly, the travel
speed also deceases when tram or bus operations cease.
However the level of reduction is relatively low (0.8% and
3.0% respectively).
When train services are not available, actual travel time
per kilometer increases by nearly 73%. This is much
higher than the increase in travel time in the event of a
tram or bus withdrawal (1.7% and 6.6% respectively).
Overall, it can be seen that train withdrawal results in a
much higher impact on traffic congestion compared to the
withdrawal of tram or bus services. Figure 5 illustrates the spa-
tial distribution of congested links in the “Base case”scenario
and “PT withdrawal”scenario. This shows that there is a con-
siderable increase in traffic congestion, particularly in the inner
and middle areas in the event of a train strike. Given the
increase in the number of car trips caused by bus strikes dou-
bles that caused by tram strikes, the traffic congestion effect of
bus strikes is more than three times than that of tram strikes.
This is because the removal of trams has higher impact on
reducing traffic congestion, particularly on roads with tram
operations.
6. Discussion and conclusion
This paper has investigated the impact of individual PT mode
withdrawals, which occur due to personnel strikes, on travel
Figure 4. Distribution of PT trip origins among respondents.
Table 4. Behavioral response of PT users when each PT mode ceases in the short term.
Train (%) (nD433) Tram (%) (nD234) Bus (%) (nD187)
Behavioral reactions Inner Middle Outer Total Inner Middle Outer Total Inner Middle Outer Total
Train ———— 29.3 37.5 43.6 34.2 28.2 26.8 18.2 23.5
Tram 45.5 21.2 4.1 20.7 ————28.2 8.5 6.5 11.8
Bus 14.3 21.8 20.5 19.4 11.4 13.9 12.8 12.4 —— ——
Car as driver 18.8 37.2 55.0 39.4 13.8 19.4 10.3 15.0 23.1 32.4 28.6 28.9
Car as passenger 4.5 6.4 8.2 6.6 1.6 5.6 5.1 3.4 2.6 8.5 13.0 9.1
Taxi/uber 3.6 1.9 1.2 2.1 4.9 1.4 5.1 3.8 0.0 4.2 1.3 2.1
Cycle 6.3 1.3 0.6 2.3 7.3 2.8 2.6 5.1 7.7 0.0 2.6 2.7
Walk 2.7 2.6 1.2 2.1 25.2 11.1 10.3 18.4 5.1 9.9 15.6 11.2
Cancel the trip 2.7 7.1 8.8 6.6 4.9 5.6 7.7 5.6 5.1 8.5 10.4 8.6
Other 1.8 0.6 0.6 0.9 1.6 2.8 2.6 2.1 0.0 1.4 3.9 2.1
Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0
Mode shift to car
*
21.1 40.4 59.1 42.7 14.6 22.2 12.9 16.7 24.4 36.7 35.1 33.5
Mode shift to car Dmode shift to car as driver C0.5 £mode shift to car as passenger.
8D. Q. NGUYEN-PHUOC ET AL.
behavior and traffic congestion. A literature review revealed
that most studies exploring the impact of PT withdrawal on
congestion have adopted simple assumptions on car diversion
from PT. The share of mode shift to car has been used in a
number of mathematical and simulation modeling studies to
examine the effect of PT withdrawal on congestion. However,
due to the lack of primary data, prior studies assumed that
either all PT users would shift to car, or that a total fixed share
would shift based on secondary data. These methodologies are
considered limited and simplistic. In this study, a primary sur-
vey was designed based on the findings of a qualitative research
conducted by Nguyen-Phuoc et al. (2016) to measure changes
in PT user behavior if PT ceases in the short term. The findings
showed that when individual PT mode withdrawal occurs, a
high proportion of users (35%–47%) would switch to other PT
modes. Other users would shift to car as a driver (15%–39%) or
a passenger (3%–9%), taxi/Uber (2%–4%) or a nonmotorized
mode (4%–24%). Only a small group (6%–9%) would choose
to cancel their trips. However, due to the ethics requirements,
PT users who are under 18 years old were not selected to partic-
ipate in the survey. In fact, a proportion of people in this age
group use PT to travel to school and they are likely to switch to
private car as a passenger in the event of a PT strike. Thus,
mode shift from PT to car explored in the survey addressing
only adults who have a higher chance of switching to car would
be overestimated. This potentially leads to overestimation of
the traffic congestion impact of PT strikes. In further research,
this bias needs to be addressed to increase the accuracy for the
developed methods.
The mode shift to travelling by a car when PT is removed is
influenced by a set of factors such as driver’s license, car owner-
ship, health concerns, the number of vehicles in a household,
trip destination (in the CBD or not), and station accessibility
(Nguyen-Phuoc et al., 2017c). Age and gender were not found
to affect mode shift to car in the event of PT strikes. Mode shift
would be different for cities with different characteristics (Exel
& Rietveld, 2001) which leads to the difference in the traffic
congestion impact of PT strikes.
In the event of a train or bus withdrawal, a high proportion
of PT users (29%–39%) would switch to driving a car. This is
because the majority of train and bus users have long distance
trips which cannot be taken by nonmotorized modes such as
cycling or walking. Shifting to other PT modes is not always an
appropriate alternative, particularly for users living in outer
areas where PT services limited. In contrast, during a tram
withdrawal, a high share of tram users (47%) would switch to
train and bus as trams operate mostly in inner and middle areas
where other PT modes can be accessed easily. Walking is also
chosen by many tram users (18%) because a large proportion
of trips taken by tram are within walking distance.
Based on the share of mode shift to car for inner, middle,
and outer areas, a transport network modeling (VITM) was
undertaken to examine the impact of individual PT mode with-
drawal on traffic congestion. The network-wide impact of train
and bus strikes on traffic congestion is firstly explored in this
study by considering the negative effect of mode shift to car on
traffic. The positive effects of train strike is not taken into
account as these effects are not significant as explored in a
study conducted by Nguyen-Phuoc et al. (2017a). On the other
hand, the effect of tram strikes is modeled by considering both
the positive and negative effect of tram strikes. The negative
effect of tram strikes is estimated by using mode shift to car
which was obtained from a field survey and varied for regions.
This differs to the method used by Nguyen-Phuoc et al.
(2017b) which used the fixed mode shift obtained from second-
ary data for all areas. Additionally, the positive effect of tram
strikes is modeled by considering only the removal of tram
operations in mixed traffic. The priority tram lanes are not con-
sidered to return to general purpose lanes in the event of tram
strikes as there is the difference in elevation between tram lanes
and normal traffic lanes.
Overall, train withdrawal shows the highest impact on traffic
congestion, followed by bus and tram withdrawal. A key reason
is that train is used by the majority of PT users (approximately
70%) and the share of mode shift to car is higher than that for
tram and bus withdrawal. The increase in car trips associated
with train withdrawal can therefore lead to significant impacts
Table 5. Increase in the number of car trips due to PT strikes in the AM peak
period.
Car trip
Additional car trip caused
by mode shift from PT
New
car trip Increase (%)
Train withdrawal 1,309,476 93,181 1,402,657 7.1
Tram withdrawal 1,309,476 12,656 1,322,132 1.0
Bus withdrawal 1,309,476 26,069 1,335,545 2.0
Table 6. Impact of individual PT mode withdrawal on the road network.
Train withdrawal Tram withdrawal Bus withdrawal
Measure Base case Difference (%) Difference (%) Difference (%)
Number of severely congested road links (V/C>D0.9) (Semcog, 2011) 2,125.0 4,945.0 132.7 2,280.0 7.3 2,491.0 17.2
Number of moderately congested road links (V/C>0.8) (Semcog, 2011) 1,931.0 2,172.0 12.5 1,992.0 3.2 2,069.0 7.1
Length of congested road links (km) 1,173.9 2,104.3 79.3 1,221.1 4.0 1,315.6 12.1
Congested road links (%) 9.1 16.4 79.3 9.5 4.0 10.3 12.1
Number of vehicles experiencing congestion (millions) 16.94 26.68 57.5 17.41 2.8 18.43 8.8
Vehicle distance travelled (million veh-km) 15.02 17.06 13.5 15.09 0.5 15.32 1.9
Vehicle time travelled (million veh-hr) 0.38 0.72 86.5 0.39 2.0 0.42 8.1
Total delay on road network (million veh-hr) 22.84 42.73 87.1 23.31 2.0 24.71 8.2
Average travel speed (km/h) 47.9 38.6 -19.5 47.5 -0.8 46.5 -3.0
Actual travel time per km (min/km) 1.82 3.16 73.1 1.86 1.7 1.95 6.6
Notes: V/C: volume to capacity ratio Dtraffic volume divided by road capacity.
The t-test is used to test the significant of these above measures. The p-value for all of measures is less than 0.001.
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 9
on traffic congestion. This is consistent with the findings of
Anderson (2013) who showed that grade-separated services
(such as trains) are considerably effective in reducing conges-
tion than other PT modes.
The congestion effect of PT strikes varies for different cities
based on mode shift from PT to car. This figure was found to
depend on the traffic characteristics of each areas (Nguyen-
Phuoc et al., 2017c). For example, Melbourne which has a
much higher car ownership rate than Amsterdam so when PT
strikes occur, the congestion effect of PT strikes in Melbourne
is higher than that in Amsterdam.
The key contributions of this paper are:
Understand the changes in travel behavior among PT
users when each PT mode ceases in the short term.
Develop a more precise method to explore the network-
wide effect of individual PT mode withdrawal on traffic
congestion by considering the spatial distribution of
expected mode shift to car as well as the removal of the
congestion creation impacts of PT operations in the event
of PT withdrawals.
Determine which PT mode trike has the most impact on
traffic congestion. Based on that, appropriate policies
could be proposed to deal with it.
The results of this paper can help authorities and policy
makers to estimate the effect of individual PT mode withdrawal
on traffic congestion. From that, a measure or a number of
measures can be better targeted to deal with these issues. For
instance, the frequency of alternative PT modes can be
increased in areas experiencing the high levels of traffic
congestion during PT strikes. Other policies could be proposed
such as allowing vehicles to travel or park in priority bus lanes
or tram lanes if these PT modes cease, thereby increasing road
capacity during strikes. In cities which have bike-sharing sys-
tems, they should be free in the event of a PT strike as it can
encourage mode shift from car users and reduce the level of
congestion. Increasing the number of real time passenger infor-
mation systems is another measure to build sustainable trans-
portation system, particularly in the event of a PT strike.
Different types of information prejourney, enroute, and post-
journey, which can be provided by using appropriate channels
before and during a journey, is extremely valuable for the road
users as it can affect their travel behavior (Beecroft &
Pangbourne, 2015, Papangelis et al., 2016). Finally, the govern-
ment should take strong balanced actions to avoid potential PT
strikes where feasible. PT systems are commonly considered to
be the most sustainable motorized transportation systems.
Given a short-term PT strike is not strong enough to break a
travel pattern or behavior, it can have a negative effect on the
future use of PT of a group of users, particularly infrequent
users who can shift to their common transport modes easily or
young travellers who can choose other modes someday (Exel &
Rietveld, 2001). Therefore, what experiences with PT will shape
an image in users which can enhance or reduce the use of PT.
The developed model can be applied for other cities to assess
the congestion effects of PT strikes. The model adopted a four-
step transport model and the mode shift from PT to car.
Recently, transport network models have been developed and
used in most major cities to predict the flow of vehicles on the
Figure 5. Distribution of congested road links in Melbourne.
10 D. Q. NGUYEN-PHUOC ET AL.
road network. Additionally, it is possible for other cities to con-
duct a field survey in an actually PT strike or using a hypotheti-
cal situation (if PT strikes do not appear often) in order to
determine the mode shift to private car. There have been a
number of studies exploring the effect of PT strikes which
assume a one hundred percent of PT users shifting to car
(Moylan et al., 2016, Schrank et al., 2012). This is unrealistic as
users can shift to other transport modes. The model developed
in this study takes into account both positive and negative
effect of a PT strike. Thus, it is considered to improve greatly
the accuracy of the estimations.
There are two key limitations associated with the find-
ings reported in this paper. Firstly, the survey was con-
ducted during autumn. Ideally, the survey should be carried
out in different seasons so that the effect of weather on
stated travel behavior can be determined. Secondly, the
switching behavior in actual withdrawal events should
be observed in order to gain a better understanding about
the mode shift of PT users. This research has estimated the
impact of short-term PT strikes on trafficcongestion.
Future research could therefore look to focus on the impact
of long-term withdrawal (Nguyen-Phuoc et al., 2016). If PT
is unavailable for longer periods, PT users, particularly for
commuters, can cancel or reschedule their trips. Others can
change their workplace or residential location to reduce
travel time (Bauernschuster et al., 2017). Thus, the effect of
the long-term withdrawal of PT is expected to be different.
ORCID
Duy Q. Nguyen-Phuoc http://orcid.org/0000-0002-2299-2368
Graham Currie http://orcid.org/0000-0001-7190-7548
References
Adler, M. W., & van Ommeren, J. N. (2016). Does public transit reduce car
travel externalities? Quasi-natural experiments’evidence from transit
strikes. Journal of Urban Economics,92, 106–119.
Aftabuzzaman, M., Currie, G., & Sarvi, M. (2010a). Evaluating the
congestion relief impacts of public transport in monetary terms.
Journal of Public Transportation,13(1), 1–23. doi:10.5038/2375-
0901.13.1.1
Aftabuzzaman, M., Currie, G., & Sarvi, M. (2010b). Modeling the spatial
impacts of public transport on traffic congestion relief in Melbourne,
Australia. Transportation Research Record: Journal of the Transporta-
tion Research Board,2144(1), 1–10. doi:10.3141/2144-01
Anderson, M. L. (2013). Subways, strikes, and slowdowns: The impacts of
public transit on traffic congestion. Berkeley, United States: University
of California.
Bauernschuster, S., Hener, T., & Rainer, H. (2017). When labor disputes
bring cities to a standstill: The impact of public transit strikes on traffic,
accidents, air Pollution, and health. American Economic Journal: Eco-
nomic Policy,9(1), 1–37.
Beecroft, M., & Pangbourne, K. (2015). Personal security in travel by pub-
lic transport: The role of traveller information and associated technolo-
gies. IET Intelligent Transport Systems,9(2), 167–174. doi:10.1049/iet-
its.2013.0166
BITRE. (2015). Yearbook 2015: Australian Infrastructure Statistical Report.
Bureau of Infrastructure, Transport and Regional Economics (BITRE),
Canberra ACT.
Blumstein, A., & Miller, H. D. (1983). Making Do: The effects of a mass
transit strike on travel behavior. Transportation,11, 361–382.
doi:10.1007/BF00150724
Currie, G., & Burke, M. (2013). Light rail in Australia –performance and
prospects. Australasian Transport Research Forum, Brisbane, Australia.
Currie, G., Delbosc, A., & Reynolds, J. (2012). Modeling dwell time for
streetcars in Melbourne, Australia, and Toronto, Canada. Journal of the
Transportation Research Board,2275(03), 22–29. doi:10.3141/2275-03
DOT. (2014). 2013–14 Annual Report. Department of Transport, Planning
and Local Infrastructure., Melbourne, Australia.
Exel, N. J. A. v., & Rietveld, P. (2001). Public transport strikes and traveller
behaviour. Transport Policy,8(4), 237–246. doi:10.1016/S0967-070X
(01)00022-1
Exel, N. J. A. v., & Rietveld, P. (2009). When strike comes to town…antici-
pated and actual behavioural reactions to a one-day, pre-announced,
complete rail strike in the Netherlands. Transportation Research Part
A: Policy and Practice,43(5), 526–535.
Garnaut, R. (2012). Transforming Transport. The Garnaut climate change
review. Australia.
Laval, J., Cassidy, M., & Herrera, J.-C. (2004). Traffic impact analysis:
Effects of the absence of bart service on major east bay corridors. Insti-
tute of Transportation Studies, UC Berkeley.
Litman, T. (2004). Rail transit in America: A comprehensive evaluation of bene-
fits. Victoria Transport Policy Institute, Victoria, British Columbia, Canada.
Lo, S. C., & Hall, R. W. (2006). Effects of the Los Angeles transit strike on
highway congestion. Transportation Research Part A,40(10), 903–917.
Moylan, E., Foti, F., & Skabardonis, A. (2016). Observed and simulated
traffic impacts from the 2013 bay area rapid transit strike. Transporta-
tion Planning and Technology,39(2), 162–179.
Nguyen-Phuoc, D. Q., Currie, G., De-Gruyter, C., & Young, W. (2016).
Understanding public transport user behavior adjustment if public
transport ceases —A qualitative study. Australasian Transport
Research Forum (ATRF), Melbourne, Australia.
Nguyen-Phuoc, D. Q., Currie, G., De Gruyter, C., & Young, W. (2017a).
Local and system-wide traffic effects of urban road-rail level crossings: A
new estimation technique. Journal of Transport Geography,60,89–97.
Nguyen-Phuoc, D. Q., Currie, G., De Gruyter, C., & Young, W. (2017b).
Net impacts of streetcar operations on traffic congestion in Melbourne,
Australia. Transportation Research Record: Journal of the Transporta-
tion Research Board,2648(1), 1–9. doi:10.3141/2648-01
Nguyen-Phuoc, D. Q., Currie, G., De Gruyter, C., & Young, W. (2017c).
Transit user reactions to major service withdrawal—a behavioral study.
Transportation Research Board 96th Annual MeetingTransportation
Research Board, Washington DC, United States.
Nguyen, P. Q. D., Currie, G., & Young, B. (2015). Public transport conges-
tion relief measurement—A new framework and its impacts. Austral-
asian Transport Research Forum, Sydney, Australia.
Papangelis, K., Velaga, N. R., Ashmore, F., Sripada, S., Nelson, J. D., & Bee-
croft, M. (2016). Exploring the rural passenger experience, information
needs and decision making during public transport disruption.
Research in Transportation Business & Management,18,57–69.
Pnevmatikou, A. M., Karlaftis, M. G., & Kepaptsoglou, K. (2015). Metro
service disruptions: How do people choose to travel? Transportation,
42(6), 933–949. doi:10.1007/s11116-015-9656-4
PTV. (2015). Public Transport Victoria [Online]. Available: http://ptv.vic.
gov.au [Accessed 22 July 2015].
Schrank, D., Eisele, B., & Lomax, T. (2012). TTI’s 2012 Urban mobility
report. Texas A&M Transportation Institute, United States.
Semcog. (2011). Congestion Management Process (CMP). Michigan
Southeast Michigan Council of Governments.
VicRoads. (2005). Austroads Metropolitan Network Performance Data-
base. Melbourne, Australia.
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 11
