A New Customer Satisfaction Index for Evaluating Transit Service Quality

A New Customer Satisfaction Index for Evaluating Transit Service Quality
A New Customer Satisfaction Index
for Evaluating Transit Service
Quality
Laura Eboli and Gabriella Mazzulla
University of Calabria, Italy
Abstract
In this paper, an index based on customer perspective is proposed for evaluating
transit service quality. The index, named Heterogeneous Customer Satisfaction Index,
is inspired by the traditional Customer Satisfaction Index, but takes into account the
heterogeneity among the user judgments about the different service aspects. The
index allows service quality to be monitored, the causes generating customer satis-
faction/dissatisfaction to be identified, and the strategies for improving the service
quality to be defined. The proposed methodologies show some advantages compared
to the others adopted for measuring service quality, because it can be easily applied
by the transit operators.
Introduction
Transit service quality is an aspect markedly influencing travel user choices. Cus-
tomers who have a good experience with transit will probably use transit services
again, while customers who experience problems with transit may not use transit
services the next time. For this reason, improving service quality is important for
customizing habitual travellers and for attracting new users. Moreover, the need
for supplying services characterized by high levels of quality guarantees compe-
tition among transit agencies, and, consequently, the user takes advantage of
21

---

Journal of Public Transportation, Vol. 12, No. 3, 2009
better services. To achieve these goals, transit agencies must measure their per-
formance.
Customer satisfaction represents a measure of company performance according
to customer needs (Hill et al. 2003); therefore, the measure of customer satisfac-
tion provides a service quality measure. Customers express their points of view
about the services by providing judgments on some service aspects by means of
ad hoc experimental sample surveys, known in the literature as “customer satisfac-
tion surveys.”
The aspects generally describing transit services can be distinguished into the
characteristics that more properly describe the service (e.g., service frequency),
and less easily measurable characteristics that depend more on customer tastes
(e.g., comfort). In the literature, there are many studies about transit service qual-
ity. Examples of the most recent research are reported in TRB (2003a, 2003b), Eboli
and Mazzulla (2007), Tyrinopoulos and Antoniou (2008), Iseki and Taylor (2008),
and Joewono and Kubota (2007). In these studies, different attributes determining
transit service quality are discussed; the main service aspects characterizing a tran-
sit service include service scheduling and reliability, service coverage, information,
comfort, cleanliness, and safety and security. Service scheduling can be defined
by service frequency (number of runs per hour or per day) and service time (time
during which the service is available). Service reliability concerns the regularity of
runs that are on schedule and on time; an unreliable service does not permit user
travel times to be optimized. Service coverage concerns service availability in the
space and is expressed through line path characteristics, number of stops, distance
between stops, and accessibility of stops. Information consists of indications about
departure and arrival scheduled times of the runs, boarding/alighting stop loca-
tion, ticket costs, and so on. Comfort refers to passenger personal comfort while
transit is used, including climate control, seat comfort, ride comfort including the
severity of acceleration and braking, odors, and vehicle noise. Cleanliness refers to
the internal and external cleanliness of vehicles and cleanliness of terminals and
stops. Safety concerns the possibility that users can be involved in an accident,
and security concerns personal security against crimes. Other service aspects char-
acterizing transit services concern fares, personnel appearance and helpfulness,
environmental protection, and customer services such ease of purchasing tickets
and administration of complaints.
The objective of this research is to provide a tool for measuring the overall transit
service quality, taking into account user judgments about different service aspects.
22

---

A New Customer Satisfaction Index for Evaluating Transit Service Quality
A synthetic index of overall satisfaction is proposed, which easily can be used by
transit agencies for monitoring service performance. In the next section, a critical
review of indexes for measuring service quality from a user perspective is made;
observations and remarks emerge from the comparison among the indexes analy-
sed. Because of the disadvantages of the indexes reported in the literature, a new
index is proposed. The proposed methodology is applied by using experimental
data collected by a customer satisfaction survey of passengers of a suburban tran-
sit service. The obtained results are discussed at the end of the paper.
Customer Satisfaction Indexes
The concept of customer satisfaction as a measure of perceived service quality was
introduced in market research. In this field, many customer satisfaction techniques
have been developed. The best known and most widely applied technique is the
ServQual method, proposed by Parasuraman et al. (1985). The ServQual method
introduced the concept of customer satisfaction as a function of customer
expectations (what customers expect from the service) and perceptions (what
customers receive). The method was developed to assess customer perceptions of
service quality in retail and service organizations. In the method, 5 service quality
dimensions and 22 items for measuring service quality are defined. Service quality
dimensions are tangibles, reliability, responsiveness, assurance, and empathy. The
method is in the form of a questionnaire that uses a Likert scale on seven levels of
agreement/disagreement (from “strongly disagree” to “strongly agree”).
ServQual provides an index calculated through the difference between perception
and expectation rates expressed for the items, weighted as a function of the five
service quality dimensions embedding the items. Some variations of this method
were introduced in subsequent years. For example, Cronin and Taylor (1994) intro-
duced the ServPerf method, and Teas (1993) proposed a model named Normed
Quality (NQ). Although ServQual represents the most widely adopted method
for measuring service quality, the adopted scale of measurement for capturing
customer judgments has some disadvantages in obtaining an overall numerical
measure of service quality; in fact, to calculate an index, the analyst is forced to
assign a numerical code to each level of judgment. In this way, equidistant num-
bers are assigned to each qualitative point of the scale; this operation presumes
that the distances between two consecutive levels of judgment expressed by the
customers have the same size.
23

---

Journal of Public Transportation, Vol. 12, No. 3, 2009
A number of both national and international indexes also based on customer per-
ceptions and expectations have been introduced in the last decade. For the most
part, these satisfaction indexes are embedded within a system of cause-and-effect
relationships or satisfaction models. The models also contain latent or unobserv-
able variables and provide a reliable satisfaction index (Johnson et al. 2001). The
Swedish Customer Satisfaction Barometer (SCSB) was established in 1989 and is
the first national customer satisfaction index for domestically purchased and con-
sumed products and services (Fornell 1992). The American Customer Satisfaction
Index (ACSI) was introduced in the fall of 1994 (Fornell et al. 1996). The Norwegian
Customer Satisfaction Barometer (NCSB) was introduced in 1996 (Andreassen
and Lervik 1999; Andreassen and Lindestad 1998). The most recent development
among these indexes is the European Customer Satisfaction Index (ECSI) (Eklof
2000). The original SCSB model is based on customer perceptions and expecta-
tions regarding products or services. All the other models are based on the same
concepts, but they differ from the original regarding the variables considered
and the cause-and-effect relationships introduced. The models from which these
indexes are derived have a very complex structure. In addition, model coefficient
estimation needs of large quantities of experimental data and the calibration pro-
cedure are not easily workable. For this reason, this method is not very usable by
transit agencies, particularly for monitoring service quality.
More recently, an index based on discrete choice models and random utility the-
ory has been introduced. The index, named Service Quality Index (SQI), is calcu-
lated by the utility function of a choice alternative representing a service (Hensher
and Prioni 2002). The user makes a choice between the service habitually used and
hypothetical services. Hypothetical services are defined through Stated Preferences
(SP) techniques by varying the level of quality of aspects characterizing the service.
Habitual service is described by the user by assigning a value to each service aspect.
The design of this type of SP experiments is generally very complex; an example
of an SP experimental design was introduced by Eboli and Mazzulla (2008a). SQI
was firstly calculated by a Multinomial Logit model to evaluate the level of qual-
ity of transit services. Hierarchical Logit models were introduced for calculating
SQI by Hensher et al. (2003) and Marcucci and Gatta (2007). Mixed Logit models
were introduced by Hensher (2001) and Eboli and Mazzulla (2008b). SQI includes,
indirectly, the concept of satisfaction as a function of customer expectations and
perceptions. The calculation of the indexes following approaches different from
SQI presumes the use of customer judgments in terms of rating. To the contrary,
SQI is based on choice data; nevertheless, by choosing a service, the user indirectly
24

---

A New Customer Satisfaction Index for Evaluating Transit Service Quality
expresses a judgment of importance on the service aspects defining the services.
In addition, the user expresses a judgment of satisfaction about the service aspects
when he/she describes the service habitual y used. Also, SQI is calculated by a
very complex procedure. Choice data can give more reliable results because the
user must make a choice and makes a simultaneous comparison of all the service
attributes; to the contrary, the evaluation of the attributes by rating general y
influence the user to assign a high level of importance to each service attribute,
and the user evaluates each attribute one by one. Nevertheless, SQI has some dis-
advantages because choice data are not usual for customer satisfaction surveys; in
addition, this type of data must be collected by well-designed SP experiments.
A more direct measure for service quality evaluation is provided by an overall
index, often called “Customer Satisfaction Index” (CSI) (Hill et al. 2003). CSI repre-
sents a measure of service quality on the basis of the user/consumer perceptions
on service aspects expressed in terms of importance rates, compared with user/
consumer expectations expressed in terms of satisfaction rates. CSI plugs the gap
of ServQual because is based on judgments expressed according to a numerical
scale. Compared to all the described indexes, CSI is based on a simple procedure,
fully described in the next section, which allows the index to be easily calculated0
by transit operators.
CSI does not take into account the heterogeneities among user judgments. To
the contrary, the index proposed by the authors provides an overall service qual-
ity measure introducing the dispersion of the importance and satisfaction rates
among users.
Methodology
The methodology adopted in this research aims to obtain a concise indicator that
provides an overall measure of service quality by considering different service
aspects. The indicator can be calculated on the basis of user judgments expressed
by a numerical scale; this kind of scale has some advantages compared to the scales
with points described by means of words (e.g., Likert and verbal scale) because it
allows quantitative techniques of analysis to be applied. To measure customer
satisfaction, different numerical values can be used, generally from 1 to 3, from 1
to 5, from 1 to 7, from 1 to 9, etc. The adopted scale can also have an even number
of levels, for example, the traditional numeric scholastic scale composed of points
from 1 to 10.
25

---

Journal of Public Transportation, Vol. 12, No. 3, 2009
As mentioned above, this research focuses on CSI, which is calculated by means of
the satisfaction rates expressed by users, weighted on the basis of the importance
rates, according to the following formula:


(1)
in which
is the mean of the satisfaction rates expressed by users on the service qual-
ity k attribute
(importance weight) is a weight of the k attribute, calculated on the basis
of the importance rates expressed by users. Specifically, is the ratio between
the mean of the importance rates expressed by users on the k attribute and
the sum of the average importance rates of all the service quality attributes:



(2)
CSI represents a good measure of overall satisfaction because it summarizes the
judgments expressed by users about various service attributes in a single score. The
more accurate the selection of the attributes, the more accurate the measure of
the overall satisfaction. For this reason, the selected attributes should describe the
service aspects exhaustively.
However, not all the attributes are important for the user in the same way; an
index based only on satisfaction rates cannot take into account these differences.
As an example, we consider five attributes with average satisfaction and impor-
tance rates reported in Table 1, according to a scale from 1 to 10. By considering
only the satisfaction rates, the overall satisfaction is 7.16, and the attribute with
the highest satisfaction score is attribute 2, which contributes to the overall sat-
isfaction with an aliquot of 1.66; on the other hand, if importance rates also are
considered, the attribute with the highest aliquot to the overall satisfaction is the
attribute 4 (weighted score equal to 1.94). The less important attribute is attribute
5, with an aliquot of 1.05. The value of CSI is 7.28 out of 10. By converting this score
26

---

A New Customer Satisfaction Index for Evaluating Transit Service Quality
into a percentage, the satisfaction index shows that the service is about 73 percent
successful in satisfying its customers. By comparing CSI with the average of all the
satisfaction scores, it can be observed that there is a difference between the value
of these two indicators, because each attribute adds up to overal satisfaction
according to a different weighted score.
Table 1. Example of Calculating CSI (Scale of 1 to 10)


Importance
Importance
Satisfaction
Weighted
Attribute
Score
Weight
Score
Score

1
7.1
0.18
6.5
1.17

2
9.2
0.23
8.3
1.91

3
7.3
0.18
6.7
1.21

4
9.5
0.24
8.1
1.94

5
6.9
0.17
6.2
1.05
Total
40.0


7.28
However, when all the importance scores are close to a certain value, the impor-
tance weights are similar, and then the CSI value is close to the average of all the
satisfaction scores. In this eventuality, CSI does not give any additional information
compared to the indicator calculated by considering only the satisfaction scores.
In addition, the average importance scores result from the rates expressed by a
sample of customers, which can be very heterogeneous; the dispersion of the rates
can be represented by the variance or the standard deviation from the mean. In
the same way, the satisfaction rates can be very heterogeneous among users. These
heterogeneities cannot be taken into account in the CSI calculation.
To overcome this lack, importance weights can be corrected according to the dis-
persion of the importance rates from the average value. Analogously, satisfaction
scores can be corrected according to the dispersion of the satisfaction rates from
the average value. These adjustments have been introduced for calculating a new
indicator, called Heterogeneous Customer Satisfaction Index (HCSI). The differ-
ences between CSI and HCSI are shown in Figure 1.
27

---

Journal of Public Transportation, Vol. 12, No. 3, 2009

Figure 1. CS Index versus Heterogeneous CS Index
From a mathematical point of view, HCSI is calculated by the following formula:


(3)
in which
is the mean of the satisfaction rates expressed by users on the k attribute
corrected according to the deviation of the rates from the average value
is the weight of the k attribute, calculated on the basis of the importance
rates expressed by users, corrected according to the dispersion of the rates
from the average value.
is calculated by the following formula:


(4)
The adjustment factor is calculated as the mean of the satisfaction rates expressed
by users on the k attribute divided by the mean of the average satisfaction rates
28

---

A New Customer Satisfaction Index for Evaluating Transit Service Quality
of all the service quality attributes, weighted on the variance of the satisfaction
rates.
is calculated as the mean of the importance rates expressed by users on the
k attribute divided by the sum of the average importance rates of all the service
quality attributes, weighted on the variance of the importance rates, according to
the following formula:


(5)
The introduction of the variance for adjusting the importance and satisfaction
rates allows the attributes characterized by more homogeneous user judgments to
be considered more significant; to the contrary, the attributes with heterogeneous
judgments are considered less significant.
The mathematical basis of the HCSI formula is demonstrated by assuming that all
the customers surveyed gave satisfaction scores of 10 out of 10 for every service
characteristic, and the average satisfaction scores would all be 10. When the vari-
ance of the satisfaction judgments expressed by the customers tends to zero for all
service characteristics, the mean of the satisfaction rates divided by the deviation
from the mean of each k attribute would tend to the maximum value of 10, and
would tend to . Therefore, total customer satisfaction on all their attributes
would produce a satisfaction index of 100 percent.
Application of Methodology
The proposed methodology was applied by considering an experimental case
study regarding transit services in a medium-sized urban area. The urban area
includes the town of Cosenza, which is a provincial capital of the Calabria region
in southern Italy. Cosenza forms a single built-up area with the town of Rende,
in a northerly direction. The urban area has grown over the years also because of
the presence of the University of Calabria, which expanded north of Rende at the
beginning of the 1970s. Cosenza and Rende represent a center of attraction for the
province because of the administrative functions, job opportunities, and supply of
services. The urban area has about 110,000 inhabitants. In addition, many univer-
29

---

Journal of Public Transportation, Vol. 12, No. 3, 2009
sity students live in Rende or Cosenza; approximately 35,000 students attend the
University of Calabria.
The analysed transit service is a suburban bus service offering the connection
between the urban area and several small villages north and south of Cosenza.
A survey was addressed to the habitual passengers of two bus lines, Line 17 and
Line 1, to measure transit service quality from a user point of view. Line 17 runs
in a southward direction and serves a catchment area of about 5,000 inhabitants;
Line 1 runs in a northward direction and serves a catchment area of about 7,000
inhabitants. Bus line characteristics are reported in Table 2.
Table 2. Transit Service Characteristics
Service
Characteristics
Line 1

Line 17
Path length
19 km

18 km
# of bus stops
23

13
Travel demand
800 pass/day

700 pass/day
Service time
14 hours (from 6:00 a.m. to 8:00 p.m.)
Service frequency 1 run/hr from 6:00 a.m to 2:00 p.m.; only 2 runs in the afternoon
Ticket cost from 0.50 to 1.50 Euros (depending on the covered distance)
The survey was conducted in the spring of 2008. An operator effected face-to-face interviews on
board during the service time; 218 passengers were interviewed.
Although the population is evenly spread between male and female, the major-
ity of the habitual transit users is female (66% of the sample). Most of the inter-
viewed users are students (49%) and younger than 20 years (44%); only 9% of the
population are students, and 22% are young people. The majority of the employed
respondents are clerks or workers (92%) and work in the private or public sec-
tor (71%); these percentages are the same for the population. About 65% of the
sample belongs to a middle class of family income and about 28% to a lower class;
the classes of income refer to the net monthly income of the family unit, expressed
in Euros (Table 3).
On average, the number of family members in a family unit is 3.8 and each family
has 1.64 cars. Of the 218 respondents, 77 get one-way tickets, 64 get one-day travel
cards, and 69 use monthly travel cards.
30

---