ARTICLE IN PRESS

Transport Policy 15 (2008) 260–272

Contents lists available at ScienceDirect

Transport Policy
journal homepage: www.elsevier.com/locate/tranpol

Public transit user satisfaction: Variability and policy implications

Yannis Tyrinopoulos a,, Constantinos Antoniou b,1
a
Hellenic Institute of Transport (HIT), Charilaou-Thermi Road, 570 01 Thermi, Thessaloniki, Greece
b
National Technical University of Athens (NTUA), 9 Heroon Polytechniou, 15773 Zografou, Athens, Greece

a r t i c l e in f o a b s t r a c t

Available online 26 July 2008 This research focuses on passenger’s perception of transit performance with an emphasis on the
Keywords: variability between operators and the policy implications of such differences. Two statistical methods
Public transit (factor analysis and ordered logit modeling) have been used to assess the quality implications of the
Users satisfaction variability of the users’ perceived satisfaction across operators. A market segmentation analysis
Policy implications of user satisfaction (between male and female respondents) provides further insight into the differences among groups of
variability the population. Five transit systems in the two major conurbations in Greece, Athens and Thessaloniki,
have been examined. The analysis demonstrated that a well-coordinated transportation environment
should be the primary aim of the policy makers in Athens, followed by other quality attributes such as
service frequency and accessibility. In Thessaloniki, the sole transit operator should include in its policy
plans immediate corrective measures addressing the service frequency, waiting time and vehicle
cleanliness attributes.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction time among different routes or even among different public

transport operators (cross-sectional analysis).
Assessing and improving quality of service in public transit
holds a high priority for the majority of public transport operators. The above approaches have been addressed at different levels
Now more than ever, public transport operators are forced to place of significance in various countries primarily in the European
particular emphasis on the monitoring and improvement of the Union (EU) and North America. The TRB Transit Capacity and
services provided in an attempt to address the increasing rate of Quality of Service Manual (TRB, 2004), the TRB Handbook for
car ownership and the deterioration of traffic conditions, and Measuring Customer Satisfaction and Service Quality (TRB, 1999),
ultimately to contribute to a sustainable urban mobility. and the EC-CEN Transportation—Logistics and Services—Public
Quality of service in public transit reflects the passengers’ Passenger Transport—Service Quality Definition, Targeting and
perception of transit performance. There are a number of Measurement (CEN, 2002) are some of the manuals produced for
approaches and techniques used to define and assess quality of measuring public transit quality. The overall process to improve
service, such as this quality entails the identification of customers’ priorities
and needs, the measurement of customers’ satisfaction using
 Customer satisfaction in public transit can be defined as the appropriate indices, the use of this feedback to evaluate the
overall level of attainment of a customer’s expectations, relevant service parameters and finally the definition and
measured as the percentage of the customer expectations, implementation of measures to improve the services provided to
which have actually been fulfilled. the customers.
 Customer loyalty is reflected by a combination of attitudes and Additional efforts have been made by various research bodies
behavior. It is usually driven by customer satisfaction. Yet, it invol- in the area of public transit quality management. Some of the
ves a commitment on the part of the customer to make a sus- most important ones are The Demand for Public Transport:
tained investment in an ongoing relationship with transit service. A Practical Guide (TRL, 2004), which aims to provide practical
 Benchmarks can also be used for comparing quality either in guidance on demand estimation for those involved in planning
different time frames (time-series analysis), or at the same and operating public transport services, and the various EU
research and development (R&D) efforts, such as PORTAL (2003),
 Corresponding author. Tel.: +30 2310 498267/498263; fax: +30 2310 498269. EQUIP (2000) and QUATTRO (1998), which addressed benchmark-
E-mail addresses: ytyrin@certh.gr (Y. Tyrinopoulos), antoniou@central.ntua.gr
ing and quality management elements in public transport. All
(C. Antoniou). these initiatives and efforts tackled different quality aspects of
1
Tel.: +30 210 7721380; fax: +30 210 7721454. public transport. For example, the project QUATTRO developed the

0967-070X/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tranpol.2008.06.002
 ARTICLE IN PRESS

Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272 261

‘‘quality loop of public transport’’ based on which the various This research has three primary objectives, in particular to
quality levels can be better coordinated and the provided service outline a methodology for analyzing the variability of the users’
optimized. behavior and their level of satisfaction from the use of diverse
In Greece, the Hellenic Institute of Transport promotes quality- transit systems, to present the results drawn from the application
control programs for public transport operators. The primary aim of this methodology to five different transit systems in Greece and
of these programs was to assess the quality and performance of to develop recommendations for public transport operators and
the transit systems in these cities using a variety of performance policy makers that will enable them to better understand the
and quality indicators, such as on-time performance, average behavior of the users and the key factors affecting their choices in
waiting time at the terminals and stops, vehicle load, average transit ridership. This feedback can also be used by public
route speed, conditions at the terminals and stops, safety, transport operators to adjust their policy plans accordingly and
information provision, accessibility and many others. Additional define actions that can better tackle the customers’ needs and
goals of these programs included: expectations.

 continuous monitoring of the quality levels of the services

provided; 2. The examined transit systems
 definition and implementation of immediate actions for the
improvement of the transit services; The analysis of the variability of users’ behavior and level of
 identification of the priorities and needs of the passengers; satisfaction from public transit services took place in five transit
 effective communication and cooperation among the parties systems in the two major cities in Greece, Athens and Thessalo-
involved; and niki. The five transport companies operate three different means
 support of the decision and policy-making processes. of transport (bus, trolley bus and rail/metro). The main characte-
ristics of these five transit systems are outlined below.
The above quality-control programs were applied to five
different public transit systems in Athens and Thessaloniki, 2.1. Attiko Metro Operation Company (AMEL)
presented in the next section. In the part of the programs referring
to the customers’ satisfaction, the passengers were asked to assess AMEL is a subsidiary of Attiko Metro S.A., with main objectives
certain attributes of these five systems both from the point of to organize, manage, operate and develop the underground
view of their importance and their level of satisfaction about the railway network on lines 2 and 3 (Fig. 1), and any extension
current transportation services. thereof within the prefecture of Attica, as well as their facilities,

Fig. 1. Map of Attika, showing fixed route networks (AMEL and ISAP) (Source: http://ametro.gr).
 ARTICLE IN PRESS

262 Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

vehicles, materials and media. It cooperates with other transpor- 2.5. Organization of urban transport of Thessaloniki (OASTH)
tation bodies and organizations of the prefecture of Attica and is
responsible for the operation of the Athens metro. The two metro OASTH serves the Greater Thessaloniki Area (GTA). GTA
lines operated by AMEL are integrated with the electric railway consists of the City of Thessaloniki and 17 nearby municipalities.
line (line 1) operated by ISAP (see Section 2.4). The public transport operation is regulated and monitored by the
Today, 23 metro stations are in operation at both METRO lines. Thessaloniki Transport Authority, under the jurisdiction of the
These lines serve 650,000 passengers per day. The frequency of Ministry of Transport. OASTH moves on a daily basis 505 out of
trips is every 3 min in rush hours and 5–10 min in non-rush hours the 536 available buses, servicing 150,000,000 passengers per
(HIT, 2003). year via 68 bus lines (HIT, 2005).

2.2. Company of Thermal Buses in Athens (ETHEL) 3. Key elements of the quality-control programs

ETHEL provides urban transport services with thermal buses in 3.1. Main activities of the two programs
the metropolitan area of Athens. The bus line network includes
the following main lines: The two quality-control programs applied to the two major
conurbations in Greece, Athens (HIT, 2003) and Thessaloniki
 Forty core lines that connect the Athens and Piraeus city (HIT, 2005), had similar aims, characteristics and methodological
centers with the centers of the peripheral municipalities, approaches. Their goals were to investigate the performance and
 Twenty inter-municipal lines that connect the municipalities quality of the current transportation services, to identify and
of the Attica region without crossing the Athens and Piraeus analyze the weak points and barriers preventing the effective
city centers, transit operation, as well as to formulate a complete quality
 One hundred and twenty-three local lines that operate within control system, based on which the relevant authorities are able to
the limits of one or a group of neighboring municipalities and monitor the quality levels of the transit services and take
act as feeders to the core lines, measures that will improve the overall transportation scene. The
 Nineteen express lines, and programs contained the following key activities:
 Seven school-bus lines.
 Initial identification and analysis of the main characteristics
and attributes of the local transportation environments;
The company serves 310 total bus routes and operates 16,000  Development of the methodologies to be applied, composed
trips daily, which represent 98.6% of all scheduled trips. This of the operational and quality indicators, measurement
percentage is considered very high, given the constant deteriora- techniques, samples, timetables, as well as templates and
tion of traffic conditions, as well as the continued decrease of the questionnaires for data collection;
buses’ average speed. ETHEL owns and operates a fleet of 2099  On site surveys for capturing the performance of the transit
buses. Currently, there are 1822 buses in operation during peak systems (operational indicators);
hours (HIT, 2003).  Customer satisfaction/dissatisfaction surveys (qualitative
indicators);
 Development of relational databases for data archiving and
2.3. Athens–Piraeus trolley busses (ILPAP) indicators calculation;
 Analysis of the indicators’ outcomes and reporting; and
ILPAP operates an electric bus network (also called ‘‘trolley  Development of a software for monitoring the performance
buses’’) of 22 lines that serve primarily the Athens and Piraeus city and quality of the transit services in the future.
centers. Ten of these lines are being monitored by a telematic
system. The company owns and operates a fleet of 315 single
3.2. Considered public transit quality attributes
trolley buses (12 m long) and 51 articulated trolley buses (18 m
long). All trolley buses are air-conditioned (142 trolley buses are
also equipped with heating system) and they are also self- The customer satisfaction/dissatisfaction surveys conducted in
powered (by diesel engine as well as electric generator), for the framework of the two programs involved a number of
continuity in cases of power failure or physical obstacles, forcing qualitative and operational service attributes (parameters). The
the vehicles to temporarily move away of their electric power list of attributes finally used was based to a large extent on the
lines. Handbook for Measuring Customer Satisfaction and Service
The number of operated trolley bus trips per day is 1943 Quality (TRB, 1999), following some adjustment to reflect existing
(first semester 2005 average). The total number of trolley bus conditions in the areas, the particular characteristics of the local
passenger trips in 2004 was 77 million. transportation systems and the key priorities of the transport
operators. The 23 selected attributes were classified into four
categories.
2.4. Athens–Piraeus electric railways (ISAP)
3.2.1. General characteristics of the public transit system
The company operates the electric railway line that runs
between Piraeus and Kifissia (metro line 1, Fig. 1), serving 24 1. Service frequency: refers to the frequency of the service in the
stations. The total length of line 1 is 25.6 km, while the total lines of the transit systems.
journey time (in one direction) is 51 min. ISAP operates 607 2. On-time performance: refers to the accuracy of the departure
trips daily (HIT, 2003). The maximum speed of the ISAP trains is times of the vehicles at terminal stations in relation to the
70 km/h. The total daily number of ISAP passenger trips is predefined schedule.
currently 450,000. ISAP has currently 84 trains, which amount 3. Service provision hours: refers to the operating hours of the
to 363 wagons. service provision on a given day.
 ARTICLE IN PRESS

Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272 263

4. Network coverage: refers to the spatial coverage of the area 22. Waiting time at transfer points: refers to the time that
under consideration with public transit services. passengers have to wait at transfer points in order to continue
5. General information provision: refers to the sufficiency of the their trip.
information provided to the passengers about the general 23. Information provision at transfer points: refers to the provision
characteristics of the transit services, such as the lines, of information to passengers at the transfer points about the
terminals and stops points, departure times, tickets and combination of the various lines and modes, and their time
passes available. schedules.
6. Types of tickets and passes: refers to the sufficiency of the
various available types of tickets and passes with respect to 3.3. Sample determination
the coverage of the needs of the public.
7. Prices of tickets and passes: refers to the price-structure of the
The sampling procedure applied to the two quality-control
various types of tickets and passes available.
programs was carefully designed, taking into account the spatial
8. Tickets selling network: refers to the sufficiency of the tickets
distribution of the population and the minimum sample per
selling network and the ease to purchase tickets from the
geographic area. Furthermore, the minimum sample sizes were
various selling points.
defined based on the following statistical method (Johnson and
9. Personnel behavior: refers to the behavior of the various types
Wichern, 1992):
of personnel of the transport operator (e.g., drivers, station
(  2 )1
officers and ticket counter officers), when communicating and N1 d
transacting with the passengers. nXN 1 þ
Pð1  PÞ za=2
10. Existence of bus lanes: refers to the sufficiency and perfor-
mance of the bus lanes to facilitate the efficiency of the transit where N is the size of the population that in this case is the
service. passenger traffic of the transit systems; P is the quality
11. Measures for environmentally friendly public transit: refers to characteristic to be measured (satisfaction); if no previous
the contribution of public transit in the protection of the experience exists then the neutral situation (P ¼ 0.5) is consid-
environment and the adequacy of the relevant actions and ered; d is the margin of error (5%); za/2 ¼ 1.64 for level of
measures taken by the relevant authorities. confidence 90%.
The samples sizes finally chosen for the five transit systems
3.2.2. Terminals and stops and used in the two quality-control programs are

12. Walking distance to terminals and stops: refers to the distance  202 passengers for AMEL,
that passengers have to walk from the origin point to the  556 passengers for ETHEL,
closest terminal and stop.  177 passengers for ILPAP,
13. Information provision at terminals and stops: refers to the  165 passengers for ISAP, and
sufficiency of the information available to the passengers  374 passengers for OASTH.
about the services provided at the terminals and stops.
14. Conditions at terminals and stops: refers to the conditions of
During the surveys conducted in the context of the two quality-
the terminals and stops concerning shelter, visibility, seating
control programs, no special events or incidents occurred that
capacity, etc.
affected the data quality or reliability.
15. Safety at terminals and stops: refers to the perceived sense of
safety of the passengers when waiting at the terminals and
stops to use the public transit service.
4. Methodology

3.2.3. Vehicles
Two statistical methods have been used for the analysis of the
survey responses. The output of these statistical methods is then
16. Onboard conditions: refers to the conditions inside the vehicle
interpreted in order to assess the quality implications of the
during the execution of a journey, mainly concerning crowded
variability of the users’ perceived satisfaction with respect to the
situations and the provision/condition of available facilities
public transit systems.
(e.g., seats and air-conditioning).
17. Vehicles cleanliness: refers to the level of cleanliness of the
vehicles from various standpoints (seats, handles, windows, 4.1. Factor analysis
doors, floor, etc.).
18. Driving behavior: refers to the driving performance of the The objective of factor analysis is to reduce the number of p
vehicle’s driver. variables in a dataset into a smaller set of Kop variables. The K
19. Onboard information provision: refers to the provision of factors are usually unobservable factors that describe the
information inside the vehicle during the trip, such as next correlation among the p variables. Factor analysis is closely
stop and estimated arrival time at the next stop. related to principal components analysis (e.g., both rely on the
20. Accessibility to disabled and mobility impaired people: refers to correlation matrix), but, unlike principal components analysis, it
the provision of facilities by the transit operator to facilitate is based on a specific statistical model (Washington et al., 2003).
the accessibility of transit services by disabled and mobility- Factor analysis was developed in the early 20th century by Karl
impaired people. Pearson and Charles Spearman with the intent to gain insight into
psychometric measurements, in particular the directly unobser-
3.2.4. Transfer points vable variable intelligence (Johnson and Wichern, 1992). Factor
analysis should not be blindly applied to a dataset with several
21. Distance between transfer points: refers to the distance that variables hoping that some underlying patterns would be
passengers have to walk between transfer points in order to uncovered; instead, a theoretical motivation should drive factor
continue their trip. analysis applications.
 ARTICLE IN PRESS

264 Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

In the remainder of this section, the presentation and 5. Analysis

notation from Washington et al. (2003) is used. The factor analysis
model can be formulated by expressing the Xi terms as linear Survey respondents were asked to rate each attribute accord-
functions, such that ing to two criteria: how important the particular attribute is to
them with respect to public transit service (importance) and how
X 1  m1 ¼ ‘11 F 1 þ ‘12 F 2 þ    þ ‘1m F m þ 1 they would rank the public transit service in question with respect
X 2  m2 ¼ ‘21 F 1 þ ‘22 F 2 þ    þ ‘2m F m þ 2 to that attribute (satisfaction). The data collected from the first set
.. of questions were used as input for the factor analysis, while the
. second set of data was used as input for the ordered logit models.
X p  mp ¼ ‘p1 F 1 þ ‘p2 F 2 þ    þ ‘pm F m þ p All models have been estimated using the R framework for
statistical computing, v. 2.7.0 (R Development Core Team, 2008).
where F is the factors and ‘ij is the factor loadings. ei is associated
only with the Xi; and p, random errors and m, factor loadings are 5.1. Factor analysis
unobservable or latent.
In matrix notation, the factor analysis model is given as Factor analysis has been performed on the collected data of the
customer satisfaction/dissatisfaction surveys on the importance of
ðX  mÞp1 ¼ Lpm Fm1 þ p1
each quality attribute according to the users. Three prevalent
The above systems have p equations and p+m unknowns, which factors are considered for each system and the obtained factor
implies that a unique solution cannot be obtained without loadings were rotated to obtain a practical and easily interpretable
additional information. Additional restrictions, in the form of set across the various systems. Each factor explains at least 10% of
factor rotation models, are typically imposed, resulting in the variance for the respective model (with the exception of the
orthogonal or oblique factor analysis models. Varimax rotation third factor for ETHEL that explains 9.4%). The results are
maximizes the sum of the variances of the factor loadings and is summarized in Table 1. Loadings below 0.3 are not shown.
commonly used for orthogonal rotation. Oblique factor analysis Furthermore, loadings above 0.6 are highlighted in bold.
models relax the restriction of uncorrelated (orthogonal) factor The objective of factor analysis is to try to discern and
loadings, striving to obtain a more easily interpretable factor recognize the underlying unobserved factors that the respondents
structure. perceive. Such an interpretation of the three main factors is
The objective of the various rotations is to move each factor as attempted in the last line of Table 1. Each of the factors has been
close to 1 or 0 as possible. Large factor loadings indicate a interpreted as
significant influence of the factor to the respective variable, while
small factor loadings suggest that the variable is not substantially  Quality of service: comprising attributes related to price,
influenced by that factor. Therefore, extreme loading values are information provision, behavior of personnel, waiting and in-
more easily interpretable, and therefore more desirable. vehicle conditions and accessibility,
 Transfer quality: comprising attributes related to trans-
fer coordination (distance, waiting time and information
4.2. Ordered logit model provision),
 Service production: reflecting service frequency and reliability,
Respondents in surveys are often asked to express their or
preferences in a rating scale. For example, respondents may be  Information/courtesy: comprising information on information
asked how satisfied they are with a public transit service in a scale provision and behavior of personnel.
of very satisfied/satisfied/somewhat dissatisfied/very dissatisfied
(TRB, 1998). Alternatively, a ranked scale of 1–4 can be used. Such Four out of the five operators are characterized by the same
scales are often called Likert scales (Likert, 1932). The main factors, albeit not always in the same order. AMEL, the only
characteristic of these questions, from a modeling perspective, is operator with somewhat different factor interpretation, has a
that the potential responses are ordered. A rating of very satisfied somewhat different factor structure. This is intuitive and
is therefore ‘‘higher’’ than satisfied (indicating, e.g., a higher representative of the reality, since Attiko metro is considered as
satisfaction towards the considered system), which is in turn a higher quality, modern metro system, with superior operating
‘‘higher’’ than somewhat dissatisfied. characteristics.
A multinomial logit model could be specified with each A convenient way to visually represent groups of three factors
potential response coded as an alternative. However, the ordering is triangle/ternary plots. In these plots, the proportions of the
of the alternatives violates the independence of the errors for each three variables plotted always sum to some constant (usually 1 or
alternative, and therefore the independence for irrelevant alter- 100). The projection of each point on each side of the triangle
natives (IIA) assumption of the logit model (Ben-Akiva and (note that the projection of points to an edge is taken parallel to
Lerman, 1985). With ordered alternatives, one alternative is the next clock-wise edge of the triangle) indicates the percentage
more similar to those close to it and less similar to more distant that is attributed to this factor. For example, in Fig. 2, the
alternatives. Nested or cross-nested models are one approach projection of ILPAP on the quality of service edge of the triangle
to overcoming this issue. Ordered logit models provide falls at around 0.4, indicating that this factor accounts for roughly
another approach that estimates parameter coefficients for the 40% of all three factors for that operator. A smaller triangle (placed
independent variables, as well as intercepts (or threshold values) to the top and left corner of each figure) indicates where the
between the choices. Assuming a ranking scale with four presented triangle falls within a general triangle ranging from 0
levels, there are three thresholds or critical values that separate to 1 across all dimensions.
the choices (Venables and Ripley, 2002). For example, respondents Fig. 2 presents a triangle plot for the four operators that share a
choose alternative very satisfied if the utility is above k3|4, common set of factors, i.e., ETHEL, ILPAP, ISAP and OASTH. This
and alternative satisfied if the probability is between k2|3 figure indicates that OASTH and ILPAP have a similar factor
and k3|4, etc. structure, dominated by quality of service (around 50%), followed
Table 1
Factor analysis results by operator (all respondents)

Loadings OASTH AMEL ILPAP ISAP ETHEL

Factor1 Factor2 Factor3 Factor1 Factor2 Factor3 Factor1 Factor2 Factor3 Factor1 Factor2 Factor3 Factor1 Factor2 Factor3

Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

Service frequency 0.864 0.921 0.877 0.965 0.763
On-time performance 0.684 0.875 0.829 0.833 0.739
Service hours 0.350 0.447 0.330 0.370
Timetable information 0.614 0.361 0.602 0.740 0.608

ARTICLE IN PRESS
Price 0.467 0.567 0.375 0.444
Behavior of personnel 0.491 0.542 0.854 0.677 0.783
(excluding driver)
Existence of bus lanes 0.303 0.469 0.508 0.332
Distance/time to access 0.325 0.412 0.426 0.419 0.354
stop
Timetable information at 0.864 0.611 0.638 0.471 0.569
stop
Waiting conditions at stop 0.553 0.724 0.739 0.314 0.422 0.468
Condition in-vehicle 0.533 0.990 0.661 0.594 0.578 0.380
Driver behavior 0.472 0.426 0.667 0.799 0.771
Information in-vehicle 0.828 1.065 0.574 0.880 0.454
Accessibility (w.r.t. 0.410 0.351 0.519 0.368 0.460
disabilities)
Transfer distance 0.948 0.575 0.744 0.808 0.920
Transfer waiting time 0.733 0.599 0.868 0.769 0.822
Information regarding 0.504 0.476 0.524 0.815 0.809 0.553
transfers

Sum of square of loadings 3.504 1.863 1.778 2.861 2.336 2.630 4.236 2.418 1.805 3.373 2.794 2.186 3.357 2.298 1.597

Proportion variance 0.206 0.110 0.105 0.168 0.137 0.155 0.249 0.142 0.106 0.198 0.164 0.129 0.197 0.135 0.094

Factor interpretation Quality of Transfer Service Service production/ Quality of Information/ Quality of Transfer Service Quality of Service Transfer Quality of Transfer Service
service quality production transfer quality service courtesy service quality production service production quality service quality production

265
 ARTICLE IN PRESS

266 Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

Fig. 3. Triangle plot for all respondents (AMEL).

Fig. 2. Triangle plots for all respondents (ETHEL, ILPAP, ISAP and OASTH).

by transfer quality (at o30%) and service production (at 420%). satisfied and very satisfied, coded as 1–4 in this order. As a result,
The factor that is perceived as the most important for ISAP is again the estimated coefficient of each attribute should be positive.
quality of service (at about 40%), followed by service production Furthermore, as the coefficients are in the same scale (since each
(at about 34%) and transfer quality (at about 26%). Customer variable takes the same values, i.e., 1–5), the unit-less coefficients
satisfaction for ETHEL is almost equally attributed to quality of are directly comparable.
service and transfers (each at about 37%), followed by service The results of the ordered logit model are presented in Table 2.
production (at about 26%). ETHEL is a bus operator, which suffers Across transit operators, the most important satisfaction
from the interaction with the general traffic in the notoriously attributes are service frequency, followed by vehicle cleanliness,
very congested traffic environment of Athens. Furthermore, ETHEL waiting conditions, transfer distance and network coverage. The
bus lines cover the majority of the geographical area in the Athens most important satisfaction attributes for AMEL are vehicle
greater region and it always interacts with the other transit cleanliness, followed by behavior of staff (other than the driver)
systems (transfers). Under these circumstances, it makes sense and ticketing systems. This reflects the perception of users that
that the factors that the passengers consider most important for the metro is a very reliable system, in which service frequency and
ETHEL are the quality of service and transfers. trip time are implicitly guaranteed (or at least perceived as such).
The quality of service is the most important factor for four out Therefore, the users notice and value these softer attributes.
of the five surveyed transport operators’ customers. In the one This result is also in line with the conclusions of the
system that it was not deemed most important (AMEL), quality of European Performance Satisfaction Index (EPSI) Rating Institute
service ranked third, but clearly still important (at 30%). One (EPSI, 2005), according to which AMEL has the highest customer
reason is that people implicitly associate a high quality of service satisfaction index compared with other means of public transport
with the particular metro system (as a result of an aggressive across Europe.
marketing/promotional campaign when the operation started, For ETHEL, on the other hand, the most important satis-
followed by a well-operating system), and therefore may take it faction attributes are service frequency, vehicle cleanliness and
for granted. Fig. 3 summarizes the percent contribution of each of network coverage, closely followed by waiting conditions and
the identified factors in customer satisfaction perception towards punctuality (service adherence). This is again consistent with the
AMEL. Service production/transfer quality and information/ nature of the operator, who has to cover a very large conurbation
courtesy of the personnel are the dominant factors for AMEL with a finite amount of resources, providing a satisfactory level of
with 36% and 34% of the satisfaction perception, respectively. service. The passengers realize that ETHEL provides service to
areas where other public transit systems cannot reach and thus
5.2. Ordered logit models the service frequency is indeed satisfactory. This finding is
noteworthy, considering that service frequency sometimes suffers,
The satisfaction data of the surveyed individuals were used to either due to the scheduled service limitations, or due to
fit ordered logit models for each operator. A full model with all unforeseen external factors, such as congestion, vehicles breaking
available attributes as explanatory variables was estimated first. down, demonstrations or road maintenance blocking roads or
The model was then gradually simplified by removing attributes strikes.
that did not contribute to the model. In general, variables whose Service frequency, transfer distance, ticketing systems and
inclusion contributed less than 1/2 AIC (Akaike Information vehicle cleanliness are the key satisfaction indices for ILPAP. ILPAP
Criterion, Akaike, 1974) point were removed. Furthermore, services a small number of lines and has recently acquired new
additional variables were removed in a couple of instances that vehicles that replaced some very antiquated equipment. Further-
counterintuitive model coefficients were obtained (in terms more, the areas in which ILPAP operates are well serviced by other
of sign). modes as well (e.g., AMEL and ETHEL) and therefore stations are
The satisfaction variables were obtained in a range of 1 (most close and the distance that needs to be covered in order to transfer
dissatisfied) through 5 (most satisfied). The overall satisfaction of to/from another operator is limited.
the operator was collected as a four-valued categorical variable ISAP users value punctuality first, followed by network
taking values from very dissatisfied, somewhat dissatisfied, coverage, in-vehicle service conditions and passenger safety.
Table 2
Ordered logit model results (all respondents)

Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

All operators AMEL ETHEL ILPAP ISAP OASTH

Value S.E. t-Test Value S.E. t-Test Value S.E. t-Test Value S.E. t-Test Value S.E. t-Test Value S.E. t-Test

ARTICLE IN PRESS
Punctuality 0.189 0.068 2.793 0.247 0.118 2.097 0.847 0.192 4.407
Passenger safety 0.415 0.175 2.370
Transfer distance 0.237 0.060 3.929 0.208 0.099 2.096 0.629 0.212 2.963
Network coverage 0.233 0.053 4.412 0.282 0.087 3.229 0.487 0.156 3.119 0.245 0.106 2.304
Vehicle cleanliness 0.288 0.060 4.765 0.660 0.152 4.338 0.361 0.105 3.450 0.462 0.171 2.698 0.277 0.111 2.493
Driver behavior 0.161 0.066 2.451 0.181 0.096 1.877
Waiting conditions 0.260 0.060 4.367 0.270 0.101 2.681
In-vehicle service conditions 0.488 0.198 2.463
Ticketing systems 0.141 0.051 2.738 0.297 0.140 2.118 0.499 0.165 3.020
Behavior of non-driver staff 0.143 0.059 2.416 0.404 0.145 2.781 0.227 0.105 2.161
Service frequency 0.416 0.072 5.744 0.380 0.126 3.005 0.669 0.194 3.452 0.539 0.114 4.745
Waiting time 0.435 0.116 3.746
Price 0.307 0.099 3.107
Information about schedule 0.267 0.106 2.522

Intercepts
1|2 10.48 0.40 26.00 8.24 0.85 9.69 10.37 0.66 15.61 12.15 1.32 9.22 10.98 1.16 9.43 10.84 0.80 13.62
2|3 7.04 0.33 21.56 4.66 0.66 7.12 6.69 0.51 13.04 8.50 1.02 8.37 7.36 0.93 7.93 7.32 0.63 11.53
3|4 3.32 0.27 12.17 1.54 0.58 2.67 2.64 0.41 6.46 4.20 0.77 5.42 3.30 0.70 4.74 3.78 0.54 7.04

Residual deviance 2457.2 364.6 868.0 263.1 263.5 620.1

AIC 2481.2 376.6 888.0 277.1 277.5 640.1

267
 ARTICLE IN PRESS

268 Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

ISAP is a rail system that offers rather frequent service and

provides significant transportation service across a single line
(metro line 1). The vehicles are spacious, albeit not as modern as
those operated by AMEL. A higher perception of passenger safety
is obtained by the fact that ISAP personnel (ticket issuer) is
present in all stations, which are also not as exposed as bus stops
in the kerbside.
Finally, OASTH users rank service frequency, waiting time,
vehicle cleanliness and schedule information higher. It is useful to
highlight that unlike the other four systems operating in the
Attika region, OASTH is the sole public transit service operating in
Thessaloniki. Survey respondents are particularly satisfied with
the service frequency and the waiting time, which indicates that
OASTH provides a service that covers the needs of the respon-
dents. Respondents from Thessaloniki were the only ones who
indicated that they are content with the price charged for the
provision of public transit. This may be attributed again to the fact
that there is no alternative for them. (Incidentally, the construc-
tion of a new metro line in Thessaloniki started in 2006.)

6. Market segmentation Fig. 5. Triangle plot for female respondents (ISAP, ETHEL).

The analysis presented in the previous section provides two

alternate approaches to quantify the public perception of the
operation of public transit services. Additional interesting insight
can be obtained through the analysis of different market segments
based on age or sex. Given the size and composition of the
available sample, market segmentation by sex (i.e., male versus
female respondents) was pursued in this paper.

6.1. Factor analysis

The factor analysis results are summarized in the triangle plots

presented in Figs. 4–7. The first two plots reflect the perceptions of
female respondents, while the latter two plots reflect those of
male respondents. Female respondents ranked quality of service
as the most important factor for OASTH (42%) followed by transfer
quality (32%). Service production was the most important factor
for female respondents (47%), while quality of service (52%) and
transfer quality (48%) have very high impact for AMEL. Note that,
as the preferences of female respondents with respect to AMEL
were captured satisfactorily by these two factors, the projection of
AMEL on the ‘‘service production’’ side of the triangle plot is zero.

Fig. 6. Triangle plot for male respondents (OASTH, ILPAP, ETHEL).

In terms of ISAP, service production and courtesy were tied in the

preference of female respondents (35% each), while for ETHEL the
most important factor was courtesy (45%).
Male respondents perceptions for OASTH, ETHEL and ILPAP are
captured by two factors, namely quality of service and transfer
quality. Therefore, the third side of the triangle plot in Fig. 6 is not
named (however, the concept of a triangle is retained for
consistency with the rest of the analysis). The preference of the
respondents with respect to OASTH and ETHEL is dominated by
quality of service (59% and 53%, respectively) with the rest going
to transfer quality. ILPAP showed a 63% loading on transfer quality,
with the rest reflecting quality of service. AMEL shows a rather
balanced loading on service quality (39%), service production
(33%) and information (28%), while the male respondents
perceptions relating to ISAP are split between information (53%)
Fig. 4. Triangle plot for female respondents (AMEL, OASTH, ILPAP). and service quality (47%).
 ARTICLE IN PRESS

Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272 269

operating three different modes of transport (bus, trolley bus and

rail/metro) in the two major cities in Greece, Athens and
Thessaloniki, were examined.
Factor analysis discerned the underlying unobserved
factors that the respondents perceive. For the transit companies
operating bus and trolley bus services, quality of service and
transfer quality appear to hold a top priority for the customers.
These are associated with quality attributes comprised prices,
information provision, waiting and in-vehicle conditions, accessi-
bility and transfer coordination. On the other hand, for the only
transit company operating a metro system, quality of service
ranked third. There, the high quality services are taken for granted
and thus customers emphasize on other quality attributes, such as
the transfer coordination with other means and information
provision. Overall and as expected, the customers expressed their
strong preference to a well coordinated and reliable transporta-
tion environment, since currently residents in the Attica region
are serviced by many transport operators (currently six in total, as
a tram service and a suburban railway service are in operation
after the data were collected.). In Thessaloniki, the situation is
different, since the passengers have no alternative to bus transit
Fig. 7. Triangle plot for male respondents (ISAP, AMEL).
mean.
According to the ordered logit models analysis, conducted
based on the satisfaction data of the surveyed individuals,
the most important satisfaction attributes across transit
6.2. Ordered logit models operators are service frequency, vehicle cleanliness, waiting
conditions, transfer distance and network coverage. The results
The ordered logit analysis has been performed for the two of the ordered logit models, however, differ between the
segments of the population (female, shown in Table 3 and male, transit systems, as their particular characteristics and service
shown in Table 4). Female respondents overall rank waiting conditions play a significant role in their performance and
conditions and punctuality higher. For AMEL (where for the most the satisfaction of the customers. For the metro operator, for
part service production is more reliable), female respondents rank example, the most important satisfaction attributes are vehicle
in-vehicle conditions and driver behavior higher. ETHEL female cleanliness, followed by behavior of staff (other than the driver)
users rank punctuality and service frequency higher, while for and ticketing systems. This reflects the perception of users
ILPAP female users, service frequency and information about that the metro is a very reliable system with maximum adherence
transfers matter the most. ISAP female users rank punctuality very to schedule. For the bus and trolley bus operators, on the
high, followed by in-vehicle service conditions, while OASTH users other hand, the most important satisfaction attributes are
rank punctuality and waiting conditions higher. It is interesting to service frequency, vehicle cleanliness and network coverage,
note that both bus services (ETHEL and OASTH) seem to have closely followed by waiting conditions and punctuality (service
similar characteristics in this respect. adherence). The congested traffic conditions in Athens and
Male respondents overall value service frequency and vehicle Thessaloniki, as well as other external factors, such as strikes
cleanliness more highly. Behavior of non-driver staff and vehicle and road constructions, explain the findings of the analysis for
cleanliness are the highest ranked attributes for AMEL (like with these operators.
the female respondents, AMEL characteristics are not directly
related to the production of service). ETHEL male users rank 7.2. Contribution to policy making
vehicle cleanliness and service frequency higher, while ILPAP male
users show a high ranking for service frequency, ticketing systems, Factor analysis and ordered logit modeling are two appropriate
behavior of non-driver staff and in-vehicle service conditions. statistical methods that can be used by policy and decision
ISAP male users rank vehicle cleanliness and punctuality higher, makers to improve the transit services provided to the passengers.
while OASTH male users rank service frequency and in-vehicle Based on the results and findings derived from the application of
service conditions higher. these methods, public transport operators, authorities and policy
makers may integrate in their strategic plans corrective actions
and measures that can better tackle users’ perception and thus
7. Conclusion increase the number of transit passengers.
The research presented in this paper demonstrates that a well-
7.1. Key concluding remarks coordinated and well-structured transportation environment
should be the primary aim of the policy makers in Athens,
The analysis of the variability of users’ behavior and their level followed by other quality attributes such as service frequency and
of satisfaction from the use of transit systems may provide useful accessibility. In Thessaloniki, the sole transit operator should
insight about the strengths and weaknesses of a given transit include in its policy plans immediate corrective measures placing
service, and recognize qualitative attributes that need special more emphasis on the service frequency, waiting time and vehicle
attention. This research focused on the passenger’s perception of cleanliness.
transit performance and by employing two statistical methods In conclusion, this research demonstrates how factor analysis
(factor analysis and ordered logit modeling) assessed the quality and ordered logit modeling may be used to analyze the variability
implications of the variability of the users’ perceived satisfaction of transit customers’ satisfaction and how the results and findings
with respect to the public transit systems. Five-transit systems of this process may contribute to policy and decision making. The
 270
Table 3
Ordered logit model results (female respondents)

All operators AMEL ETHEL ILPAP ISAP OASTH

Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value

Punctuality 0.364 0.091 3.996 0.387 0.169 2.280 1.163 0.319 3.645
Passenger safety

ARTICLE IN PRESS
Transfer distance 0.201 0.083 2.408 0.262 0.134 1.955
Network coverage 0.288 0.071 4.033 0.224 0.113 1.983 0.393 0.235 1.668 0.444 0.223 1.993 0.314 0.137 2.285
Vehicle cleanliness 0.179 0.083 2.153
Driver behavior 0.202 0.092 2.206 0.522 0.251 2.074 0.285 0.123 2.305
Waiting conditions 0.404 0.079 5.073 0.432 0.238 1.812 0.301 0.129 2.334 0.626 0.145 4.305
In-vehicle service conditions 0.597 0.242 2.461 0.713 0.310 2.299
Ticketing systems 0.344 0.216 1.589
Behavior of non-driver staff 0.131 0.082 1.600 0.355 0.127 2.784
Service frequency 0.236 0.092 2.556 0.349 0.181 1.925 0.617 0.249 2.471 0.445 0.148 3.004
Travel time 0.411 0.228 1.803
Waiting time
Price 0.114 0.070 1.624 0.395 0.124 3.179
Information about schedule 0.219 0.132 1.665
Information about transfers 0.549 0.232 2.365

Intercepts
1|2 10.90 0.57 19.26 10.52 1.47 7.15 10.94 1.05 10.40 – – – 11.10 1.63 6.82 11.41 1.04 10.98
2|3 7.19 0.45 15.91 6.32 1.06 5.98 6.47 0.67 9.71 7.65 1.22 6.24 7.63 1.30 5.85 7.57 0.81 9.37
3|4 3.37 0.38 8.93 2.59 0.86 3.00 2.29 0.52 4.40 3.62 0.92 3.89 3.18 0.99 3.21 3.83 0.65 5.84

Residual deviance 1345.1 160.0 465.7 153.6 133.0 367.3

AIC 1369.1 171.0 483.7 165.6 145.0 385.3
Table 4
Ordered logit model results (male respondents)

All operators AMEL ETHEL ILPAP ISAP OASTH

Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value Value S.E. t-Value

Punctuality 0.694 0.236 2.933

ARTICLE IN PRESS
Passenger safety 0.275 0.139 1.980
Transfer distance 0.215 0.096 2.242
Network coverage 0.167 0.080 2.096 0.309 0.141 2.193 0.531 0.221 2.394
Vehicle cleanliness 0.422 0.084 4.998 0.534 0.203 2.627 0.654 0.156 4.176 0.713 0.244 2.917
Driver behavior 0.361 0.187 1.930
Waiting conditions
In-vehicle service conditions 0.844 0.311 2.705 0.636 0.182 3.493
Ticketing systems 0.201 0.077 2.585 0.320 0.138 2.317 0.932 0.293 3.176
Behavior of non-driver staff 0.181 0.077 2.327 0.556 0.191 2.901 0.875 0.352 2.483 0.408 0.207 1.972
Service frequency 0.674 0.096 7.056 0.465 0.162 2.863 1.004 0.314 3.194 0.796 0.174 4.558
Travel time
Waiting time
Price
Information about schedule 0.254 0.129 1.972 0.439 0.173 2.535
Information about transfers 0.181 0.092 1.977

Intercepts
1|2 10.20 0.58 17.57 7.08 1.11 6.39 11.25 1.06 10.66 16.17 2.92 5.52 11.58 1.81 6.38 10.16 1.22 8.31
2|3 6.98 0.48 14.59 3.70 0.89 4.17 7.82 0.88 8.82 12.87 2.56 5.02 7.56 1.40 5.40 6.83 0.97 7.01
3|4 3.28 0.39 8.33 0.86 0.81 1.07 3.76 0.72 5.19 8.01 2.14 3.73 3.88 1.07 3.61 3.25 0.82 3.96

Residual deviance 1085.9 195.8 384.5 99.9 126.9 236.6

AIC 1105.9 205.8 402.5 113.9 140.9 250.6

271
 ARTICLE IN PRESS

272 Y. Tyrinopoulos, C. Antoniou / Transport Policy 15 (2008) 260–272

public transit operators and authorities should include in their Hellenic Institute of Transport (HIT), 2005. Application manual of the service and
quality-control programs such methods, and facilitate the for- performance indicators. Stage 2 Final Report of the Project: An Integrated
Quality Control System for the Public Transport Services in Thessaloniki.
mulation of short- and medium-term action plans. Depending on Johnson, R., Wichern, D., 1992. Multivariate Statistical Analysis, third ed. Prentice-
the aim and nature of a survey and the level of detail of the Hall, Englewood Cliffs, NJ.
collected data, this proposed approach may assist in operational Likert, R., 1932. A technique for the measurement of attitudes. Archives of
Psychology 140, 55.
adjustments related to service frequencies, network expansion, PORTAL, 2003. Promotion of results in transport research and learning. Final
transfer points, network coverage and many others. Report, EU, RTD Programme.
QUATTRO, 1998. Quality approach in tendering/contracting urban public transport
operations. Final Report, EU, RTD Programme.
References R Development Core Team, 2008. R: A Language and Environment for
Statistical Computing. R Foundation for Statistical Computing, Vienna,
Akaike, H., 1974. A new look at the statistical model identification. IEEE Austria.
Transactions on Automatic Control 19 (6), 716–723. Transportation Research Board, 1998. A handbook: integrating market research
Ben-Akiva, M., Lerman, S.R., 1985. Discrete choice analysis. MIT Press, into transit management. TRCP Report 37.
Cambridge, MA. Transportation Research Board, 1999. A handbook for measuring customer
EQUIP, 2000. Extending the quality of public transport. Final Report and its Annex: satisfaction and service quality. TRCP Report 47.
Practical Handbook, EU, RTD Programme. Transportation Research Board, 2004. Transit Capacity and Quality of Service
European Committee for Standardization (CEN), 2002. Transportation—logistics Manual. TRCP Report 100, second ed.
and services—public passenger transport—service quality definition, targeting TRL Limited, 2004. The demand for public transport: a practical guide. TRL Report
and measurement. TRL593. ISSN 0968-4107, Crowthorne, UK.
European Performance Satisfaction Index (EPSI) Rating Institute, 2005. Customer Venables, W.N., Ripley, B.D., 2002. Modern Applied Statistics with S, fourth ed.
Satisfaction Index 2005. Springer.
Hellenic Institute of Transport (HIT), 2003. Handbook for the implementation of Washington, S.P., Karlaftis, M.G., Mannering, F.L., 2003. Statistical and
the quality control system of OASA. Stage 2 Final Report of the Project: An Econometric Methods for Transportation Data Analysis. Chapman & Hall/
Integrated System for the Quality Assessment of the OASA Passenger Services. CRC, London.