You are on page 1of 24

Transportation Planning and Technology

ISSN: 0308-1060 (Print) 1029-0354 (Online) Journal homepage: http://www.tandfonline.com/loi/gtpt20

Methodology to determine service delivery levels


for public transportation

Shreya Das & Debapratim Pandit

To cite this article: Shreya Das & Debapratim Pandit (2016): Methodology to determine
service delivery levels for public transportation, Transportation Planning and Technology, DOI:
10.1080/03081060.2015.1127541

To link to this article: http://dx.doi.org/10.1080/03081060.2015.1127541

Published online: 12 Jan 2016.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=gtpt20

Download by: [Orta Dogu Teknik Universitesi] Date: 13 January 2016, At: 04:01
TRANSPORTATION PLANNING AND TECHNOLOGY, 2016
http://dx.doi.org/10.1080/03081060.2015.1127541

Methodology to determine service delivery levels for public


transportation
Shreya Das and Debapratim Pandit
Department of Architecture & Regional Planning, Indian Institute of Technology Kharagpur, Kharagpur, West
Bengal, India

ABSTRACT ARTICLE HISTORY


Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

When designing transit services, the Level-of-Service concept has Received 29 July 2014
been widely used by transport planners and service providers to Accepted 30 July 2015
assess the service quality of an existing transit system. In addition
KEYWORDS
to the service quality assessment, service providers also need to Bus transit; service delivery
estimate the service levels that will satisfy a maximum number of levels; Level-of-Service; user
users and potential users, across all socio-economic groups, so as satisfaction level; zone of
to maximize patronage. This paper demonstrates a method using tolerance; total utility;
the concept of ‘user satisfaction levels’ and their ‘zone of marginal utility
tolerance’, along with total utility and marginal utility for service
providers, to provide a range of service delivery levels for
individual transit service attributes in the city of Kolkata. This
range of service levels provides a guideline for service providers
within which they can consider making an improvement in
service level. However, the final decision on service improvement
is an outcome of both financial and infrastructural feasibility.

1. Introduction
Transit service planning requires service providers to assess users’ perception of service
quality so that service levels that will satisfy the maximum number of users can be
delivered. For this purpose, many service providers and government agencies believe in
benchmarking service levels using the Level-of-Service (LOS) concept to evaluate the per-
formance of the existing services as well as future services (Das and Pandit 2013a; Gupta
2012; Zaworski 2003; Taylor and Brogan 1978; TCRP 1999). For example, in the United
States, LOS is used to benchmark service quality of highways, signalized intersections as
well as public transportation across six levels/grades from LOS A (best) to LOS F
(worst) (Zaworski 2003). In practice, service providers provide service levels at LOS C
or LOS D, which are perceived as satisfactory service levels for public transportation
according to the Transit Capacity and Quality of Service Manual developed by the US
Transportation Research Board (Zaworski 2003). However, the ranges of LOS C or LOS
D are often quite wide with equally broad implications in terms of cost and user
satisfaction.
The transit service delivery level is a function of a number of factors including the
expected service levels of users and potential users, the level of satisfaction against the
service levels, the cost incurred on providing those service levels and the revenue

CONTACT Shreya Das shreyadas_planner@yahoo.com


© 2016 Taylor & Francis
2 S. DAS AND D. PANDIT

earned by service providers (MRSCW 1994). Researchers have found that users often have
higher expectations than their perceptions would indicate (Cullen 2001; Somaratna, Peiris,
and Jayasundara 2010). Users’ expectations on service levels can be expressed as their
‘zone of tolerance’ for a given service attribute which is bounded by the users’
minimum acceptable service level and their desired service level (Zeithaml, Berry, and
Parasuraman 1993) which differs between socio-economic and demographic groups
(Bhat et al. 2005; Cirillo, Eboli, and Mazzulla 2011; Dantas et al. 2001; dell’Olio, Ibeas,
and Cecin 2010; Zeithaml, Berry, and Parasuraman 1993). Some researchers suggest
that users’ zone of tolerance, that is, their minimum acceptable service, and the desired
service should be used to determine service delivery levels when designing transit services,
which should also incorporate the requirements of minority groups including the phys-
ically challenged population etc. (Mistretta et al. 2009; Pandit and Das 2013).
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

While the zone of tolerance depicts the requirements of the users, it does not reflect the
service providers’ willingness to deliver those service levels. The service providers’ willing-
ness to provide a certain service level is determined by the total utility/disutility obtained
by service providers to provide that service level. The utility/disutility for transit service
providers is represented by the total gain in patronage (Beirão and Cabral 2007; dell’Olio,
Ibeas, and Cecin 2011) that can be translated into the total revenue earned against a given
change in service level. The aim of the service providers is to maximize total utility.
However, like all utility curves, the change in total utility against a given change in
service level diminishes after a certain maximum point – the point of maximum marginal
utility (Bittinger 1999; Yang 2003). Improving service levels beyond the point of maximum
marginal utility may increase the total utility/patronage for the service providers but the
gain in patronage is insufficiently large compared to the investment made to improve
the service level. This is particularly relevant in the case of developing countries where
the transit fare is often controlled and regulated by the government for the public
service operators to cater to a large section of poor and captive riders (Pucher, Korattys-
waroopam, and Ittyerah 2004). The remaining travel demand is supplied by private service
operators (Badami and Haider 2006). The transit fares for these private operators are not
regulated by the government but are often kept low to compete with the government oper-
ators (Badami and Haider 2006). As a consequence, these private operators provide extre-
mely poor transit service quality, hazardous operating and driving practices and
overloading of buses in order to recover the infrastructure costs (Badami and Haider
2006; Mfinanga and Ochieng 2006; Pucher, Korattyswaroopam, and Ittyerah 2004).
While private operators will attempt to maximize their total collection or total utility,
in a service delivery environment with subsizied transit fares, service providers must
restrict their service provision to the point that gives them the maximum return/utility
against the investment made to provide that service level.
In a previous study (Das and Pandit 2015) the authors developed LOS scales to assess
transit service quality based on user perception against individual service attributes, for the
city of Kolkata in West Bengal. In continuation of this work in this paper, the authors have
addressed the next component of transit service quality improvement, that is, determi-
nation of the service delivery levels that will satisfy the maximum number of users and
potential users. The method is demonstrated in this paper using the case study of
Kolkata. The approach adopted is to identify a ‘range’ of service levels against each
service attribute that should be considered for service improvement and a method has
TRANSPORTATION PLANNING AND TECHNOLOGY 3

been designed to determine this using users’ satisfaction levels, their expected service levels
(zone of tolerance) and the maximum marginal utility derived by service providers against
the service improvement. This ‘range’ of transit service delivery level will guide service
providers on what should be their ‘minimum’ and ‘maximum’ service provision against
individual transit service attributes.
This paper is divided into five main sections. Section 2 describes the study area of
Kolkata for which the bus user and potential bus user perception data were collected.
Section 3 describes the bus transit service quality indicators considered in this research.
Section 4 describes the research methodology to determine the range of service delivery
levels for each service attribute. Section 5 discusses the results on the ranges of transit
service delivery levels for different transit service attributes for Kolkata. The paper then
concludes with the major findings.
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

2. Study area
The present research has been carried out using user perception data from bus users and
potential bus users in the city of Kolkata, one of the largest metropolitan cities in India,
with a population of 4,580,544 according to the 2001 Census. Surface transport in
Kolkata includes buses, trams, para-transit (rickshaws and three-wheeler auto rickshaws)
and taxis. The city is also served by an underground metro service and new metro corri-
dors are under construction. The majority of the urban population is heavily dependent on
public transportation, primarily the bus (Maparu and Pandit 2010). Bus services in
Kolkata or any other city in India are the only mode choice available for a considerable
section of society who cannot afford any other mode for making trips which cannot be
covered by walking or bicycling. Thus, government bus transit services in most cities
are highly subsidized with the bus fare being fixed by the government, whereas, private
operators, bounded by the same fare structure and no subsidy to support them, can
only manage to provide a poor level of bus transit operations. The Indian Ministry of
Urban Development has reported a constant decline in the share of public transport in
the last decade across all major cities in India, including Kolkata, and an increasing
trend for private vehicles and intermediate public transport especially the autorickshaw
(MoUD 2008b). Other issues with surface transportation in Kolkata include lack of suffi-
cient road network, traffic congestion, mixed use of motorized and non-motorized surface
transport, an unreliable bus transit system and lack of enforcement of rules and regu-
lations on bus operations and driving practices (Maparu and Pandit 2010, 2012). Travel
demand is served through 401 bus routes operated by both public and private operators
in the city (Maparu and Pandit 2010).

3. Service quality indicators for urban bus transit in Kolkata


Nine bus transit service attributes have been considered in this research which are ‘quan-
titative’ in nature, that is whose service levels can be quantified. These include: ‘delay in
total journey time’, ‘bus stop nearness’, ‘waiting time’, ‘service hours’, ‘crowding level’,
‘seat availability’, ‘number of bus transfers’, ‘on-time performance’ and ‘boarding & alight-
ing time’. These attributes have been identified in three stages following a literature review,
expert opinion survey and pilot bus user survey in Kolkata. The detailed method of the
4 S. DAS AND D. PANDIT

attribute identification process can be obtained from Das and Pandit (2011, 2012). In
addition to these nine attributes, some others were also identified in the process which
are ‘qualitative’ in nature like bus design, ticket purchasing system, etc. which have not
been discussed in this paper. The complete list can be found in Das and Pandit
(2013b). In this paper, the proposed method to determine transit service delivery levels
is demonstrated in detail using two of these nine attributes.

4. Research methodology
A review of the existing literature highlighted that transit service delivery levels should be
determined by incorporating: a) users’ and potential users’ expected service levels,
especially of the minority groups, b) service levels that give maximum satisfaction to
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

both users and potential users, and c) service levels that provide maximum marginal
utility to the service providers (Zaworski 2003; MoUD 2008a; Mistretta et al. 2009;
Walker 2008). In addition to these, the transit riders affordability, willingness to pay
and infrastructure availability or financial capability of the service providers should also
be considered to determine the optimal transit service levels. However, due to limited
data availability, the scope of this research has been limited to the determination of
only a ‘range’ of service levels based on these three identified parameters and not the
actual optimal transit service delivery level.
The method proposes that for each individual quantitative transit service attribute, the
range of transit service delivery level should be bounded by a ‘lower limit’ defined by the
‘minimum acceptable service’ of the ‘users’ or ‘potential users’, or the ‘lower threshold of
LOS C’ as obtained from the LOS scale developed for the given service attribute for the city
concerned, whichever is higher. The LOS scale for Kolkata has been obtained from Das
and Pandit (2015). The ‘upper limit’ of this range is to be determined by the service
level that provides the ‘maximum marginal utility’ for the ‘service providers’, which is
to be discussed later in this section. In case the service level at which marginal utility is
maximum is found to be lower than the minimum acceptable service of the users or poten-
tial users, then ideally there will be no upper limit for the range of service delivery level for
the given service attribute. However, service providers may choose to provide service levels
at the desired service level of the users or potential users (whichever is higher) when there
are no financial constraints. The desired service level, on the other hand, does not incor-
porate the perception of transit service providers but the requirements of the users.
In order to incorporate the expectations of different user groups in the transit service
delivery levels, that reflect the differences in their socioeconomic profile and travel needs,
different user groups were first identified in this research, including captive and choice
riders (as defined in ABS (2008) and Krizek and El-Geneidy (2007)). There were five
income groups: the Economically Poor or EWS (with income <Rs 4000/month), the
Lower Income Group or LIG (with income between Rs 4000–8000/month), the Lower
Middle Income Group or LMIG (with income between Rs 8000–15,000/month), the
Higher Middle Income Group or HMIG (with income between Rs 15,000–30,000/
month) and the High Income Group or HIG (with income >Rs 30,000/month). There
were also three age groups (<30 years, 31–55 years and >55 years), male and female
riders, peak and non peak riders, and physically handicapped riders. User perception
data were then collected from bus users and potential bus users in Kolkata covering all
TRANSPORTATION PLANNING AND TECHNOLOGY 5

these user groups (and discussed in Section 4.1). The percentage composition of these user
groups in the total sample of bus users surveyed in Kolkata is summarized in Table 1.
In the second stage of this research, the ‘minimum acceptable service’ of the users for a
given service attribute was then determined as the ‘average of the mean minimum accep-
table service’ across all user groups identified. This calculation is explained in Section 4.2.
The ‘minimum acceptable service’ is the lower limit of the users’ zone of tolerance and is
defined as the LOS customers are willing to accept, while the ‘desired service’ is the upper
limit of the zone of tolerance and is the LOS representing a blend of what customers
believe ‘can be’ and ‘should be’ provided (Zeithaml, Berry, and Parasuraman 1993). The
minimum acceptable service for the potential users for a given service attribute was deter-
mined in this research as an average of the minimum acceptable service stated by the total
sample of potential users as demonstrated in Section 4.2.
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

In the third stage of the research, the perception of the transit service providers was
incorporated by determining the total utility derived by the service providers by making
service level improvements and was calculated as the cumulative percentage of users
‘very satisfied’ or ‘satisfied’ at different perceived service levels for the service attribute con-
sidered. Then a relationship was established between users’ perceived service levels and the
cumulative percentage of users at designated states of satisfaction against the perceived
service levels, by performing a regression analysis and curve fitting task. Two states of sat-
isfaction were considered: the ‘very satisfied’ users included those who rated their

Table 1. User group share in total sample of bus users in Kolkata.


User groups/categories Symbolic representation Number
Total sample of users Ts 496

Percent share of total users


Captive riders Cap 52
Choice riders Ch 48

Income groups Percent share of total users


<Rs 4000 per month EWS 7
Rs 4000–8000 per month LIG 12
Rs 8000–15,000 per month LMIG 47
Rs 15,000–30,000 per month HMIG 26
>Rs 30,000 per month HIG 8

Gender Percent share of total users


Male M 71
Female F 29

Age Percent share of total users


<30 years Ay 33
31–55 years Am 63
>55 years Aa 4
Percent share of total users
Physically handicapped population Ph 1

Time of the day Total no. of respondents


Morning peak hour Tmp 120
Evening peak hour Tep 41
Non peak hour Tnp 335

Total no. of respondents


Total non-users Non-users 89
6 S. DAS AND D. PANDIT

satisfaction as ‘1’ (very good) or ‘2’ (good), and the ‘satisfied’ users were those who rated
their satisfaction as ‘1’ (very good), ‘2’ (good) or ‘3’ (average). The relationship thus estab-
lished was defined here as the user satisfaction level (USL) derived at different service
levels, which alternately represented the total utility that can be derived by service provi-
ders by making a given change in service level (Beirão and Cabral 2007; dell’Olio, Ibeas,
and Cecin 2011).
In the fourth stage, a first order differentiation was carried out on the total utility curve
obtained for each service attribute to identify the point of maximum marginal utility.
Finally, the range of service delivery level for a given service attribute was determined
by defining the ‘lower limit’ and the ‘upper limit’ of the range of service delivery level
for each quantitative bus transit service attribute as described in the second paragraph
in this section. The proposed method developed in this research has been graphically sum-
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

marized in Figure 1 as an input-output table.


The method described above provides only a range of service delivery level for quanti-
tative transit service attributes. The exact service level to be provided by service providers
against different service attributes is a function of a number of additional factors, including
the cost incurred on providing those service levels and the revenue earned by the service
providers. Due to the limited time frame for this research, the data required to carry out an
optimization between the cost incurred in service provision and the revenue earned from it
could not be collected from the required sources. Hence, the scope of this method was

Figure 1. Input-output table for determination of range of transit service delivery level for quantitative
bus transit service attributes.
TRANSPORTATION PLANNING AND TECHNOLOGY 7

limited to the determination of only a range of transit service delivery levels for transit
service attributes. A different approach needs to be adopted for determining service
levels for qualitative attributes which is not within the scope of this research.

4.1. Collection of user perception data


The data required for this research included users’ satisfaction levels against their per-
ceived service levels and their minimum acceptable and desired service levels. This
research was conducted using customers’ ‘perceived’ service levels and not the ‘actual’
service levels, which may differ from each other (Mishalani, McCord, and Wirtz 2006;
Yen, Teng, and Chen 2001). This was due to difficulty of conducting on-board surveys
of bus users using actual service levels and also due to the unavailability of revealed
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

data from government records. However, it is strongly recommended that future research-
ers should establish a relationship between actual service levels and perceived service levels
for further practical application of the method proposed here.
The final survey asked respondents to state their perceived existing service levels for
different transit service attributes and their levels of satisfaction against these service
levels on a five point ordered categorical scale of measurement from 1 = very good to 5
= very poor. In addition to this, respondents were asked to state their ‘minimum accepta-
ble service’ and their ‘desired service’ against each service attribute. In order to differenti-
tate user perceptions between different user groups, the survey asked respondents to state
their income, age, gender, status on private vehicle ownership, regularity in bus usage,
alternative mode of transport used, time of day of travel and whether physically challenged
or not. The user perception data thus collected were then grouped according to different
user group categories.
The sample size that was estimated for the bus user survey was 400. After incorpor-
ating an expected response rate of 80%, the sample size was revised to 500. Random on-
board surveys were conducted on bus users in Kolkata for this research, along 25 bus
routes across the city, on five different types of buses operating in the city and for all
working days in a week, covering morning peak, evening peak and off-peak hours.
The surveyors started the survey from the trip origin of a given bus route and surveyed
passengers en-route till the bus reached its final destination for both onward and return
journeys. A total of 496 completed responses were collected. In addition to the survey on
bus users, a similar user perception survey was conducted on auto-rickshaw riders that
represented only a section of the non/potential bus user community in Kolkata.
However, unlike the survey conducted on bus users, it was not possible to obtain non
bus users’ levels of satisfaction against existing service levels because these non-users
have either little or no recent experience in using the existing bus transit service.
Thus, the USL curves were only determined for the bus users in Kolkata and not for
the potential bus users and the point of maximum marginal utility determined on the
USL curve/total utility curve is only for the bus users in Kolkata because of the limit-
ation on data collection as discussed above. However, an attempt has been made to
incorporate the expectations of potential users by considering the minimum acceptable
service level of the non/potential users to determine the lower limit of the range of
transit service delivery levels.
8 S. DAS AND D. PANDIT

4.2. Overall zone of tolerance for bus users and potential bus users
In the next step of this research, the overall zone of tolerance (OZOT) for users and poten-
tial users in Kolkata was determined which included the overall minimum acceptable
service of the users and potential users (Smino – users; Smino – p-users) and the
desired service of the users and potential users (Sdeso – users; Sdeso – p-users). In
order to capture this difference in user perception amongst different user groups, initially
the mean zones of tolerance (MZOT) were estimated for each user group for each service
attribute, which is an average of the zones of tolerance amongst all respondents in a par-
ticular user group. The results demonstrated that the user perception and expectation
varies significantly amongst user groups for certain transit service attributes, while for
the others the differences were minimal. In developing countries like India, bus transit
is provided as a social welfare service. Hence, it is important to give importance to all
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

user groups and not to any particular user group, because ultimately the service is to be
used by all and not by any particular user group. Therefore, in this research, equal
weight was given to all the user groups. Finally the overall zone of tolerance (OZOT)
was determined for each attribute, as shown in Equation (1). The zones of tolerance of
different user groups have been summarized in Table 2 and briefly discussed later in
this section.
m
MZOT
OZOT = i=1 , (1)
N
where OZOT = the overall zone of tolerance across all user groups,MZOT = the mean zone of
tolerance amongst users in each user group represented by the mean minimum acceptable
service across all user groups (Sminm) and mean desired service across all user groups
(Sdesm), andm = total number of user groups (i) considered in the research. And,
n
j=1 ZOTi
MZOT i = , (2)
n
where ZOTi = the zone of tolerance, represented by the minimum acceptable service and
the desired service, for an individual user (j) in user group (i) where there are (m) user
groups and n = total number of users in each user group
The overall zone of tolerance represents the average of the mean of zones of tolerance
(MZOT) covering all user groups and provides a broad range of service level that service
providers should consider when designing bus transit services. Giving equal weight to
all user groups ensures that the requirements of all user groups are taken into account
without any undue bias toward any particular user group having a skewed zone of toler-
ance. OZOT indicates the service levels that should be provided to retain the loyality of
existing users. However, in order to attract new users to bus services, the overall zone
of tolerance for non-users also needs to be considered, which will ensure that these
users will consider the bus as a viable transport alternative. The results of the mean
zones of tolerance for different user groups and the overall zone of tolerance for different
bus transit service attributes for bus users and potential bus users (p-users) in Kolkata are
summarized in Table 2. This shows that certain transit service attributes are more impor-
tant for only some users groups, compared to others, and the existing service levels on
these attributes, to some extent, influence the mode choice decision of these user
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Table 2. Mean zone of tolerance (MZOT) and overall zone of tolerance (OZOT) for bus users and potential bus users for bus transit service attributes in Kolkata.
Seat availability
Delay in total Crowding level (percent of On-time No. of mode Boarding-
Waiting time Distance to bus Service hours journey time (passengers/ times seat is performance transfers alighting time
(minutes) stop (meters) (hours) (minutes) seat) available) (percent) (number) (minutes)
User group Sminm Sdesm Sminm Sdesm Sminm Sdesm Sminm Sdesm Sminm Sdesm Sminm Sdesm Sminm Sdesm Sminm Sdesm Sminm Sdesm
Cap 18 8 426 203 14 19 21 3 1.5 0.8 0.6 0.9 0.6 0.9 2 0 1.8 2.9
Ch 15 8 453 244 14 18 17 4 1.3 0.8 0.6 0.8 0.6 0.9 2 0 1.4 1.9
EWS 17 9 624 326 14 17 16 4 1.3 0.9 0.5 0.8 0.7 0.9 2 1 1.3 2.0
LIG 19 8 401 179 15 20 18 3 1.4 0.7 0.7 0.9 0.7 0.9 2 0 1.4 2.1
LMIG 17 8 372 176 14 19 22 4 1.6 0.8 0.6 0.9 0.6 0.9 2 0 1.8 2.8
HMIG 17 10 556 304 15 19 17 4 1.3 0.8 0.6 0.9 0.6 0.9 2 0 1.6 2.1
HIG 13 8 484 288 15 19 14 4 1.2 0.9 0.6 0.8 0.6 0.9 1 0 1.5 1.9
M 17 8 439 220 14 19 20 4 1.5 0.8 0.6 0.8 0.6 0.9 2 0 1.7 2.6

TRANSPORTATION PLANNING AND TECHNOLOGY


F 17 8 423 206 14 18 19 3 1.3 0.7 0.6 0.9 0.7 0.9 2 0 1.7 2.6
Ay 17 8 451 218 14 19 18 3 1.4 0.8 0.6 0.9 0.7 0.9 2 0 1.5 2.3
Am 17 8 433 219 14 19 20 4 1.5 0.8 0.6 0.9 0.6 0.9 2 0 1.8 2.7
Aa 15 8 303 142 15 19 17 4 1.1 0.8 0.6 0.8 0.6 0.9 1 0 1.9 3.0
Ph 20 6 540 280 19 21 36 7 1.0 0.8 1.0 1.0 0.7 0.9 1 0 3.2 4.0
Tmp 15 8 463 238 14 18 21 5 1.4 0.9 0.6 0.8 0.6 0.9 2 0 1.9 2.6
Tnp 18 8 403 189 14 19 20 3 1.5 0.8 0.6 0.9 0.6 0.9 2 0 1.7 2.7
Tep 18 10 631 387 15 19 13 3 1.3 0.8 0.6 0.9 0.6 0.9 2 1 0.9 1.5
Smino Sdeso Smino Sdeso Smino Sdeso Smino Sdeso Smino Sdeso Smino Sdeso Smino Sdeso Smino Sdeso Smino Sdeso
OZOT(users) 17 8 461 237 15 19 19 4 1.3 0.8 0.6 0.9 0.6 0.9 2 0 1.7 2.5
OZOT(p-users) 16 8 369 201 16 20 14 2 1.5 0.9 0.6 0.9 0.6 0.9 1 1 2.2 3.4
Notes: Cap: captive riders, Ch: choice riders, EWS: income <Rs 4000/month, LIG: income Rs 4000–8000/month, LMIG: income Rs 8000–15,000/month, HMIG: income Rs 15,000–30,000/month, HIG:
income >Rs 30,000/month, F: female riders, M: male riders, Ay: bus users of age <30 years, Am: bus users of age 30–55 years, Aa: bus users of age >55 years, Ph: physically handicapped users,
Tmp: users traveling during morning peak hours, Tnp: users traveling during non peak hours, Tep: users traveling during evening peak hours.
The Minimum Acceptable Service is also referred to as the Adequate Service Level which may indicate a higher value of service level for some attribute like on-time performance, while for some
other attributes these may be represented by lower values service levels like waiting time.

9
10 S. DAS AND D. PANDIT

groups. For example, choice riders, users with income >Rs 30,000 per month and morning
peak travelers were found to have a ‘lower’ zone of tolerance for waiting time as compared
to other groups. Similarly, physically handicapped users were found to possess a lower
zone of tolerance for service hours as compared to the other groups. On the other
hand, choice riders, the higher income group population, female riders, the elderly and
the physically handicapped users displayed lower zones of tolerance for crowding level.
These findings may be used by transit service providers in a given city to design specialized
services for specific user groups like higher frequencies of buses for the morning peak
riders, special high-end buses for the higher income groups and choice riders or
women-only buses for the working women population. Such specialized services can
always be provided when there are no financial constraints for the service providers or
when there is enough demand for these services. However, under the given socioeconomic
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

condition in Kolkata, it is not possible to provide such specialized services. Thus it is judi-
cous to give equal weightage to the requirements of all user groups and design services that
will satisfy most of the user and potential user groups.

4.3. Determination of the USL


The USL is defined in this research as the cumulative percentage of users at designated
states of satisfaction, at different perceived service levels, and is expressed in the form
of a relationship trend line for each designated state of satisfaction. In the current research,
two states of satisfaction are considered, one being ‘very satisfied’ and the other just ‘sat-
isfied’ wherein ‘very satisfied’ users are the cumulative population of users who stated their
level of satisfaction as 1 = very good or 2 = good, while the ‘satisfied’ users are those who
stated their levels of satisfaction as 1 = very good, or 2 = good, or 3 = average, for a given
perceived service level. Table 3 shows the number of responses for each satisfaction rating

Table 3. Cumulative percentage of bus users in Kolkata ‘very satisfied’ at different perceived waiting
times.
Number of responses against service level
x and rating k (nxk)
Perceived Very Very Total Cumulative no. Cumulative
service good Good Average Poor poor no. of Cumulative of users ‘very percent of
levels k= users no. of users N satisfied’ (Satis users ‘very
(minutes) x k=1 k=2 k=3 4 k=5 t(x) (m) = 1 + 2) M(m) satisfied’ F1(x)
0 0 1 0 0 0 1 1 1 100
1 1 0 1 0 1 3 4 2 50
2 0 1 1 0 0 2 6 3 50
3 1 2 0 0 0 3 9 6 67
4 0 2 0 0 0 2 11 8 73
5 1 25 27 2 0 55 66 34 52
6 0 1 0 0 0 1 67 35 52
7 0 0 1 1 0 2 69 35 51
8 0 0 3 0 0 3 72 35 49
10 0 32 83 16 0 131 203 67 33
12 0 1 5 1 0 7 210 68 32
15 0 16 99 58 3 176 386 84 22
18 0 1 0 0 0 1 387 85 22
20 1 4 15 58 5 83 470 90 19
25 0 1 1 7 1 10 480 91 19
30 0 0 1 6 4 11 491 91 19
TRANSPORTATION PLANNING AND TECHNOLOGY 11

category (k) against each perceived service level (x) of ‘waiting time’ for bus users in
Kolkata. The USL against ‘very satisfied’ users is defined as a causal relationship
between the cumulative percentage of users rating a given perceived service level as 1 =
very good or 2 = good and the perceived service level. This is explained through the follow-
ing equations:
Let M(m) be the cumulative total number of respondents ‘very satisfied’ at a given
service level (x).
Then,

M (m) = f (x); m ≤ x, (3)
m
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

where


2
f (x) = nxk , (4)
k=1

nxk = number of responses with satisfaction rating (k) for a service level (x).
Let N(m) be the cumualtive total number of respondents rating a service level (x) from 1
to 5.
Then,

N (m) = t(x); m ≤ x, (5)
m

where


5
t(x) = nxk , (6)
k=1

nxk = number of responses with satisfaction rating (k) for a service level (x).
Then, the LSat against ‘very satisfied’ users for a given service level (x) is represented as:

M (m)
F 1 (x ) = × 100. (7)
N (m)

Similarly, the LSat against ‘satisfied’ users for a given service level (x) can be represented as
below. The LSat values for ‘satisfied’ users on perceived waiting times in Kolkata are shown
in Table 4.

Q(m)
F2 (x) = × 100, (8)
N(m)

where

Q(m) = v́(x); m ≤ x (9)
m
12 S. DAS AND D. PANDIT

Table 4. Cumulative percentage of bus users in Kolkata ‘satisfied’ at different perceived waiting times.
Number of responses against service level
x and rating k (nxk) Cumulative no.
Perceived Very Very Total of users Cumulative
service good Good Average Poor poor no. of Cumulative ‘satisfied’ (Satis percent of
levels k= users no. of users N = 1 + 2 + 3) Q users ‘satisfied’
(minutes) x k=1 k=2 k=3 4 k=5 t(x) (m) (m) F2(x)
0 0 1 0 0 0 1 1 1 100
1 1 0 1 0 1 3 4 3 75
2 0 1 1 0 0 2 6 5 83
3 1 2 0 0 0 3 9 8 89
4 0 2 0 0 0 2 11 10 91
5 1 25 27 2 0 55 66 63 95
6 0 1 0 0 0 1 67 64 96
7 0 0 1 1 0 2 69 65 94
8 0 0 3 0 0 3 72 68 94
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

10 0 32 83 16 0 131 203 183 90


12 0 1 5 1 0 7 210 189 90
15 0 16 99 58 3 176 386 304 79
18 0 1 0 0 0 1 387 305 79
20 1 4 15 58 5 83 470 325 69
25 0 1 1 7 1 10 480 327 68
30 0 0 1 6 4 11 491 328 67

and

3
v́(x) = nxk (10)
k=1

nxk = number of responses with satisfaction rating (k) for a service level (x).
The method discussed above has been exemplified in this research considering the attri-
bute of ‘bus stop waiting time’ in Kolkata. Tables 3 and 4 provide the observed USL values
against ‘very satisfied’ and ‘satisfied’ users. In order to obtain the USL for a given service
attribute, a causal relationship was established between the cumulative percentage of users
‘very satisfied’ and ‘satisfied’ against perceived service levels, that is, F1(x) and F2(x) and
perceived service level (x). Thereafter a curve fitting task was undertaken to identify the
family of curve that best fits the observed data. This curve is also called the maximum like-
lihood estimate of the true curve. This was done by identifying the curve whose parameters
generate the smallest weighted sum of squared errors (SSE) when the distribution of the
responses at a specific concentration was approximately normal (Gottschalk and Dunn
2005). Figures 2–6 show five types of curve that were fitted to the observed data on cumu-
lative percentage of ‘very satisfied’ users on different perceived waiting times using Sigma-
Plot version 12.5 software. The results are reported in Table 5 which shows the 4PL curve
to be the best fit which has the highest R 2 and Adjusted R 2 values and the lowest standard
error of estimate as calculated by the SigmaPlot software along with the calculated SSE
which was also lowest for the 4PL curve.
Once the best curve fit was identified, the quality of fit was measured by determining
the residual variance and the cumulative χ2 distribution of the SSE with the given
degrees of freedom (number of data points minus number of parameters) that gives the
best fit probability estimation as discussed in Gottschalk and Dunn (2005).
A similar test was carried out on the observed data of the cumulative percentage of
users ‘satisfied’ on perceived waiting time (from Table 4). The results again indicated
TRANSPORTATION PLANNING AND TECHNOLOGY 13
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Figure 2. Linear curve fitting – waiting time.

that the 4PL curve fit has the highest R 2 value, the lowest SSE and standard error of esti-
mate as compared to other curves. The curve fitting program was also carried out on
observed data on the cumulative percentage of users ‘very satisfied’ and ‘satisfied’ on
the remaining transit attributes considered in this research. In all these cases, the 4PL func-
tion described by Equation (11) was predominantly found to be the most satisfactory in
meeting all prerequisite conditions of best curve fit. The three parameters of the 4PL
curve, that is, min., max., and EC50 represent the lowest point, highest point and midpoint
between the max. and min. parameters of the standard s-curve respectively, while the hill-
slope characterizes the slope of the curve at its midpoint. Large hillslope values result in a

Figure 3. Quadratic curve fitting – waiting time.


14 S. DAS AND D. PANDIT
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Figure 4. Cubic curve fitting – waiting time.

Figure 5. Exponential curve fitting – waiting time.

steep curve whereas small values demonstrate a shallow curve. The 4PL curve will increase
with x if the hillslope is positive and decrease if it is negative.
max − min
F (x) = min + . (11)
1 + (x/EC50)(−Hillsope)
Thus, the 4PL curve, amongst all the family of curves, was found to be the curve that
best explains the variation in the percentage of users that will be satisfied or dissatisfied for
a given change in service level which are expected to have direct implications on transit
patronage [as explained in some human behavioral studies in the field of medical
TRANSPORTATION PLANNING AND TECHNOLOGY 15
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Figure 6. 4PL curve fitting – waiting time.

Table 5. Goodness of fit statistics for linear, quadratic, cubic, exponential decay and four-parameter
curves as shown in Figures 1–5.
Statistic Linear Quadratic Cubic Exponential 4PL
2
R 0.72 0.97 0.99 0.89 1.00
Adjusted R 2 0.70 0.97 0.99 0.88 1.00
Std. error of Estimate 14.01 4.52 3.00 1.15 0.00
F-stat 26.19 169.69 260.27 84.04 +inf
p .0005 <.0001 <.0001 <.0001 <.0001
SSE (calculated) 4.20 0.29 0.11 0.75 0.10
Degrees of freedom (calculated) 10 9 8 10 8
Residual variance (calculated) 0.53 0.03 0.01 0.08 0.01
Fit probability (calculated) 0.82 1.00 1.00 0.99 0.99

science that have also identified the 4-parameter logistic curve for humans’ response to a
given change in medical dose (Gottschalk and Dunn 2005), human adoption of technology
or product life cycle (Wright and Charlett 1995) as well as for other natural processes that
demonstrate ‘learning curves’ or ‘cumulative distribution functions’ (Leibowitz et al.
2010)]. In the above case of perceived waiting time in Kolkata, the s-curve clearly explains
that at higher waiting times, lower percentages of users are ‘very satisfied’ or ‘satisfied’. As
waiting time decreases, the cumulative percentage of users satisfied increases, but the
growth rate is slow up to a certain service level and then at a certain service level (perceived
waiting time) the cumulative percentage of users satisfied suddenly increases steeply till it
reaches a level when the growth rate again diminishes and stabilizes. Equations (12) and
(13) represent the 4PL functions that explain the observed data on cumulative percentage
of users ‘very satisfied’ and ‘satisfied’, respectively, on perceived waiting time for Kolkata.

99.9 − 13.8
F1 (x) = 13.8 + , (12)
1 + (x/5.9)(2.0)
16 S. DAS AND D. PANDIT

where F1(x) = cumulative percent of users ‘very satisfied’ and x = perceived waiting
time.
98.3 − 62.3
F2 (x) = 62.3 + , (13)
1 + (x/15.2)(3.3)

where F2(x) = cumulative percent of users ‘satisfied’ and x = perceived waiting time.
In the field of research that predicts users’ mode choice behavior, many researchers
have applied a binary logistic model to determine the probability that a given individual
will choose the bus as a mode of transport under given changes in service level (cf. Choo,
You, and Lee 2013; Gebeyehu and Takano 2008). While these models can be useful to
predict the sensitivity of different user groups toward changes in service levels of different
transit attributes, this research does not focus on emphasizing the difference in perception
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

of different user groups. Rather, the objective is to provide a guideline by which service
providers will be able to assess the total change in patronage, including all user groups,
under a given change in service level for a given service attribute. Moreover, no data
was available in this research on the satisfaction levels of non-users which did not ration-
alize the application of the ordered logit model in this research. Thus the 4–parameter
logistic model was adopted to represent the USLs for ‘very satisfied’ and ‘satisfied’
users, which are demonstrated for ‘waiting time’ in Kolkata in Figure 7. The USL curve
was then used to assess the performance of the existing service levels on waiting time in
Kolkata with respect tothe LOS scale obtained from Das and Pandit (2015) and the
overall zones of tolerance (OZOT-users, OZOT-p-users) developed for users and potential

Figure 7. USL for waiting time in Kolkata.


TRANSPORTATION PLANNING AND TECHNOLOGY 17

users as discussed in Section 4.2. For simplicity of representation, OZOTs have been
referred to as the ZOT for both users and potential users in this figure.
Figure 7 shows that the existing average perceived waiting time at bus stops in Kolkata
is 13.5 minutes. According to the 5-point LOS scale developed for waiting time in Kolkata
(Das and Pandit 2015), 13.5 minutes is perceived as LOS C by bus users. At this service
level, only 27% of total users are ‘very satisfied’ while 84% are just ‘satisfied’. The existing
perceived waiting time also falls within the overall zone of tolerance for both bus users and
potential bus users in Kolkata but fails to meet their desired service level by a significant
margin. Since only 27% of the bus users are ‘very satisfied’ at this service level, transit
service providers need to improve the existing waiting times in Kolkata.
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

4.4. Estimation of total utility and marginal utility


As explained earlier, USLs are directly related to patronage levels. Improving user satisfac-
tion results in improved patronage and vice versa. While the USL curve developed for
different bus transit service attributes could be used to determine the change in the cumu-
lative percentage of users ‘very satisfied’ or ‘satisfied’ at perceived transit service levels, it
also represents the change in patronage at different service levels which represent the total
utility (TU) obtained by transit service providers. The change in total utility made against
change in service level is called the marginal utility (MU) (Bittinger 1999; Yang 2003). The
4PL functions F1(x) and F2(x) as obtained in Equations (12) and (13) represent the USL for
waiting time in Kolkata which also represents the total utility derived in terms of the total
patronage gained or lost against a given change in service level (x) (Koppelman and Bhat
2006; Yang 2003). Thus the marginal utility derived for a given change in service level can
be represented as ∂TU/∂x and can be obtained as the first order derivative of the function
F(x).
Therefore,
∂TU dF (x)
MU = = = F ′ (x). (14)
∂x dx
According to the properties of marginal utility, if F′ (x) >0 for all values of x in an interval I,
then the function F(x) is increasing on I, whereas if F′ (x)<0 for all values of x in an interval
I, then F(x) is decreasing on I (Bittinger 1999). Further, the law of diminishing marginal
utility explains that, as the rate of service level increases, marginal utility decreases.
However, every function has a critical point in its domain where the tangent to the
curve is horizontal or where F′ (x) = 0. This is the point where the function has either
the maximum or the minimum value (Bittinger 1999). For a given utility function F(x),
it is essential to know the point at which the marginal utility F′ (x) is maximum,
because beyond this point an increase in service level will diminish the rate of increase
in utility although the total utility may increase (Bittinger 1999; Yang 2003). With this
information, service providers may consider limiting their service provision, in times of
resource constraint, up to a level where the marginal utility is maximum (Yang 2003).
This concept can also be adopted and applied when making a decision of transit service
level improvement. Thus, by finding the point where the marginal utilities, that is,
F1’(x) and F2’(x) are maximum on the USL curves denoted by F1(x) and F2(x) for ‘very
satisfied’ and ‘satisfied’ users, transit service providers may limit their service improvement
18 S. DAS AND D. PANDIT

to this level. As discussed earlier, the 4-parameter logistic curve F(x) denotes the USL or
the total utility for the transit service providers. The marginal utility is derived by estimat-
ing the first order derivative of the function F(x) in Equation (11). Therefore, transit
service providers may consider limiting the service improvement at the point where the
value of marginal utility as obtained from Equation (15) is maximum.
 
′ d max−min
MU = F (x) = min +
dx 1+ (x/EC50)(−Hillsope)

 x −Hillsope 2
     −(Hillsope+1)
= (max−min)× Hillslope × EC50Hillslope
× x 1+ .
EC50
(15)
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

5. Range of transit service delivery level


The lower boundary of OZOT, that is, the minimum acceptable service level as described in
Section 4.2, indicates the minimum service level that service providers must aim for. If
existing service levels are below the minimum acceptable service (Smino), then achieving
at least the service level at Smino for bus users is a bare minimum necessity. On the other
hand, in order to attract potential users, service providers must improve services up to the
Smino obtained for the non bus users. Simultaneously, the Smino for both bus users and
potential bus users should be within the range of LOS C (average) or above to ensure that
these service levels are not perceived as poor by users and potential users. Further
improvement in service levels will further increase the USLs or the total utility for
service providers. In the case of ‘delay in total journey time’, reducing the delay to 4
minutes, that is, at the Sdeso for bus users, about 98% of the users would be ‘very satisfied’
(Figure 7); however, the target of 4 minutes delay may often appear to be nearly impossible
in practical situations given the existing surface transport issues in a developing city like
Kolkata. Such high standard of service levels can only be achieved through, for example,
the provision of dedicated bus lanes and additional facilities like signal priority for buses,
etc. when financial resources are available with the service providers. In an alternative and
more realistic scenario service, providers can rationalize the service level improvement up
to the point of maximum marginal utility on the USL curves beyond which the rate of
return on investment made by service providers to improve service levels diminishes, as
demonstrated in Figures 8 and 9.
Figure 8 shows that the marginal utility for the USL curve for ‘waiting time’ on ‘very
satisfied’ users is a maximum at 4 minutes. At this point, 73% of users are ‘very satisfied’.
On the other hand, Figure 9 shows that the marginal utility for the USL curve for ‘waiting
time’ on ‘satisfied’ users is a maximum at 12 minutes. At this point, 87% of the total users
are at least moderately satisfied, which is quite an achievable target. Thus service providers
in Kolkata may consider reducing waiting time at bus stops to either 12 or 4 minutes
depending on the state of satisfaction they wish to address, based on their resource avail-
ability and total utility (patronage) gained. The service level at which marginal utility is a
maximum on the ‘very satisfied’ USL curve for waiting time in Kolkata (i.e. 4 minutes) is
higher than the service level at which the marginal utility is a maximum on the ‘satisfied’
USL curve (12 minutes). This indicates that, ideally, service providers should attempt to
maximize the total number of ‘very satisfied’ users and thus the upper limit of the
TRANSPORTATION PLANNING AND TECHNOLOGY 19
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Figure 8. Total utility and marginal utility against ‘very satisfied’ users for service providers in Kolkata –
waiting time.

range of service delivery level for ‘waiting time’ in Kolkata should be 4 minutes. However,
under resource constraints, service providers may want to focus just on the ‘satisfied’ users
and thus consider 12 minutes of waiting time as the upper limit of the range of service
delivery level. On the other hand, the minimum acceptable service for waiting time
amongst bus users in Kolkata is 17 minutes, whereas the minimum acceptable service
for waiting time amongst potential users is 16 minutes, which is lower than that of the
users. Service providers must ensure that the waiting time at bus stops does not exceed
16 minutes, which will be within the maximum tolerance limit of both users and potential
users. On the other hand, the lower threshold of LOS C is 20 minutes, which is higher than
the minimum acceptable service of the users or potential users. Thus the ‘range of service
delivery level’ for waiting time in Kolkata is 4–16 minutes when service providers want to
maximize the ‘very satisfied’ users and 12–16 minutes when the service providers want to
address the just ‘satisfied’ users.
Using the above method, the range of transit service delivery levels were determined for
the other transit service attributes in Kolkata, as summarized in Table 6. The choice of the
maximum marginal utility on the USL curves depends on the service providers’ goal to
address a particular state of satisfaction, that is, ‘very satisfied’ or just ‘satisfied’ and the
availability of resources.
The ranges of transit service delivery levels as demonstrated in this research are only
indicative ranges whose lower limit is determined based on the minimum service level
requirement of the users and potential users but the upper limit may vary based on the
20 S. DAS AND D. PANDIT
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Figure 9. Total utility and marginal utility against ‘satisfied’ users for service providers in Kolkata –
waiting time.

Table 6. Range of target service levels for quantitative transit service attributes in Kolkata.
Range of target service level for ‘very Range of target service level for ‘satisfied’
satisfied’ users users
Maximum service Minimum service Maximum service Minimum service
Service attribute level/upper limit level/lower limit level/upper limit level/lower limit
Waiting time (minutes) 4 16 12 16
Distance to bus stop (meters) 50 400 100 400
Crowding level (number of 0.7 1.3 1.1 1.3
passengers per seat)
Seat availability (percentage of 65 60 60 60
times seat is available)
Delay in journey time (minutes) 7 14 12 14
On time performance 80 60 60 60
(percentage of times bus is on
time)
Service hours (hour) 16 16 16 16
Number of mode transfers 2 2 2 2

operators perspective and goal. However, the actual service level to be provided by the
service providers can be of any value within this range based on the subsidy available
and the amount of fare recovered.

6. Conclusions
This paper has demonstrated a method that can be adopted by transit service providers in
a given city to take decisions on service level improvements. The approach used the
TRANSPORTATION PLANNING AND TECHNOLOGY 21

concepts of overall zone of tolerance, USLs, total utility and marginal utility. The research
discussed the notion that service providers must attempt to provide service levels within
the lowest boundary of the overall zone of tolerance for bus users to retain their existing
loyalty and at least within the lowest boundary of the overall zone of tolerance for non bus
users to attract potential users, which should be within LOS category C as perceived by
users. However, if resources are available, service providers may provide better service
levels than the aforesaid thresholds. However, once maximum marginal utility is
reached, it is not ideal for service providers to increase service levels because their rate
of return, that is the change in patronage, diminishes. In developing countries like
India, public transport fares are subsidized and regulated by the government to make
them affordable for lower income groups, who form the majority of transit service
users in urban areas. Under such circumstances, a judicious decision can be taken,
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

using the method described in this research, by the transit service providers in Indian
cities to make improvement plans, optimize costs but at the same time maximize patron-
age. However, in scenarios where no subsidy is required and transit users are willing to pay
for services, service providers may choose to provide service levels up to the desired service
level of the users and potential users.
Although this research has proposed a method that can be adopted by transit service
providers to determine a feasible range of transit service levels, this however, does not
provide an idea on which service areas should be prioritized for improvement. For this
purpose, it is recommended that an importance-satisfaction analysis be carried out by
transit service providers to identify the key service area gaps in an existing transit
system in a given city and make a priority plan according to the resources available.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
The authors would like to thank the Sponsored Research & Industrial Consultancy (SRIC) cell of
Indian Institute of Technology, Kharagpur, for funding this study.

References
ABS. 2008. “Australian Social Trends-Public Transport Use for Work and Study.” Australian
Bureau of Statistics. http://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/4102.0Chapter10102008.
Badami, M. G., and M. Haider. 2006. “An Analysis of Public Bus Transit Performance in Indian
Cities.” Transportation Research Part A: Policy and Practice 41A (10): 962–981.
Beirão, G., and J. A. S. Cabral. 2007. “Understanding Attitudes Towards Public Transport and
Private Car: A Qualitative Study.” Transport Policy 14 (6): 478–489.
Bhat, C. R., J. Guo, S. Sen, and L. Weston. 2005. Measuring Access to Public Transportation Services:
Review of Customer-Oriented Transit Performance Measures and Methods of Transit Submarket
Identification. Report 0–5178–1. Centre for Transportation Research, University of Texas, Austin
TX.
Bittinger, M. L. 1999. Calculus. 5th ed. Manchester: Addison-Wesley.
Choo, S., S. You, and H. Lee. 2013. “Exploring Characteristics of Airport Access Mode Choice:
A Case Study of Korea.” Transportation Planning and Technology 36 (4): 335–351.
22 S. DAS AND D. PANDIT

Cirillo, C., L. Eboli, and G. Mazzulla. 2011. “On the Asymmetric User Perception of Transit Service
Quality.” International Journal of Sustainable Transportation 5: 216–232.
Cullen, R. 2001. “Perspectives on User Satisfaction Surveys.” Library Trends 49 (4): 662–686.
Dantas, A. S., R. T. Antunes, Y. Yamashita, and M. V. Lamar. 2001. “Marketing in Public
Transportation: A Neural Network Approach.” 9th World Conference in Transportation
Research: D1–4137, Seoul, South Korea, July 22–27.
Das, S., and D. Pandit. 2011. “Level of Service Parameters for Bus Transit in the Indian context.” 1st
Conference of the Transportation Research Group of India. Bangalore, Transportation Research
Group of India, December 7–10.
Das, S., and D. Pandit. 2012. “Methodology to Identify the Gaps in the Level of Service Provided for
Urban Bus Transit: Case Study Kolkata.” SPANDREL 4 (Spring): 59–71.
Das, S., and D. Pandit. 2013a. “Importance of User Perception in Evaluating Level of Service for Bus
Transit for a Developing Country Like India: A Review.” Transport Reviews 33 (4): 402–420.
Das, S., and D. Pandit. 2013b. “Methodology to Determine Level of Service for Bus Transit in a
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Developing Country Like India.” 13th International Conference on Computers in Urban


Planning and Urban Management, Faculty of Geosciences, Utrecht University, The
Netherlands, July 2–5.
Das, S., and D. Pandit. 2015. “Determination of Level-of-Service Scale Values for Quantitative bus
Transit Service Attributes Based on User Perception.” Transportmetrica A 11 (1): 1–21.
dell’Olio, L., A. Ibeas, and P. Cecin. 2010. “Modelling User Perception of Bus Transit Quality.”
Transport Policy 17: 388–397.
dell’Olio, L., A. Ibeas, and P. Cecin. 2011. “The Quality of Service Desired by Public Transport
Users.” Transport Policy 18 (1): 217–227.
Gebeyehu, M., and S. Takano. 2008. “Modeling the Relationship Between Travelers’ Level of
Satisfaction and Their Mode Choice Behavior Using Ordinal Models.” Journal of the
Transportation Research Forum 47 (2): 103–118.
Gottschalk, P. G., and J. R. Dunn. 2005. “The Five-Parameter Logistic: A Characterization and
Comparison with the Four-Parameter Logistic.” Analytical Biochemistry 343: 54–65.
Gupta, S. 2012. “A Methodology for Assessing Service Quality of Infrastructure at Indian Railway
Stations.” Dissertation, Indian Institute of Technology, Kharagpur, India.
Kittelson & Associates, Inc., KFH Group, Inc., Parsons Brinckerhoff Quade, Douglass, Inc., K.
Zaworski. 2003. Transit Capacity and Quality of Service Manual (TCRP Report 100, 2nd ed.).
Washington, DC: Transportation Research Board.
Koppelman, F. S., and C. Bhat. 2006. A Self Instructing Course in Mode Choice Modeling:
Multinomial and Nested Logit Models. Report prepared for U.S. Department of
Transportation, Federal Transit Administration, Washington, DC.
Krizek, K. J., and A. El-Geneidy. 2007. “Segmenting Preferences and Habits of Transit Users and
Non-Users.” Journal of Public Transportation 10 (3): 71–94.
Leibowitz, N., B. Baum, G. Enden, and A. Karniel. 2010. “The Exponential Learning Equation as A
Function of Successful Trials Results in Sigmoid Performance.” Journal of Mathematical
Psychology 54: 338–340.
Maparu, T. S., and D. Pandit. 2010. “A Methodology for Selection of Bus Rapid Transit Corridors: A
Case Study of Kolkata.” Institute of Town Planners, India Journal 7 (4): 21–36.
Maparu, T. S., and D. Pandit. 2012. “Bus Route Merging and Frequency Setting Methodology for
Bus Rapid Transit Trunk Corridor: A Case Study of Kolkata.” Space: The SPA Journal of Planning
and Architecture 16 (2): 16–37.
Mfinanga, D. A., and M.-O.-A. Ochieng. 2006. “Development of a Model for Assessing Urban
Public Transport Level of Service in Cities of Developing Nations.” African Journal of Science
and Technology (AJST) 7: 35–53.
Mishalani, R. G., M. McCord, and J. Wirtz. 2006. “Passenger Wait Time Perceptions at Bus Stops:
Empirical Results and Impact on Evaluating Real-Time Bus Arrival Information.” Journal of
Public Transportation 9 (2): 89–106.
Mistretta, M., J. A. Goodwill, R. Gregg, and C. DeAnnuntis. 2009. Best Practices in Transit Service
Planning, Project Report, Florida Department of Transportation Research Center.
TRANSPORTATION PLANNING AND TECHNOLOGY 23

MoUD. 2008a. Service Level Benchmarks for Urban Transport, Ministry of Urban Development,
Government of India, New Delhi.
MoUD. 2008b. Study on Traffic and Transportation Policies and Strategies in Urban Areas in India,
Ministry of Urban Development, Government of India, New Delhi.
MRSCW. 1994. Level of Service Standards: Measures for Maintaining the Quality of Community
Life, Report 31, Municipal Research & Services Center of Washington, Kirkland, Washington,
DC.
Pandit, D., and S. Das. 2013. “A Framework for Determining Commuter Preference along a
Proposed Bus Rapid Transit Corridor.” Procedia Social and Behavioral Sciences 104: 894–903.
Pucher, J., N. Korattyswaroopam, and N. Ittyerah. 2004. “The Crisis of Public Transport in India:
Overwhelming Needs but Limited Resources.” Journal of Public Transportation 7 (4): 1–20.
Somaratna, S. D., C. N. Peiris, and C. Jayasundara. 2010. “User Expectation Verses User Perception
of Service Quality in University Libraries: A Case Study.” ICULA, Colombo. [online] Accessed
March 14, 2012. http://archive.cmb.ac.lk/research/bitstream/70130/168/1/ccj4.pdf.
Downloaded by [Orta Dogu Teknik Universitesi] at 04:01 13 January 2016

Taylor, W., and J. Brogan. 1978. Level-of-Service Concepts in Urban Public Transportation. Report
No. UM-HSRI-78-50. Michigan Transportation Research Program, Highway Safety Research
Institute, University of Michigan, Lansing, MI.
TCRP. 1999. A Handbook for Measuring Customer Satisfaction and Service Quality. Report 47,
Transit Cooperative Research Program, Transportation Research Board, Washington, DC.
Walker, J. 2008. “Purpose-Driven Public Transport: Creating a Clear Conversation about Public
Transport Goals.” Journal of Transport Geography 16: 436–442.
Wright, M., and D. Charlett. 1995. “New Product Diffusion Models in Marketing: An Assessment
of Two Approaches.” Marketing Bulletin 16: 32–41.
Yang, C. C. 2003. “Improvement Actions Based on the Customers’ Satisfaction Survey.” TQM and
Business Excellence 14 (8): 919–930.
Yen, J., C. Teng, and P. Chen. 2001. “Measuring the Level of Services at Airport Passenger
Terminals: Comparison of Perceived and Observed Time.” Transportation Research Record:
Journal of the Transportation Research Board 1744 (1): 17–33.
Zeithaml, V. A., L. Berry, and A. Parasuraman. 1993. “The Nature and Determinants of Customer
Expectations of Service.” Journal of the Academy of Marketing Science 21 (1): 1–12.

You might also like