Public Transp

DOI 10.1007/s12469-014-0096-x

ORIGINAL PAPER

Relationships between rail passengers’ satisfaction

and service quality: a framework for identifying key
service factors

Laura Eboli • Gabriella Mazzulla

Ó Springer-Verlag Berlin Heidelberg 2014

Abstract Increasing the use of public transport is one of the most convenient
strategies for alleviating the problems resulting from the excessive use of the private
car in most metropolitan areas (congestion, pollution, noise, etc.). In order to improve
public transport, developing appropriate tools for measuring and monitoring service
quality is necessary. Among the various methods for measuring transit service quality
the authors choose to adopt a method based on customer perspective because they
retain that customers have the right elements for appropriately judging the used ser-
vice. Specifically, a structural equation model is formulated to explore the impact of
the relationship between global customer satisfaction and service quality attributes,
such as safety, cleanliness, main and additional services, information about the ser-
vice, and personnel, and to verify which of these attributes are key factors of service
quality. Services offered by rail operators in Northern Italy are analysed (32 regional
lines, 9 suburban lines, 2 express lines). Data collected by a survey addressed to a
sample of more than 16,000 passengers are used for calibrating the model. The pro-
posed model can be useful both to transport agencies and planners to analyze the
correlation between service quality attributes and identify the most convenient attri-
butes for improving the supplied service. The main findings are that information,
cleanliness, and service characteristics like punctuality and frequency of runs have the
highest positive effect on service quality. In addition, the authors experimented with
the introduction of indicators explaining the endogenous latent variable different from
the traditionally used indicators, such as an index based on both satisfaction and
importance rates.

L. Eboli  G. Mazzulla (&)

Department of Civil Engineering, University of Calabria, P. Bucci, cubo 46/B, 87036 Rende, CS,
Italy
e-mail: gabriella.mazzulla@unical.it
L. Eboli
e-mail: laura.eboli@unical.it

123
 L. Eboli, G. Mazzulla

Keywords Service quality  Railway service  Structural equation modelling

1 Introduction

Promoting more sustainable transport modes to alleviate the problems resulting

from the excessive use of the private car in most metropolitan areas (congestion,
pollution, noise, etc.) is one of the main concerns of transportation planners (de Ona
et al. 2012). At this aim increasing the use of public transport is the most important
strategy; this is convenient for both private car users and present public transport
users. In order to improve public transport, developing appropriate tools for
measuring and monitoring service quality is necessary. These tools should take into
account the different characteristics of a service, which concern various aspects
such as service programming, reliability, comfort, information, personnel, and so
on. Transit service quality, in fact, depends on the quality levels of the factors
describing the service. Given the numerousness of these factors, it is very difficult to
preliminarily select the factors to be considered for evaluating service quality, so the
analyst would have a deep knowledge about the various service aspects (for a
description of the service quality factors see TRB 2004).
In the literature, there are very different methods for measuring service quality.
Some methods are based on the use of customer satisfaction data, being the
customers as one of the key determinants to measure the quality of a service (an
exhaustive review is reported in Eboli and Mazzulla 2010); other methods are based
on other kinds of data (e.g. objective indicators and measures) (Eboli and Mazzulla
2011).
Among the various methods for measuring transit service quality the authors
choose to adopt a method based on customer’s perspective because they retain,
supported also by the literature review of the sector, that customers have the right
elements for appropriately judging the used service; moreover, passengers are the
direct users of the service, and for this reason an analysis based on their perceptions
allows the establishment of the elements retained as critical by the same users, and
the possibility of improving more effectively the service.
In addition, passengers’ opinions about the characteristics of the service allow
preferences about the different service quality attributes to be determined. Several
approaches have been used to estimate the relative importance of each attribute with
regards to the service quality perceived by each customer. There are methods based
on asking customers to rate each attribute on an importance scale, or methods based
on deriving a measure of attribute importance by statistically testing the strength of
the relationship of individual attributes with overall satisfaction (de Ona et al.
2012). Recent years have seen the development of these methods based on
traditional customer satisfaction surveys (Cavana et al. 2007; Dell’Olio et al. 2010;
Eboli and Mazzulla 2007; Jen et al. 2011; Joewono and Kubota 2007; Nurul-Habib
et al. 2011; Pakdil and Aydin 2007; Weinstein 2000) and from stated preference
surveys (Cirillo et al. 2011; Eboli and Mazzulla 2008a, b, 2010; Hensher and Prioni
2002; Hensher et al. 2003; Dell’Olio et al. 2011; Imaz et al. 2014).

123
Relationships between rail passengers’ satisfaction and service quality

The objective of this research is to propose a methodology aimed to establish the

importance of the various characteristics of a transit service on the overall service
quality, in order to identify the key factors for the users of transit services. The
authors have chosen the Structural Equation Modelling (SEM) methodology
because they retain it as appropriate for describing a complex phenomenon like
transit passenger perception of the used service. SEM is a technique which can be
considered as similar to the regression modelling but more advanced, in fact it
permits to introduce latent constructs really appearing in such a phenomenon where
there are some latent factors due to the subjectivity of users’ perceptions and the
variety of the service attributes characterizing a transit service. SEM allows latent
variables to be introduced for better describing the relationship among the service
factors and with the overall service quality, expressed in terms of customer
satisfaction.
SEM was adopted in several fields of research and generalized by Joreskog and
Wiley (Joreskog 1973; Wiley 1973). SEM was applied in Psychology and Social
Science, Natural Science, and especially in the field of Economics and Statistics. In
the field of transportation research some applications of SEM were proposed to
analyze land-use and transport interactions (e.g. Tschopp and Axhausen 2007; Van
Acker et al. 2007; de Abreu e Silva and Goulias 2009; de Abreu e Silva et al. 2012;
Eboli et al. 2012). Also in public transport some authors proposed SEM
applications, such as Bamberg and Schmidt (1998), Fillone et al. (2005), Tam
et al. (2005), Wen et al. (2005), Webb (2010), Lai and Chen (2011), Bass et al.
(2011), Trépanier et al. (2012). More specifically, SEM was also adopted for
investigating on customer satisfaction on public transport services, but there are not
many studies in this field (e.g. Eboli and Mazzulla 2007, 2012; de Ona et al. 2013).
Therefore, the main purpose of this study is to examine whether or not SEM can
effectively identify the key factors affecting transit service quality. The authors now
propose a work aimed on an investigation of the influence of a series of service
quality attributes on the overall service quality of a railway service. In the model
structure there are latent exogenous variables representing service quality macro-
factors, each explained by a series of observed service quality attributes, and a latent
endogenous variable representing the overall service quality, explained by observed
global indicators.
In the literature, a similar structure based on SEM was proposed by Irfan Syed
et al. (2011), who aimed to investigate the passengers’ perceptions about service
quality of rail transport system in Pakistan while travelling between the major cities.
Several latent variables, such as tangible, empathy, assurance, responsiveness,
timeliness, food, information, safety, as Parasuraman et al. (1988) defined in the
SERVQUAL methodology, were investigated as affecting passenger satisfaction.
Another similar work was proposed by Ngatia et al. (2010) who introduced
unobserved attributes such as service quality, safety and travel cost related to the
overall satisfaction. In both works the latent endogenous variable representing the
overall service quality is measured through the satisfaction rates. These two works
are following Eboli and Mazzulla (2007) in which a model for investigating the
impact of bus transit aspects on global customer satisfaction was proposed, based on
an urban bus service of an area of the Southern Italy habitually used by university

123
 L. Eboli, G. Mazzulla

students. The model relates three exogenous latent variables concerning network
design, service planning and reliability, and comfort to an endogenous latent
variable representing the overall satisfaction; differently from the works mentioned
above, in this case the overall satisfaction was expressed by two observed variables,
satisfaction rates and importance rates. Just in this previous work the authors
attempted to explain the latent endogenous variable through another observed
indicator in addition to passengers’ satisfaction. A similar structure was proposed in
another work of Eboli and Mazzulla (2012) based on the data collected by the
sample survey adopted in the present work. Specifically, a structural equation model
was formulated to explore the impact of the relationship between global customer
satisfaction and service quality attributes, such as safety, cleanliness, main and
additional services, information, and personnel. Also in this case, satisfaction rates
and importance rates were used to express the overall satisfaction.
In the proposed work, we adopt the same structure reported in Eboli and
Mazzulla (2012) where users’ perceptions are expressed in terms of satisfaction, but
the most relevant difference compared to the previous works of the authors and the
other similar works is the introduction of more sophisticated indicators measuring
the overall service quality. In fact, we chose to introduce a synthetic indicator based
on user perceptions and expectations, respectively, expressed in terms of
satisfaction and importance rates, and an indicator based on statements of the
users concerning their recent problems with the service factors. A particular aim of
this work is just to verify the goodness and appropriateness of these two more
elaborate measures for expressing the overall service quality, and to propose better
indicators for measuring the latent endogenous variable representing the overall
service quality.
To support the research, an original questionnaire survey to public transport users
was designed and conducted in the city of Milan, one of the most important cities of
Italy, and its hinterland.
The paper is structured in four sections. After this introduction containing also a
brief literature review of works proposing tools for measuring service quality, there
is a section describing the methodology and framework adopted in this context. The
third section is about the study case adopted as a support of the research: the survey
conducted for collecting data is briefly described, the main characteristics of the
sample are shown together with an analysis of the importance and satisfaction rates;
then, the results of the introduced model are shown and discussed. Finally, we
propose a brief conclusive discussion about the work.

2 Methodology

2.1 Structural equation modeling: theoretical framework

SEM is a relatively new method whose use is now rapidly expanding as user
friendly software, like AMOS (Arbuckle and Wothke 1995), becomes available.
SEM is a specific type of regression analysis and explains relationships between
independent (exogenous) and dependent (endogenous) variables. It is composed of

123
Relationships between rail passengers’ satisfaction and service quality

up to three sets of simultaneous equations, estimated at the same time: (1) a

measurement model for the endogenous variables, (2) a measurement model for the
exogenous variables, and (3) a structural model. This full model is known as ‘‘SEM
with latent variables’’. Latent variables are constructs which cannot be directly
observed, but they must be defined in terms of underlying observed variables, called
indicators. Each latent variable is defined by a measurement model, whereas the
structural model represents the relationships between exogenous and endogenous
variables. The basic equation of the latent variable model is the following (Bollen
1989):
g ¼ Bg þ Cn þ f ð1Þ
in which g (eta) is an (m 9 1) vector of the latent endogenous variables, n (xi) is an
(n 9 1) vector of the latent exogenous variables, and f (zeta) is an (m 9 1) vector
of random variables. The elements of the B (beta) and C (gamma) matrices are the
structural coefficients of the model; the B matrix is an (m 9 m) coefficient matrix
for the latent endogenous variables; the C matrix is an (m 9 n) coefficient matrix
for the latent exogenous variables.
The basic equation of the measurement model (2) is for the exogenous variables,
and Eq. (3) is for the endogenous variables:
x ¼ Kx n þ d ð2Þ

y ¼ Ky g þ e ð3Þ
in which x and d (delta) are column q-vectors related to the observed exogenous
variables and errors, respectively; Kx (lambda) is a (q 9 n) structural coefficient
matrix for the effects of the latent exogenous variables on the observed variables;
y and e (epsilon) are column p-vectors related to the observed endogenous variables
and errors, respectively; Ky is a (p 9 m) structural coefficient matrix for the effects
of the latent endogenous variables on the observed ones.
The structural equation system is generally estimated by using the Maximum
Likelihood method (ML). In other cases, the structural equation model parameters
can be estimated by using other estimation methods, such as Unweighted Least
Squares (ULS), Weighted Least Squares (WLS), Generalized Least Squares (GLS),
and so on. These estimation methods are described in Bollen (1989) and
Washington et al. (2003), in which useful information about goodness-of-fit
measures and their statistical interpretation are also provided. For a more detailed
discussion on structural equation models one should refer to Joreskog (1973),
Bollen (1989), Bagozzi (1994) and Golob (2003).

2.2 The proposed framework

In this work SEM methodology was applied for modelling the phenomenon relating
to the measurement of service quality based on passengers’ perceptions; unobserved
latent constructs representing the main service quality characteristics are introduced,
together with indicators representing all the observed service quality factors. The

123
 L. Eboli, G. Mazzulla

issue of service quality measurement based on passengers’ perceptions can be

conveniently described through structures simulating the presence of latent
constructs, considering that service quality depends on the quality of various
factors; this large number of service quality attributes presumes a presence of latent
constructs grouping some service attributes. In the proposed model we supposed the
presence of seven n latent exogenous constructs representing seven main
characteristics of a railway service, Safety, Cleanliness, Comfort, Service,
Additional Services, Information and Personnel; these latent factors are supposed
to be linked to a latent construct representing the overall service quality, simulated
by a g latent endogenous variable, named as Service Quality. Each latent construct
is explained by really observed service quality factors; so, the latent endogenous
variables are linked to 33 observed variables, as represented in Fig. 1, while the
latent endogenous variable is linked to three observed indicators of global service
quality. The first one is simply the satisfaction rate expressed by each user on the
overall service. However, not all the attributes are important for the user in the same
way, and then, these differences cannot be taken into account by an index based
only on the satisfaction rates. For this reason, we considered a second indicator, the
Customer Satisfaction Index (CSI) (Hill et al. 2003), which is a measure of service
quality based on users’ perceptions on some service aspects expressed in terms of
satisfaction rates, compared with users’ expectations expressed in terms of
importance rates. CSI is calculated by means of the satisfaction rates expressed
by the users, weighted on the basis of the importance rates; specifically, it is

F4 F5 F6 F7 F8 F9

F1 F2 F3 Cleanliness F10 F11 F12

Safety Comfort
F13

F14
Satisfaction
F15
CSI Service Quality Service
F16
Critical Event
F17

F18
Additional
Personnel Services

Information
F33 F32 F31 F30 F22 F21 F20 F19

F29 F28 F27 F26 F25 F24 F23

Fig. 1 Structural equation model

123
Relationships between rail passengers’ satisfaction and service quality

introduced an importance weight, which is the ratio between the mean of the
importance rates expressed by the users on each attribute and the sum of the average
importance rates of all the service quality attributes. CSI represents a good measure
of the overall satisfaction because it summarizes the judgments expressed by the
users about various service attributes in a single score (Eboli and Mazzulla 2009).
The third indicator is represented by the number of factors for which the user has
experienced problems in the last 30 days before the interview; this kind of
information was introduced in the literature of the sector on the occasion of the
development of the method of the Impact Score proposed by the Transportation
Research Board (TRB 1999).
The regression weights were obtained by solving a system of 37 equations (33
equations relating to the x endogenous variables, 3 equations to the y endogenous
variables, 1 equation relating to the g latent endogenous variable), and by using the
data collected through the survey described in the next section.

3 Application of the framework: a case study of Milan

3.1 Survey

A survey was addressed to a sample of users of a railway service operating in the

North of Italy. The service offers different types of connections: 32 regional lines
and 9 suburban lines connecting different towns of the hinterland of the city of
Milan, and 2 express lines connecting Milan with the Malpensa airport.
Face-to-face interviews were realized in the period from the end of June to the
middle of July 2011, during the whole week and then considering weekday, before
a holiday, and holiday days, in a time slot between 6.00 a.m. and 10.00 p.m.
Passengers were prevalently interviewed on board during their journey. A final
simple randomly selected sample of 16,718 people aged above 16 was considered
for analyzing the investigated services, guaranteeing a sample rate of 3 %
(calculated on the basis of the registered number of passengers per day of
572,040).
The questionnaire was structured into two main sections. The first section had the
aim of collecting data concerning: general information (e.g. time period of the
interview, train, line, station, and operator); socio-economic characteristics (e.g.
gender, age, qualification, professional condition, car availability, income); travel
habits (e.g. trip scope and frequency, ticket, transport modes connecting with the
train stations). The second section was more oriented on the passenger perceptions
about the used services. Specifically, users expressed importance and satisfaction
rates, on a cardinal scale from 1 to 10, about 33 service quality factors concerning
safety, cleanliness, main and additional services, information, and personnel. In
addition, users indicated for each service quality factor the occurrence of a critical
event in the last 30 days before the interview.

123
 L. Eboli, G. Mazzulla

Users also provided deeper information about some railway services such as
convenience of this kind of services compared to the others, or ticket purchasing, or
information services.
About 53 % of the interviewed people are passengers of the regional lines, about
40 % of the suburban lines, and the remaining 7 % travels through the Malpensa
express services. 80.5 % of users were interviewed on a weekday, 14.2 % on a day
before a holiday, and 5.3 % on a holiday. Most of the passengers were interviewed
in the off-peak hours (56.0 %), 18.7 % in the afternoon peak-hours, 13.8 % in the
evening peak hours, and 11.5 % in the morning peak-hours. 30.6 % of sample are
people travelling to work, 13.6 % to study, and the remaining 55.8 % travel for
other purposes.

3.2 Sample characteristics

The users of the railway services who answered to the questionnaire build up a
sample almost equally spread between females and males, even if females prevail
over males (Table 1). The major part of the interviewed passengers is young (more
than 70 %), that is, they are aged at most 40. Most of the sampled people are
employed (about 60 %), and a considerable part is composed of students (about
30 %); the remaining part is spread among unemployed people, housewives and
pensioners. About 35 % do not give any kind of information about the income, but
most people state that their income is lower than 2,000 Euros per month (56 % of
the whole sample). Only one-third of the sample has a degree, whereas two-thirds
obtained just a diploma of second level. More than half of the sampled people own a
private car (58 %).
The interviewed passengers travel by train prevalently for reaching the place of
work (42 %) or the study places (21 %), but a considerable part of the sample
(36 %) travels for bureaucratic or personal activities, and for tourism. Almost 65 %
of the sample habitually travel by train, whereas only 35 % of the passengers travel
occasionally. Interviewed passengers purchase tickets (One-way, return, or multi-
runs ticket) or travel cards (weekly, monthly or yearly travel card) in equal measure.
Finally, responses about the transport modes used by passengers for arriving to/
moving from the stations indicated that most of the sample arrive at the stations on
foot (about 37 %) and move from the station also on foot (about 38 %). Less than
one-third of the passengers reach the station or move from the station by car, and the
remaining part by other transit systems.
Definitively, the interviewed passenger is prevalently a young employed, with a
monthly net income lower than 2,000 Euros, who habitually travels by train for
reaching the place of work, and then by using prevalently a travel card; the traveler
arrives to/moves from the station on foot.

3.3 Service quality evaluation

A preliminary evaluation of service quality can be made by means of the importance

and satisfaction rates expressed by the interviewed users, on a scale from 1 to 10, on

123
Relationships between rail passengers’ satisfaction and service quality

Table 1 Sample characteristics

Characteristics Categories Percentages (%)

1. Gender Male 45.5

Female 54.5
2. Age B40 years old 70.5
[40 years old 29.5
3. Professional condition Employed 58.7
Student 30.5
Other (unemployed, housewife, pensioner, etc.) 10.8
4. Monthly net income level B2,000 Euros 56.3
[2,000 Euros 8.7
No answer 35.0
5. Qualification Degree 32.9
Diploma 67.1
7. Car ownership Have car 58.1
Have no car 41.9
8. Scope of journey Work 42.3
Studying 21.3
Other (bureaucratic and/or personal activities, tourism) 36.4
9. Frequency of journey Habitually (daily or weekly) 64.8
Occasionally 35.2
10. Ticket kind Ticket (One-way, return, or multi-runs ticket) 48.9
Travel card (weekly, monthly or yearly travel card) 51.1
11. Access transport mode On foot 37.3
By car, bicycle or motorcycle 29.9
By other transit systems 27.8
No answer 5.0
12. Egress transport mode On foot 38.6
By car, bicycle or motorcycle 24.8
By other transit systems 29.1
No answer 7.5

33 service quality attributes, which are the observed variables in the model. Table 2
shows the average rates calculated from the collected data.
Users retain all the attributes as very important (having an average rate of
importance around 8 and 9) with the exception of the attribute linked to the possibility
to transport the bicycle on board, for which an average importance rate of 7.3 was
obtained. On the contrary, by observing the average satisfaction rates we can say that
people judge most of the service characteristics as being not very good; in fact, only
nine attributes have an average rate higher than the sufficiency ([6). More specifically,
the attributes considered as the most important are the three service aspects linked to
travel safety and personal security, which have an average importance rate higher than
9. Among the attributes considered as relatively less with a relatively smaller

123
 L. Eboli, G. Mazzulla

importance there are the services linked to the possibility of parking at stations,
maintenance of the stations, activities of communication to the offices, activities of
ticket inspection, which, however, show an importance rate higher than 8.

Table 2 Importance and Satisfaction rates, and Customer Satisfaction Index (CSI)
Service quality attribute Importance Importance Satisfaction Satisfaction CSI
rate var rate var

F1 Travel safety 9.2 2.6 7.4 4.6 0.24

F2 Personal security on board 9.1 2.6 6.7 4.9 0.24
F3 Personal security at station 9.1 2.8 6.5 5.1 0.21
F4 Cleanliness of vehicles 8.9 3.4 5.0 5.5 0.16
F5 Cleanliness of seats 8.9 3.5 4.8 5.6 0.15
F6 Maintenance of seats 8.6 3.5 5.1 5.6 0.16
F7 Cleanliness of toilet 8.8 3.8 4.4 5.7 0.14
facilities
F8 Cleanliness of stations 8.5 3.3 5.3 4.9 0.16
F9 Maintenance of stations 8.3 3.7 5.4 4.9 0.16
F10 Crowding on board 8.4 3.5 5.4 5.7 0.16
F11 Air-conditioning on 8.7 3.1 5.1 6.0 0.16
board
F12 Comfort on board 8.4 3.1 5.6 5.2 0.16
F13 Fare/service ratio 8.8 3.4 5.1 5.5 0.16
F14 Frequency of runs 8.9 2.6 5.9 5.1 0.19
F15 Punctuality of runs 9.0 2.9 5.4 5.8 0.17
F16 Regularity of runs 9.0 2.6 5.7 5.2 0.18
F17 Integration with PT 8.7 3.0 6.0 4.9 0.18
F18 Localization of Stations 8.6 2.7 6.5 4.4 0.20
F19 Parking 8.0 4.9 5.7 5.5 0.16
F20 Bicycle transport on 7.3 5.8 5.8 4.6 0.15
board
F21 Facilities for the disabled 8.8 3.8 5.2 5.7 0.16
F22 Substitute services 8.4 4.0 5.4 4.9 0.16
F23 Information at stations 8.7 2.9 5.9 4.7 0.18
F24 Information on board 8.5 3.3 5.5 5.0 0.17
F25 Info timeliness at stations 8.7 3.0 5.5 5.0 0.17
F26 Info timeliness on board 8.6 3.2 5.3 5.2 0.16
F27 Complaints 8.5 3.7 5.0 5.5 0.15
F28 Communication to office 8.3 3.7 5.1 5.3 0.15
F29 Info connections with PT 8.5 3.3 5.4 5.0 0.16
F30 Kindness on board 8.5 2.7 6.6 4.2 0.20
F31 Competence on board 8.7 2.5 6.6 4.1 0.20
F32 Ticket inspection 8.3 3.8 6.3 5.2 0.18
F33 Kindness at station 8.6 2.9 6.4 4.9 0.19
Overall service 5.8 4.6 5.7

123
Relationships between rail passengers’ satisfaction and service quality

Table 3 Model results

RW SE P st. RW

Latent endogenous variable Latent exogenous variable

Service quality (g1) Safety (n1) 0.080 0.003 0.000 0.191
Service quality (g1) Cleanliness (n2) 0.154 0.003 0.000 0.525
Service quality (g1) Comfort (n3) 0.111 0.003 0.000 0.287
Service quality (g1) Service (n4) 0.200 0.004 0.000 0.488
Service quality (g1) Additional Services (n5) 0.136 0.004 0.000 0.289
Service quality (g1) Information (n6) 0.240 0.005 0.000 0.560
Service quality (g1) Personnel (n7) 0.114 0.003 0.000 0.280
Observed endogenous variable Latent exogenous variable
F1 (x1) Travel safety Safety (n1) 1 – – 0.737
F2 (x2) Personal security on board Safety (n1) 1.354 0.012 0.000 0.962
F3 (x3) Personal security at station Safety (n1) 1.225 0.011 0.000 0.858
F4 (x4) Cleanliness of vehicles Cleanliness (n2) 1 – – 0.941
F5 (x5) Cleanliness of seats Cleanliness (n2) 1.018 0.004 0.000 0.952
F6 (x6) Maintenance of seats Cleanliness (n2) 0.967 0.004 0.000 0.911
F7 (x7) Cleanliness of toilet facilities Cleanliness (n2) 0.817 0.005 0.000 0.808
F8(x8) Cleanliness of stations Cleanliness (n2) 0.737 0.006 0.000 0.736
F9 (x9) Maintenance of stations Cleanliness (n2) 0.691 0.006 0.000 0.697
F10 (x10) Crowding on board Comfort (n3) 1 – – 0.715
F11 (x11) Air-conditioning on board Comfort (n3) 1.268 0.013 0.000 0.872
F12 (x12) Comfort on board Comfort (n3) 1.134 0.012 0.000 0.845
F13 (x13) Fare/service ratio Service (n4) 1 – – 0.682
F14 (x14) Frequency of runs Service (n4) 1.094 0.012 0.000 0.775
F15 (x15) Punctuality of runs Service (n4) 1.235 0.013 0.000 0.820
F16 (x16) Regularity of runs Service (n4) 1.168 0.012 0.000 0.836
F17 (x17) Integration with PT Service (n4) 0.983 0.011 0.000 0.758
F18 (x18) Localization of stations Service (n4) 0.899 0.011 0.000 0.713
F19 (x19) Parking Additional Services (n5) 0.899 0.013 0.000 0.587
F20 (x20) Bicycle transport on board Additional Services (n5) 0.848 0.011 0.000 0.699
F21 (x21) Facilities for the disabled Additional Services (n5) 1.091 0.013 0.000 0.754
F22 (x22) Substitute services Additional Services (n5) 1 – – 0.767
F23 (x23) Information at stations Information (n6) 1.137 0.010 0.000 0.811
F24 (x24) Information on board Information (n6) 1.254 0.011 0.000 0.864
F25 (x25) Info timeliness at stations Information (n6) 1.258 0.010 0.000 0.876
F26 (x26) Info timeliness on board Information (n6) 1.269 0.010 0.000 0.879
F27 (x27) Complaints Information (n6) 1.055 0.010 0.000 0.762
F28 (x28) Communication to office Information (n6) 1.020 0.010 0.000 0.764
F29 (x29) Info connections with PT Information (n6) 1 – – 0.757
F30 (x30) Kindness on board Personnel (n7) 1.164 0.009 0.000 0.923
F31 (x31) Competence on board Personnel (n7) 1.158 0.009 0.000 0.941
F32 (x32) Ticket inspection Personnel (n7) 0.967 0.011 0.000 0.694
F33 (x33) Kindness at station Personnel (n7) 1 – – 0.745
Observed endogenous variable Latent endogenous variable
Satisfaction (y1) Service quality (g1) 1 – – 0.423
CSI (y2) Service quality (g1) 1.115 0.020 0.000 0.748
Critical Event (y3) Service quality (g1) -3.110 0.095 0.000 -0.287

123
 L. Eboli, G. Mazzulla

Concerning satisfaction rates, we can see that the service characteristics

considered as the most satisfying are linked to all the characteristics regarding
safety, to all the aspects concerning the personnel, and to the integration with the
other modes of public transport and the localization of the stations. On the contrary,
the characteristics judged as the less satisfying are linked to aspects concerning
cleanliness, comfort, and information. Specifically, two attributes registered a
satisfaction rate lower than 5: cleanliness of seats and cleanliness of toilet facilities.
By observing the variance of the importance and satisfaction rates (Table 2), we
can easily observe that users’ perceptions in terms of satisfaction are more
heterogeneous than importance judgments. In fact, the average variance calculated
by considering the variance of all the satisfaction rates is about 5.1, while the
corresponding average variance relating to the importance rates is about 3.3. The
attribute judged as the most heterogeneous in terms of satisfaction is the aspect
relating to air-conditioning on board, while the less heterogeneous is the aspect
linked to the competence of the personnel on board. Concerning importance rates,
the attribute judged as the most heterogeneous is relating to the possibility to
transport the bicycle on board, while the less heterogeneous is also in this case the
aspect linked to the competence of the personnel on board.
For the analyzed service, there is no substantial difference between the overall
service satisfaction rate (5.8) and CSI (5.7), because of the homogeneity in the
judgments expressed by the user in terms of importance rate (Table 2). However,
both the indicators show a good service quality level.
Finally, some observations can be made about the critical events registered by
users. There is only one attribute for which a very large part of the user experienced
a critical event in the last 30 days before the interview; this attribute is linked to
personal security on board (14,715 occurrences out of 16,623 interviewed users).
This result appears conflicting with the high registered satisfaction rate (6.7);
evidently, most of the problems experienced by the passengers are not substantial,
and they feel quite safe on board. For certain service aspects, like cleanliness of
vehicles, seats and toilet facilities, crowding and air-conditioning on board, and
punctuality of runs, the number of registered critical events is, in any case, notable
(over than 6,000 occurrences). For the other service quality aspects the number of
occurrences is, instead, relatively less important (about 2,000 or 3,000 occurrences
out of 16,623 interviewed users).

3.4 Model results

The structural equation system was estimated by using the ML method on the basis
of 16,623 observations, and specifically 8,782 regarding regional services, 6,718
suburban services, and 1,123 express services. The model was calibrated by using
the AMOS 4.0 package from SmallWaters Corporation (Arbuckle and Wothke
1995).
Table 3 shows the model results. In the first and second column the model
variables are reported; the third column shows the values of the regression weights
(RW) of the coefficients; the fourth and fifth column contain the values of the
standard error (SE) of each coefficient and the probability levels (P) that the

123
Relationships between rail passengers’ satisfaction and service quality

estimated coefficient is significantly different from zero; finally, in the last column
the values of the standardized regression weights (st. R.W.) are shown.
All the parameters have a correct sign and assume a statistically different from
zero value, at a good level of significance. The minimum value of discrepancy
function is statistically significant according to the Chi squared test. Because the Chi
squared test of absolute model fit is sensitive to sample size and non-normality in
the underlying distribution of the input variables, various descriptive fit statistics
may be used to assess the overall fit of a model to the data. The Goodness of Fit
Index (GFI) is 0.651, the Adjusted Goodness of Fit Index (AGFI) is 0.605, and the
Comparative Fit Index (CFI) is 0.785. These tests are quite satisfactory. The Root
Mean Square Error of Approximation (RMSEA) has a value of 0.105 and its lower
and upper confidence interval boundaries are 0.105 and 0.106, respectively; besides,
Root Mean Residual (RMR) is lower than 2.1. These indexes have low values and
therefore are quite good.
The latent exogenous variables with the highest positive effect on Service Quality
are Information (0.560), Cleanliness (0.525) and Service (0.488). Directly after,
there is another group of three service aspects presenting similar coefficients:
Additional Services (0.289), Comfort (0.287), and Personnel (0.280). Finally, Safety
has the lowest effect on Service Quality (0.191).
The relationship between the latent endogenous variable and observed variables
shows that the CSI indicator widely explains Service Quality (being the coefficient
value equal to 0.748), while satisfaction and critical event indicators explain a
smaller part of Service Quality latent construct (0.423, and -0.287, respectively).
By observing the relationship between the latent exogenous variables and their
observed indicators, some interesting results can be highlighted. First, it is clearly
evident that comparative weights of the observed indicators for each latent variable
are very similar to each other; it is surely due to the fact that service attributes
included in the same macro-factor registered very similar satisfaction rates.
However, the following distinctions can be underlined. The values of the
coefficients of the indicators explaining Safety latent variable vary from 0.962,
for personal security on board (the most important indicator) to 0.737, for travel
safety (the least important one). Cleanliness is prevalently understood as degree of
cleanliness of seats and of vehicles (0.952 and 0.941, respectively), and the
maintenance of the seats (0.911); the other indicators have lower weights, up to a
minimum value of 0.697, registered for the attribute linked to the maintenance of
the stations. The relationship between Comfort and its observed indicators shows
unexpected results; in fact, Comfort is prevalently interpreted by the passengers as
air-conditioning (0.872) and degree of comfort on board (0.845), but less as level of
crowding on board (0.715), which often is the only comfort indicator used by the
researchers who analyze service quality in public transport. The latent variable
representing Service characteristics is well explained by the attributes of service
reliability: regularity (0.836) and punctuality (0.820); directly after, frequency of
runs and the degree of integration between the services analyzed and others local
transit services present similar weights (0.775 and 0.758, respectively); finally,
localization of station has a weight of 0.713. Additional Services are moreover
understood as services substitute of irregular services (0.767), besides as facilities

123
 L. Eboli, G. Mazzulla

for disabled (0.754). Information is prevalently explained by the timeliness in

having the information both on board and at stations (having very similar
coefficients equal to 0.879 and 0.876, respectively), and also by the possibility to
have some information about the services (0.864 on board and 0.811 at station).
Finally, the latent variable representing Personnel characteristics is best explained
by competence and kindness of the personnel on board (presenting coefficient
values equal to 0.941 and 0.923, respectively).

4 Conclusions

The weights that passengers give to the service attributes, and the measure of their
satisfaction with them help in improving service quality and preparing better
investment plans; this goal is ever more important nowadays because of the growing
worldwide tendency for cost reduction. In the literature, there are many studies
analyzing transit service quality based on users’ perceptions in terms of satisfaction
judgments. Some of them adopt SEM methodology, but no research was about
railways services in Italian context. Specifically, in this work, a railway service was
analysed, and a detailed and relevant number of service characteristics was
investigated.
The results suggest that SEM is a suitable methodology for establishing the
relationships between rail passengers’ satisfaction and service quality, and
identifying the key service factors that determine passengers’ satisfaction.
We found that macro-factors regarding information, cleanliness and service have
the highest positive effect on the overall service quality. This result agrees with the
importance rates directly stated by the users about the indicators explaining the
macro-factors. In fact, by calculating the average importance rate for each macro-
factor, we obtain very similar values, that is 8.5 for information, 8.7 for cleanliness,
and 8.8 for service. A contrasting finding concerns the attribute linked to safety,
which is judged by the users as the most important factor (9.1) while the model
results suggest that it is the least important one. According to the authors’
knowledge, this is a not surprising result because passengers incline to express the
highest rates of importance to the indicators concerning safety and security although
the number of accidents and crimes on board and at station is very low; however,
passengers consider these aspects as particularly important because they interest
users’ personal integrity. We retain that model results are more reliable in providing
for the weights of the various factors, being calculated starting from the correlation
of the satisfaction rates of the factors with the overall satisfaction. This last one
represents a calculated importance and it is preferred by researchers and academics
because of their numerous advantages, such as a more reliable evaluation of the
users, who otherwise tend to indifferently give importance to all the attributes if
they have to directly state a rate of importance (de Ona et al. 2013).
In addition, SEM permits to determine for each service factor the indicators (or
observed variables) better explaining it. Each latent service factor is better explained
by some observed attributes. As an example, safety is best explained by personal
security on board, cleanliness is prevalently understood as degree of cleanliness of

123
Relationships between rail passengers’ satisfaction and service quality

vehicles and seats, while comfort as air-conditioning and degree of comfort on

board. In this way, we can understand which service attributes are mostly
considered by passengers, and which service factors mostly affect the overall
service quality.
In addition, the authors experimented the introduction of indicators explaining
the endogenous latent variable different from the traditionally used indicators.
Specifically, the introduction of CSI showed good findings. In fact, this is the
indicator best explaining the overall service quality; this finding is relevant because
CSI allows both importance and satisfaction rates to be considered. Being CSI
correlated with satisfaction rate and widely explaining the service quality, the
findings suggest that only CSI should be used as indicator of the latent construct.
Also the indicator linked to the critical events showed interesting results, explaining
20 % of the overall service quality. We can retain that interesting findings emerged
regarding the ways of measuring the latent endogenous variable representing the
overall service quality.
The framework obtained by analyzing passenger responses can be used for an
improvement program for similar railways services. This will provide a cost
effective solution to achieve higher passenger satisfaction and a tool for a better
planning fund allocation.

References

Arbuckle JL, Wothke W (1995) AMOS 4.0 User’s Guide. SmallWaters Corporation, Chicago
Bagozzi RP (1994) Structural equations models in marketing: basic principles. In: Bagozzi RP (ed)
Principles of marketing research. Blackwell, Cambridge, pp 317–385
Bamberg S, Schmidt P (1998) Changing travel-mode choice as rational choice: results from a longitudinal
intervention study. Ration Soc 10:223–252
Bass P, Donoso P, Munizaga M (2011) A model to assess public transport demand stability. Transp Res
Part A Policy Pract 45:755–764
Bollen KA (1989) Structural equations with latent variables. Wiley, New York
Cavana RY, Corbett LM, Lo YL (2007) Developing zones of tolerance for managing passenger rail
service quality. Int J Qual Reliab Manag 24:7–31
Cirillo C, Eboli L, Mazzulla G (2011) On the asymmetric user perception of transit service quality. Int J
Sustain Transport 5:216–232
de Abreu e Silva J, Goulias KG (2009) A structural equations model of land use patterns, location choice,
and travel behavior in Seattle and comparison with Lisbon. In: Proceedings of the 88th annual
transportation research board meeting. January 2009: pp 11–15, Washington, DC
de Abreu e Silva J, Morency C, Goulias KG (2012) Using structural equations modeling to unravel the
influence of land use patterns on travel behavior of workers in Montreal. Transport Res Part A
46:1252–1264
de Ona J, de Ona R, Calvo FJ (2012) A classification tree approach to identify key factors of transit
service quality. Expert Syst Appl 39:11164–11171
de Ona J, de Ona R, Eboli L, Mazzulla G (2013) Perceived service quality in bus transit service: a
structural equation approach. Transport Pol 29:219–226
dell’Olio L, Ibeas A, Cecı́n P (2010) Modelling user perception of bus transit quality. Transport Pol
17:388–397
dell’Olio L, Ibeas A, Cecı́n P (2011) The quality of service desired by public transport users. Transport
Pol 18:217–227

123
 L. Eboli, G. Mazzulla

Eboli L, Mazzulla G (2007) Service quality attributes affecting customer satisfaction for bus transit.
Journal of Public Transportation 10:21–34
Eboli L, Mazzulla G (2008a) A stated preference experiment for measuring service quality in public
transport. Transport Plan Techn 31:509–523
Eboli L, Mazzulla G (2008b) Willingness-to-pay of public transport users for improvement in service
quality. European Transport 38:107–118
Eboli L, Mazzulla G (2009) A new customer satisfaction index for evaluating transit service quality.
Journal of Public Transportation 12:21–37
Eboli L, Mazzulla G (2010) How to capture the passengers’ point of view on a transit service through
rating and choice options. Transport Rev 30:435–450
Eboli L, Mazzulla G (2011) A methodology for evaluating transit service quality based on subjective and
objective measures from the passenger’s point of view. Transport Pol 18:172–181
Eboli L, Mazzulla G (2012) Structural equation modelling for analysing passengers’ perceptions about
railway services. Procedia-Soc Behav Sci 54:96–106
Eboli L, Forciniti C, Mazzulla G (2012) Exploring land use and transport interaction through structural
equation modelling. Procedia-Soc Behav Sci 54:107–116
Fillone AM, Montalbo CM, Tiglao NC (2005) Assessing urban travel: a structural equations modeling
(SEM) approach. Proc East Asia Soc Transp Stud 5:1050–1064
Golob TF (2003) Structural equation modeling. In: Goulias KG (ed) Transportation systems planning.
CRC Press, Boca Raton, pp 1–23
Hensher DA, Prioni P (2002) A service quality index for area-wide contract performance assessment.
J Transp Econ Policy 36:9–113
Hensher DA, Stopher P, Bullock P (2003) Service quality-developing a service quality index in the
provision of commercial bus contracts. Transport Res A 37:499–517
Hill N, Brierley G, MacDougall R (2003) How to measure customer satisfaction. Gower Publishing,
Hampshire
Imaz A, Habib KMN, Shalaby A, Idris A (2014) Investigating the factors affecting transit users’ loyalty.
Public Transport. doi:10.1007/s12469-014-0088-x
Irfan Syed M, Mui Hung Kee D, Saman S (2011) Service quality in rail transport of Pakistan: A Passenger
Perspective. In: Proceedings of 3rd SAICON: International Conference on Management, Business
Ethics and Economics (ICMBEE) December 28–29, 2011. Lahore, Pakistan
Jen W, Tu R, Lu T (2011) Managing passenger behavioral intention: an integrated framework for service
quality, satisfaction, perceived value, and switching barriers. Transport 38:321–342
Joewono TB, Kubota H (2007) User perception of private paratransit operation in Indonesia. J Public
Transp 10:99–118
Joreskog KG (1973) Analysis of covariance structures. Multivariate Analysis-III ed. In: Krishnaiah PR
(ed). Academic Press, New York, p 263–285
Lai W-T, Chen C-F (2011) Behavioral intentions of public transit passengers—the roles of service
quality, perceived value, satisfaction and involvement. Transp Policy 18:318–325
Ngatia GJ, Okamura T, Nakamura F (2010) The structure of users’ satisfaction on urban public transport
service in a developing country: the case of Nairobi. J East Asia Soc Transp Stud 8:1288–1300
Nurul-Habib KM, Kattan L, Islaam T (2011) Model of personal attitudes towards transit service quality.
J Adv Transport 45:271–285
Pakdil F, Aydin Ö (2007) Expectations and perceptions in airline services: an analysis using weighted
SERVQUAL scores. J Air Transport Manag 13:229–237
Parasuraman A, Zeithaml V, Berry L (1988) SERVQUAL: a multiple item scale for measuring consumer
perceptions of service quality. J Retail 64:12–40
Tam ML, Tam ML, Lam WHK (2005) Analysis of airport access mode choice: a case study in Hong
Kong. J East Asia Soc Transport Stud 6:708–723
TRB (Transportation Research Board) (1999) A handbook for measuring customer satisfaction and
service quality. Transit Cooperative Research Program, Report 47, National Academy Press.
Washington, DC
TRB (Transportation Research Board) (2004). Transit capacity and quality of service manual, second edn.
Transit cooperative research program, Report 100, National Academy Press, Washington, DC
Trépanier M, Habib KMN, Morency C (2012) Are transit users loyal? Revelations from a hazard model
based on smart card data. Can J Civ Eng 39:610–618

123
Relationships between rail passengers’ satisfaction and service quality

Tschopp M, Axhausen KW (2007) Transport infrastructure and spatial development in Switzerland

between 1950 and 2000. In: Proceedings of 86th annual meeting of the transportation research
board, January 2007. Washington, DC
Van Acker V, Witlox F, Van Wee B (2007) The effects of the land use system on travel behavior: a
structural equation modeling approach. Transport Plan Techn 30:331–353
Washington SP, Karlaftis MG, Mannering FL (2003) Statistical and econometric methods for
transportation data analysis. Chapman & Hall/CRC Press, Florida
Webb V (2010) Customer loyalty in the public transportation context. Massachusetts Institute of
Technology
Weinstein A (2000) Customer satisfaction among transit riders. How customer rank the relative
importance of various service attributes. Transp Res Rec 1735:123–132
Wen C-H, Lan L, Cheng H-L (2005) Structural equation modeling to determine passenger loyalty toward
intercity bus services. Transp Res Rec 1927:249–255
Wiley DE (1973) The identification problem for structural equation models with unmeasured variables.
In: Goldberger AS, Ducan OD (eds) Structural equation models in the social science. Seminar Press,
New York, pp 69–83

123