Review: attached doc and following linksNam, S, Ha, C, Lee, H. (2018). Redesigning in-flight service with service blueprint based on text analysis. (Links to an external site.) MDPI.com, Sustainability|an Open Access Journal. 10 (12):4492.Next, select an airline and research information pertaining to the services provided and customer reviews. There are many online resources to aid in your search effort, including but not limited to:Airline Quality Rating (AQR) (ERAU Scholarly Commons) (Links to an external site.)A-Z Airline Reviews (SkyTrax) (Links to an external site.)Airlines Traveler Rating (TripAdvisor) (Links to an external site.)Develop a comprehensive report of the airline-passenger service encounters (as identified by Nam and Lee, 2018) and propose recommendations for improving service quality. Review the following report requirements:At a minimum, your report should include the following topics:Identify the airline’s position (as a legacy or low-cost carrier) and describe how the service mix caters to their target market.Evaluate the eight service encounters based on customer reviews and secondary research.Explain the importance of innovative in-flight services.Describe the significance of service quality and how it can be measured.Comparisons of service quality perceptions
between full service carriers and low cost
carriers in airline travel
We apply latent Dirichlet allocation topic modeling to a vast number of
passenger-authored online reviews for airline services to compare service quality
between full service carriers (FSCs) and low cost carriers (LCCs). Representing
key features of airline service quality, topics are extracted from the reviews and
matched to the five typical dimensions used by the SERVQUAL model. Based on
the measure of word frequency statistically distributed to topics, we quantitatively
determine the dimensions of service quality that are deemed as most essential
by travelers. The results show that the most significant dimensions for FSCs and
LCCs are tangibles and reliability, respectively. The least significant dimensions
are assurance and empathy, respectively. By comparing extracted features in
detail, we discover specific differences in traveler perceptions between FSCs and
LCCs. Air carriers should be aware of these differences, as it would help them
better differentiate themselves. Moreover, inflight meal services and seats, which
have typically been regarded as tangible features, are subdivided into different
topics, and the subdivisions are simultaneously matched to multiple dimensions
(eg tangibles, empathy, and reliability). This suggests that research needs to
reflect the diverse aspects of traveler perceptions for primary service items.
KEYWORDS:
Airline serviceairline travelerslatent Dirichlet allocationonline reviewservice quality featuretext analysis
Previous articleView issue table of contentsNext article
Introduction
Competition between low cost carriers (LCCs) and full service carriers (FSCs)
has intensified in the global air travel market (Han & Hwang, 2017; O’Connell &
Williams, 2005). For example, LCCs, relatively recently introduced in Asian and
emerging air travel markets, are gradually increasing their market share while
concentrating on cost reduction strategies to capture cost-sensitive travelers
(Baum & Kua, 2004; Martinez-Garcia & Royo-Vela, 2010; O’Connell &
Williams, 2005). To respond to the challenges from LCCs, FSCs are strategically
focusing on their hub airports, a strategy that runs counter to the point-to-point
strategy used by LCCs. FSCs are also providing higher levels of service quality
and strengthening their alliances to retain their loyal customers and avoid
customer switching behavior (Dennis, 2007, 2010). In such a competitive
environment, increased importance has been placed on acquiring a better
understanding of the key differences in perceived service quality between LCC
and FSC customers to differentiate service strategies and achieve business
sustainability (Koklic, Kukar-Kinney, & Vegelj, 2017; Lee et al., 2018).
To measure service quality, researchers rely mainly on surveys that are designed
using the existing literature. Using surveys, many studies have explored
customer perceptions regarding the service quality of airlines and showed
significant differences in perceptions between LCCs and FSCs (Ahn &
Lee, 2011; Chiou & Chen, 2010; Curras-Perez & Sanchez-Garcia, 2016; Koklic
et al., 2017; O’Connell & Williams, 2005; Rajaguru, 2016). Despite the utility of
being able to draw upon standardized research designs, circumstances have
highlighted typical drawbacks – the amount of time needed to collect complete
datasets (Kothari, 2004), the restricted expandability of research (Lee &
Bradlow, 2011), sample size limitations (Bartlett, Kotrlik, & Higgins, 2001) and so
forth. Online reviews have the exact the opposite disadvantage – the absence of
a standardized research process. Compared to survey data, they can serve as
objects of exploratory research and reveal new aspects of service quality.
Nevertheless, because online reviews are regarded as one of the most critical
factors in customer purchase decisions (Archak, Ghose, & Ipeirotis, 2011; Duan,
Gu, & Whinston, 2008; Godes & Mayzlin, 2004), they have lately received
considerable attention in a number of business research areas for numerous
reasons, including the following. First, they have sizable volumes. The inferencederived results can be considered reliable when drawing the results from trusted
review sites with very large datasets. Second, online reviews preserve the realtime perceptions held by customers (Dellarocas, Zhang, & Awad, 2007; Duan
et al., 2008). They are one of the most immediate measures of service
experience. Finally, they show voluntary, unrefined, and direct experience or
feedback from customers (Mudambi & Schuff, 2010).
In this study, we apply latent Dirichlet allocation (LDA) topic modeling, a widelyused text analysis technique, to a vast number of passenger-written online
reviews for airline services to analyze and compare service quality (Blei &
Lafferty, 2007; Blei, Ng, & Jordan, 2003). Representing key features of airline
service quality, the topics discerned by the modeling are (with the help of
academic researchers) matched to the five traditional dimensions –
responsibility, assurance, tangibles, empathy, and responsiveness – employed
by the SERVQUAL model to determine experience-based service quality. This
enables us to quantitatively determine the dimensions of service quality that
travelers deem most essential based on the measure of word frequency
statistically distributed across topics. A major incentive to employ the five
dimensions is that it is the most widely accepted measure of service quality, and
it is thus easy to compare the current results to previous results. To minimize the
validity issue with the five dimensions, we thoroughly review varied versions of
newly-developed and different dimensions for airline service quality in the
following section. This helps us understand how dimensions should be defined
and which dimension type is required to incorporate airline service-specific
characteristics. Accordingly, it is possible to match topics to the dimensions as
appropriately as possible while integrating domain-specific characteristics that
varied models have proposed. We also carry out a sentiment analysis to uncover
customer emotions or attitudes regarding the quality of airline services (Liu &
Zhang, 2012). In summary, the aim of this study is to answer the following
research questions (RQs).
RQ1. Can service features of airline service quality be extracted from online
reviews and be properly represented in a service quality model?
RQ2. How can the significance of service features be quantified?
RQ3. What are the differences between FSCs and LCCs in terms of SERVQUAL
dimensions and features?
RQ4. Compared to previous studies using surveys, what are the new aspects of
the customers’ perceptions of airline service quality?
RQ5. What are the customers’ sentiments about service quality?
By answering these questions, this study offers meaningful insights for air
carriers with respect to providing differentiated services to travelers. Likewise, it
provides understandings for researchers attempting to determine the types of
features that should be considered or added when designing studies on airline
service quality.
Literature review
Airline service quality models
Many studies on airline service quality have worked with SERVQUAL, which was
postulated by Parasuraman, Zeithaml, and Berry (PZB) (1988), and its variations
that reorganize the structure of dimensions to improve model validity by adopting
domain-specific characteristics. Fick and Brent Ritchie (1991) and Gourdin and
Kloppenborg (1991) studied service quality in the air transport industry using
PZB’s model. Fick and Brent Ritchie (1991) measured service quality for four
kinds of businesses, including airline service, but they could not measure the
relative effects of the SERVQUAL items (Young, Cunningham, & Lee, 1994).
Using PZB’s model of service quality developed in 1985, Gourdin and
Kloppenborg (1991) surveyed customers, airline employees, and officials in the
US Department of Transportation and the Federal Aviation Agency. While they
showed statistical significance in several variables, their approach was not
complete in terms of sample representativeness, variable origin, and model
reference (Young et al., 1994).
Tsaur, Chang, and Yen (2002) implemented the fuzzy set theory to resolve the
briefness issue of the Likert scale for service quality measurement. Using the
fuzzy approach, they tried to measure vague human judgements such as
customer satisfaction in a more explicit manner. Their study concluded that the
most and least important dimensions were tangibles and empathy, respectively,
among the five dimensions in PZB’s SERVQUAL. Gilbert and Wong (2003)
measured airline service quality by modifying SERVQUAL’s original form. The
tangible dimension was subdivided into facilities, employees, and flight patterns,
and the empathy dimension was renamed as customization. Assurance was
considered to be most important, whereas customization and facilities were not
important. They also found that service expectations varied in different market
segments by showing statistical differences depending on ethnic groups,
nationalities, and travel purposes. Park, Robertson, and Wu (2005) proposed the
application of structural equation modeling to test simultaneous relationships.
However, the applicability of their research results was restricted because their
data only represented international economy class travelers.
Studies on LCCs have only appeared recently, and there are thus fewer of them
than studies on FSCs. Saha and Theingi (2009) found that service quality was
still a determinant of customer satisfaction in LCCs, and behavioral intentions
such as repurchase intentions and feedback were affected by service quality and
customer satisfaction. They also showed that customer satisfaction and feedback
were positively correlated. Chiou and Chen (2010) adopted the research frame
provided by Park, Robertson, and Wu (2004) to investigate factors that affected
the behavioral intentions of travelers between FSCs and LCCs. Their study
showed that service perception had a considerable influence on the behavioral
intentions of FSC travelers whereas service value had a large influence on those
of LCC travelers. These contrasting results suggested that there existed a
nontrivial gap between customer perspectives for FSC and LCC airline services.
In particular, price might have been much more critical for LCC passengers than
for FSC passengers. In Table 1, we summarize selected published results that
have used the conventional five dimensions or variations thereof. We retain the
meanings of the original or modified dimensions in the selected literature when
matching topics to the five dimensions.
Table 1. Selected studies for airline service quality. We finally select 13
results from 21 papers with five dimensions or varied dimensions
employed among the total of 45 papers reviewed for airline service
quality. Due to the lack of space, features of the dimensions are displayed
in Appendix A (in Supplementary Material). The dimensions used by
Aksoy et al. (2003) are only for domestic airline service. The dimensions
for international airline service are included in Appendix A.
CSVDisplay Table
Text analysis-based service quality in airline travel
With the growth of mobile web platforms, online reviews have become one of the
most popular methods of customer assessment for overall service quality (Lee &
Lin, 2005; Mudambi & Schuff, 2010; Palese & Piccoli, 2016). Reviews are mainly
composed of comments reflecting direct perceptions of service performance and
experience (Guo, Barnes, & Jia, 2017; Humphreys & Wang, 2017; Miguéis &
Nóvoa, 2017). Many studies have analyzed reviews for products and service
areas such as tourism and hotel businesses (Archak et al., 2011; Berezina,
Bilgihan, Cobanoglu, & Okumus, 2016; Mankad, Han, Goh, & Gavirneni, 2016;
O’connor, 2010). Recently, several studies in the air travel area have used online
text data to investigate customer perceptions (Gitto & Mancuso, 2017, 2019; Lee
& Yu, 2018; Martin-Domingo, Martín, & Mandsberg, 2019; Misopoulos, Mitic,
Kapoulas, & Karapiperis, 2014; Yee Liau & Pei Tan, 2014).
Misopoulos et al. (2014) utilized 67,953 Tweets to identify important customer
service factors. They produced relevance ratings of Tweet messages based on
the similarity coefficient, and they analyzed customer sentiments to investigate
opinions regarding airline service. They found that services related to flight
delays, lost baggage, and check-in/boarding problems caused negative
sentiments, while those related to check-in in mobile applications, reasonable
prices, and on-board entertainment generated positive sentiments. However,
their study was limited in that they only analyzed 20 keywords in the dataset.
Similarly, Yee Liau and Pei Tan (2014) analyzed 10,895 Tweets to study
customer opinions about LCCs in Malaysia. They employed a k-means clustering
algorithm to group Tweets and spherical k-means clustering to enhance the
efficiency of the analysis. The results reported that clusters of customer service,
booking management and ticket promotions collected more positive emotions.
On the contrary, the flight cancelation cluster acquired more negative sentiments.
Gitto and Mancuso (2017) worked with online reviews collected from five major
European airports on the Skytrax website. They analyzed 895 sentences, two
third of which were related to non-aviation services and one third of which were
aviation services. They found that slightly more than half (55%) of the sentiments
were positive in the non-aviation services, while one third (33%) of them were
positive in the aviation services. Related to the non-aviation services, the most
frequent opinions referred to food and beverages and shop service. On the
contrary, check-in and baggage claim services were most frequently addressed
in the aviation services. Lee and Yu (2018) investigated Google reviews for the
top 100 airports to show that online reviews could be used to measure airport
service quality (ASQ). They found that the sentiment scores of reviews
adequately predicted Google star ratings. They also demonstrated that sentiment
scores and Google star ratings had a sizable relationship with ASQ ratings.
Furthermore, they revealed that 25 topics extracted from the LDA analysis were
well matched to the ASQ service attributes. They proposed a future study on a
relative importance investigation for attributes of different groups such as FSCs
versus LCCs, which could be one of the results of the present study.
Gitto and Mancuso (2019) used the Twitter accounts of 118 airports to determine
the brand perceptions of airports based on attributes of the airport industry,
including environment, disability, and luxury. Using a cluster analysis and social
perception scores, they explained the passengers’ clustered perceptions of
airports. Martin-Domingo, Martín, and Mandsberg (2019) attempted to measure
ASQ using sentiment analysis with a dataset of 4,392 Tweets. They determined
23 service attributes composed of 108 keywords and compared them to 34
attributes of ASQ. The research results revealed that passengers frequently
mentioned attributes regarding waiting and ground transport, but they mentioned
shopping or washrooms (WC) only in 1% of their Tweets. In the sentiment
analysis, customers had a positive attitude regarding WiFi, WC, food and
beverages, and lounge services, while they showed negative sentiments
regarding waiting, parking, arrival, staff, and passport control.
Research model
With respect to analyzing the meaning of the words and content in the
documents in the topic model, the topic model assumes that a topic is a
probability distribution of words, and a document comprises a mixture of topics
(Steyvers & Griffiths, 2007). LDA is the most common topic model (Blei
et al., 2003). It generates topics that have been latent in documents based on the
Dirichlet distribution. LDA can be easily implemented via software (eg R, Matlab
and Python) after model input parameters such as the number of topics (=k) are
adequately set. (See details of data preprocessing and model parameter setting
in Appendix B.) As a result of the LDA modeling, a topic is the probability
distribution of words from online reviews that contain customer perceptions on
service experience, and it represents a reorganized form that expresses the
feature of service quality in the k-dimensional space. That is, a topic suggests a
feature of a specific dimension of service quality.
The research model is depicted in Figure 1. First, online reviews are collected via
web crawling. Second, the collected reviews are preprocessed to make them
suitable as input for the LDA modeling. Data is arranged into a document-term
matrix format, ie a matrix of a preprocessed corpus. Third, when extracting topics
from the online reviews, the LDA algorithm reduces the uncontrollable
dimensional space into a controllable k-dimensional space (30 topics for FSCs
and 20 topics for LCCs; see explanations in Appendix B). This makes the data
sufficiently manageable in the subsequent stage. Fourth, the extracted topics are
named and matched to the dimension regarded as the best fit among the five
choices (RQ1). This is achieved with the help of an advisory group. Namely, we
perform another dimension reduction (from 30 and 20 to five each for the FSCs
and LCCs, respectively) based on the group members’ survey and interview. The
group is composed of three professors and six graduate students whose
specialties cover diverse majors in aviation management, including airline
marketing, airport operations, airline service, human resources, finance, MIS,
and aviation policy and strategy. Next, we decide which dimension among the
five dimensions is the most significant by measuring word frequency (RQ2). All of
the stages are repeated twice for both the FSCs and LCCs. Then we compare
the differences between the service qualities (RQ3) and explain what aspects of
the research in airline service quality should be newly considered (RQ4). Finally,
a sentiment analysis is applied to topics and dimensions to investigate customer
attitudes based on the widely-used word dictionary (Hu & Liu, 2004) (RQ 5).
Figure 1. Research model.
Display full size
Data
We use online reviews from airlinequality.com in which travelers voluntarily and
individually write reviews of their service experiences. To compare, we choose
the top 10 of the world’s top 100 ranked airlines for both categories of service
carriers (airlinequality.com/review-pages/top-10-airlines/). Using a data crawling
package in R, we retrieve all of the online reviews of the selected airlines for both
LCC and FSC categories at the time of review gathering. Table 2 summarizes
the data employed in this study.
Table 2. Data summary.
CSVDisplay Table
We statistically examine whether collected online reviews are representative of
the population using the method offered by Aggarwal and Singh (2013). They
determined sample representativeness by conducting t-tests for critical variables
between sample and population groups. If there was no significant difference in
the critical variable between groups, it was concluded that the sample was
representative. We consider two critical variables for the LDA algorithm. One is
the number of words per review, and the other is the number of occurrences of
words per review (Blei et al., 2003; Wei & Croft, 2006). For the first t-test, the
average number of words per review is compared. When we collect the online
reviews (11,031) for 20 airlines, there are 52,506 reviews for 84 airlines,
including 23 LCC carriers. The test results are summarized in Table 3. For the
most frequent 1,000 words, which explain 82.6% of the total occurrences of all
words, the average number of occurrences per review is compared at the second
t-test. Because a lager number of reviews leads to a larger number of word
occurrences, we use statistics through dividing word occurrences by the number
of reviews in each group for the most frequent 1,000 words. Table 4 displays the
test outcomes, which demonstrate that there are no statistically significant
differences between the groups in terms of the critical variables in topic
modeling.
Table 3. t-test for the number of words.
CSVDisplay Table
Table 4. t-test for the average number of occurrences of 1,000 words that
are the most frequent in each group.
CSVDisplay Table
Topic modeling
Topic naming
The LDA modeling results in Appendix C reveal the extracted features of the
customer online reviews (ie topics), and are now composed of probabilistic
distributions of words. If we can give an appropriate name to every topic while
considering the meaning of words distributed to each topic, the contents of all of
the reviews will be expressed by interpretable topics. This naming process is
carried out in two steps. In the first step, we provide the topic modeling results
with a questionnaire form (Appendix C without topic names (second row), topic
significance (third row), and highlighting) to members of the advisory group, and
they fill in the empty name of each topic independently. This step is designed to
collect various interpretations from the modeling results. During this step, we
provide the subjects with two explanations. First, we inform them that a word with
a larger probability in a topic has more explanatory power than a word with a
smaller probability. Second, we ask them to focus on distinguishable words that
could represent differences among topics rather than similar words that
simultaneously exist in multiple topics. For example, the names for FSC Topic 5
(inflight meal (punctuality)) and FSC Topic 11 (inflight meal (menu variety)) are
more likely to be determined by words such as short, quick, and prior for Topic 5
and choic(e), cours(e), and option for Topic 11 rather than by words such
as serv(e), breakfast, and dinner that simultaneously exist in both topics. The
second step is needed only when there exists an obvious disagreement among
the collected names. Opinions from the professors are adjusted until reaching an
agreement, and then the agreed upon name is finalized via consensus from the
graduate students. If opposition still exists as the students finalize their work, the
adjusting process with the professors is repeated until all conditions are
satisfied. Figure 2 depicts the process.
Figure 2. Topic naming and matching process.
Display full size
Topic matching with five dimensions
After being named, the topics are matched to the five dimensions based on the
opinions of the advisory group. The matching process is conducted in a similar
manner to the naming process, but only the second step (in Figure 2) is carried
out. During the naming process, both broad interpretations as well as precise
interpretations are necessary from the experts. However, when matching,
precision alone is sufficient. We gather the reliable and valid opinions provided
by the professors from the initial process stage and concentrate on the
consensus within the advisory group while striving to integrate the airline servicespecific characteristics that we have outlined in the broad literature review. The
topic matching results with five dimensions are summarized in Table 5. Four of
the topics – customer satisfaction, recommendation intention, and model residual
for the FSCs; and price for the LCCs – are not assigned. They are excluded
because they are not regarded as typical dimensions of service quality. The
model residual indicates an uninterpretable topic driven by noisy data and is
used to enhance the coherence of the rest of the topics in the topic modeling
(DiMaggio, Nag, & Blei, 2013).
We use word frequency to quantify the significance of the topic (Yu, Zha, Wang,
& Chua, 2011; Zhang, Narayanan, & Choudhary, 2010). In terms of word
frequency, we assume that a topic is mentioned more frequently when it is more
significant. Thus, the more frequently that words are mentioned by travelers, the
more significant they are, and the more those words are contained in a particular
topic, the more significant the topic is. For example, the words seat, comfortable,
and pitch are more probable than the words meal, food, and water in the reviews
given by travelers who express that seat comfort is significant. The significance
of the kth topic (k = 1 … K) based on word frequency is calculated
asTk=∑Nn=1wn/k/∑Kk=1∑Nn=1wn/kTk=∑n=1Nwn/k/∑k=1K∑n=1Nw
n/k(1)wn/kwn/k: the frequency of the nth word in the kth topic
(n = 1 … N, k = 1 … K).
Note that the probability of the nth word in the kth topic (n = 1 … N, k = 1 … K) in
Appendix C is represented
by p(wn/k)=wn/k/∑Nn=1wn/kp(wn/k)=wn/k/∑n=1Nwn/k.
Table 5. Matching results in order of significance. The percentage value
in parenthesis indicates the dimension significance excluding topics not
matched. Highlighted topics are unique topics that simultaneously do not
belong to either type of air carrier.
Display Table
Result comparisons and discussions
Dimension level comparisons
Carrying out comparisons in the dimension level facilitates an overview of the
differences in traveler perceptions of service quality between FSCs and LCCs.
Tangibles and reliability are the most significant dimensions for FSCs and LCCs,
respectively. On the contrary, the least significant dimensions for FSCs and
LCCs are assurance and empathy, respectively. As we can see in Table 5, seatrelated topics, which have usually been considered as tangible features, are
prevalent in FSCs. In addition, they are principle contributors to the significance
of tangibles. For LCCs, at almost 50% (≈48%), the significance of reliability
dominates all other dimensions. This means that LCC customers are particularly
critical of how accurately promised services are performed, and they are less
focused on the performance of additional services (eg inflight entertainment
(IFE), wider seats, and faster staffing). Because of the low price of the service,
LCC travelers tend to have lowered levels of confidence in the service quality,
and this lowered confidence affects the significance of reliability (Bhadra, 2009;
Seo, Moon, & Lee, 2015; Wittman, 2014). Likewise, it can be understood that
empathy appears as the least significant dimension for LCCs. Assurance has a
relatively small level of significance for both service carriers. This is similar to the
results reached by Tsaur et al. (2002) whose study determined that assurance
was ranked fourth in terms of importance within the five dimensions for FSCs.
Kim and Lee (2011) also showed that assurance, including reliability and
empathy, did not significantly affect customer satisfaction for LCCs. These
comparisons indicate that there exists a considerable difference in passenger
perceptions between FSCs and LCCs. This conclusion was also reached by
O’Connell and Williams (2005).
Feature level comparisons
We can differentiate the service quality between FSCs and LCCs by examining
topics that simultaneously do not belong to either type of air carrier since those
topics represent the uniqueness of each category. When we exclude FSC topics
identical or similar to LCC topics, there exist 15 unique topics out of the 27 total
topics. The unique topics are highlighted in Table 5. Of note, frequent flyer
program (Topic 1), lounge service (Topic 14), and inflight entertainment (Topic
30) are topics exclusive to FSCs. These distinctive services can increase the
probability that passengers will choose a particular FSC (Baker, 2013). Topics of
employee quality (language skill) (Topic 10), sleep comfort (Topic 17), and
inflight entertainment (Topic 30) are also more relevant for FSCs because FSCs
are more likely to frequently cover long distances and fly international routes
(Fourie & Lubbe, 2006; Gillen & Morrison, 2003; Kappes & Merkert, 2013). The
fact that language skill appears for FSCs suggests that FSCs place greater focus
on international routes than LCCs (Kappes & Merkert, 2013). Sleep comfort also
represents a major feature of FSCs that usually operate long-distance flights
(Francis, Dennis, Ison, & Humphreys, 2007). Longer flights lead to longer
sleeping times. Therefore, as flight time increases, the need for more attentive
service to ensure sleep comfort also increases.
Interestingly, unique FSC topics are distributed in all of the five dimensions of
service quality. On the contrary, four of the five unique topics, out of 19 LCC
topics, are placed only in the reliability dimension. This indicates that FSCs
should not ignore any dimension of service quality, though tangibles is the most
critical dimension. On the other hand, LCCs need to focus on service items
related to the dimension of reliability for better differentiation. This is because
FSC travelers tend to recognize quality in various aspects of airline service
(Gillen & Morrison, 2003; Hunter, 2006; Zhang, Lin, & Newman, 2016). They
consider the entire range of services from basic to sophisticated, even including
the physical appearance of the aircraft (Topic 8). When the most basic and
sophisticated services are assumed to be related to reliability and empathy, the
unique topics of service consistency (Topic 26) and family seat request (Topic
12) are categorized into each dimension. The rest of the unique FSC topics are
related to seats and inflight meal services, which almost match with the tangibles
and empathy dimensions. These dimensions are closely related to the lower
price sensitivity of FSC customers when compared to LCC customers (O’Connell
& Williams, 2005; Wittman, 2014). FSC travelers tolerate higher expenses to
obtain greater benefits via various services (Seo et al., 2015).
Of the 19 LCC topics, five of them are unique. Among them, inflight meal
purchase (Topic 4), paid ancillary service (Topic 8), and carry-on baggage (Topic
20) are closely related to specific LCC properties. The inflight meal purchase
topic aptly reflects the fact that most LCCs charge fees for inflight meals
(O’Connell & Warnock-Smith, 2013). Words such as intern(et) and prebook in
this topic reveal a recent trend in the LCC business environment (Baker, 2013;
Bigné, Hernández, Ruiz, & Andreu, 2010; Hunter, 2006). The paid ancillary
service topic has been stressed as a major business strategy for LCCs and is an
important contributing factor for LCC profits (Doganis, 2006; O’Connell &
Warnock-Smith, 2013). This service incorporates seat selection, seat upgrades,
IFE, WiFi service, and inflight food and beverages, and related words such
as snacks, wifi, legroom, and drinks appear in the topic. LCCs need to
concentrate on services related to these words to achieve better differentiation.
The carry-on baggage topic also reflects the business nature of LCCs. Most
LCCs try to implement strict policies regarding checked baggage, including the
implementation of fees, to reduce related costs (O’Connell & WarnockSmith, 2013). To save on travel costs, LCC travelers tend to prefer using carryon baggage (Aldamari & Fagan, 2005).
Further divided features
Also of interest, the inflight meal service and seat topics are subdivided. For
FSCs, meal services are divided into four subtopics – inflight meal (punctuality)
(Topic 5), inflight drink service (Topic 7), inflight meal (menu variety) (Topic 11),
and inflight meal (special demand) (Topic 21). Also for FSCs, seats are divided
into five subtopics – seat comfort (space and location) (Topic 2), seat comfort
(flight distance) (Topic 9), family seat request (Topic 12), seat comfort (aircraft
type) (Topic 18), and seats (overall evaluation) (Topic 22). For LCCs, seats are
divided into two subtopics – seat comfort (space and location) (Topic 6) and seat
comfort (flight distance) (Topic 7). However, for LCCs, there is only a single
subtopic for meal services – inflight meal purchase (Topic 4). The subdivision is
specifically evident for FSCs. Moreover, the subdivided topics do not match
single dimensions, but are instead separated into multiple dimensions. In most of
the previous studies, features related to inflight meal services and seats have
belonged to tangibles. Although they mainly belong to tangibles in this study (five
of nine for FSCs and two of three for LCCs), some of the topics match with
empathy and reliability. This indicates that diverse aspects of customer
perceptions exist with respect to quality, especially for primary service items. As
such, they cannot be simply measured as belonging to tangible dimensions as
they have been in previous studies. If studies are carried out using surveys, the
surveys must be designed in a more sophisticated and nuanced manner to
accurately reflect the diverse aspects of traveler perceptions. This is particularly
true for meal services and seats. Indeed, food service is a complex mixture of
multiple components, including ingredient freshness, menus, drinks, moods,
employee courtesy, and so forth. The service is also affected by cultural and
social factors (Aksoy, Atilgan, & Akinci, 2003).
Inflight meal services have played an important role in airline service marketing.
Good meal services create a positive effect on word of mouth among customers,
which serves as important information in airline selection (Heide, Grønhaug, &
Engset, 1999). From this, it is reasonable to think that the importance of the meal
service leads to the division into four distinctive subtopics in this study. Through
words such as short, quick, takeoff, welcome, and prior, we see that customers
recognize the punctuality of meal services. From the word distribution
(chicken, fruit, salad, bread, snack, beef, and so forth) for the topic of inflight
meal (menu variety), customers recognize which type of menu is served and how
that menu varies throughout the service. Aksoy et al. (2003) showed that the
punctuality and menu variety of inflight meals were important service measures
in both foreign and domestic airline services. Laws (2005) also found that there
was a more diverse demand for meals as the flight distance increased.
The inflight meal (special demand) topic differs from the inflight meal (menu
variety) topic in terms of the degree of service customization. Offering a variety of
items on a menu is indicative of good service, but providing customized meals for
individual passengers suggests a higher level of quality. When we look at the
word distribution for inflight meal (special demand), words such
as special, avail(able), request, and vegetarian appear. This indicates that FSCs
should attend to customer requests. Examples include meals for passengers that
have specific dietary restrictions, including vegetarians, infants, and religious
adherents. Dana (1999) also asserted that airlines should provide special meals
that reflect the dietary or religious needs of travelers. The inflight drink service
topic seems to be recognized as a separate service by customers rather than as
a part of the overall meal service. In fact, words associated with drinks are not
present in the other three meal-related topics. This suggests that the importance
of drink service cannot be overlooked by FSCs. For the meal-related topics in
LCCs, inflight meal purchase is the only topic.
Together with the meal service, seats are another primary service item when
evaluating airline service quality. Seat comfort (space and location) (Topic 2 for
FSCs and Topic 6 for LCCs) is a service feature dealt with in almost every study
on airline service quality (Aksoy et al., 2003; Chang & Yeh, 2002; Gilbert &
Wong, 2003; Hussain, Al Nasser, & Hussain, 2015; Jiang, 2013; Ostrowski,
O’Brien, & Gordon, 1993; Pakdil & Aydın, 2007; Park et al., 2005; Saha &
Theingi, 2009; Tsaur et al., 2002; Young et al., 1994), and it typically has a tradeoff relationship with price (Balcombe, Fraser, & Harris, 2009). Based on the
measure of seat comfort with seat width and pitch, low values in terms of seat
comfort are associated with increased seating capacity. This increased capacity
results in lower unit operating costs, but it also reduces the level of service
quality onboard (Lee & Luengo-Prado, 2004). On the other hand, LCC strategies
generally rely on this idea. Because the same negative word (uncomfort(able)) is
found in both the carriers, we see that customers are sensitive to this feature of
service quality.
Looking at the words associated with the seat comfort (flight distance) topic
(Topic 9 for FSCs and Topic 7 for LCCs), we can infer that customer perceptions
for seat comfort are dependent on flight distance. Recently, many LCCs have
tried to expand their business territories to include long haul markets (Francis
et al., 2007). Therefore, LCCs need to pay special attention to seat comfort for
long-distance flights. For example, LCCs might consider the introduction of an
upper class service such as premium economy (see Topic 19) to compensate for
this weakness (Morrell, 2008).
Sentiment analysis
The sentiment analysis is a useful tool to elicit customer perceptions of service
and new perspectives to improve service features based on customer opinions
(Misopoulos et al., 2014; Wei, Chen, Yang, & Yang, 2010). An opinion can be a
positive, negative, or neutral emotion or attitude from customers towards a
service, product, or topic (Liu & Zhang, 2012). We employ a well-established
word dictionary made by Hu and Liu (2004) to analyze sentiments. By applying
the same stemming method in Appendix B to exactly match our data to the
dictionary, we produce the optimal form of the dictionary to the data. To read
sentiments in terms of topic level, the sentiment score of the kth topic (k = 1 … K)
is calculated as (Guzman &
Maalej, 2014).TSk=∑Nn=1wn/k⋅wsn/∑Nn=1wn/kTSk=∑n=1Nwn/k⋅wsn
/∑n=1Nwn/kwhere wsnwsn has a value of + 1 if the nth word in the kth
topic matches with the positive word in the dictionary, a value of −1 if the nth
word in the kth topic matches with the negative word in the dictionary, and a
value of 0 otherwise. wn/kwn/k has the same definition as in Equation (1).
Similarly to Hu and Liu (2004), we obtain a sentiment score of the specific
dimension by adding up the sentiment scores of topics (TSkTSk) that belong to
the specific dimension. Then we compare the sentiment scores of the FSCs and
LCCs. We briefly introduce the comparison results of sentiments within the
dimension level. In terms of the total sum of sentiment scores, LCC travelers
have slightly lower negative perceptions of the airline service than FSC travelers.
When we observe the sentiment distributions in Figure 3, negative words appear
more in the three dimensions of the FSCs, although there is not a large
difference in empathy. In a similar vein, LCCs have two negative dimensions.
Reliability and responsiveness are negatively perceived for both FSCs and
LCCs. While reliability is the most significant dimension of LCCs, the dimension
includes features that are likely to cause negative perceptions such as flight
delay (Topic 1), punctuality (Topic 13), and change & cancelation (Topic 18). In
FSCs, reliability is the second most significant dimension, and it covers
negatively perceived features such as punctuality (Topics 5 and 6) and
reservation including cancelation (Topic 29). This result is consistent with
previous studies (Misopoulos et al., 2014; Yee Liau & Pei Tan, 2014). Such
studies showed that Twitter messages regarding flight delay and cancelation
received more negative sentiments. Although responsiveness has a smaller
significance than reliability, its features are mainly related to waiting situations
such as waiting for check-in & boarding, flight delays and employee reactions.
Unless the waiting time is almost zero, it is likely to elicit negative sentiments
from travelers. In contrast, tangibles and assurance are positively perceived for
both FSCs and LCCs. Features regarding seats and aircraft conditions, which
principally comprise the tangibles dimension, contribute to obtain positive
sentiments for both types of air carriers. Assurance is composed of features of
general employee quality, and positive perceptions are dominant for both types
of carriers. FSCs should sharpen tangibles features to show more positive
sentiments since the dimension is positively perceived and the most significant.
Figure 3. Sentiment score distributions without neutral scores. Top panel is for
FSCs and bottom panel is for LCCs. Distributions in the topic level are available
upon request from the authors.
Display full size
Conclusions
Competition between FSCs and LCCs has intensified. To determine differences
in service quality, we investigated customer perceptions of quality based on the
LDA topic model. As a result of the LDA modeling, we extracted 27 and 19
features of perceived service quality for FSCs and LCCs, respectively, from
passenger-authored online reviews. Also we further reduced the dimensions of
service quality by matching them to five dimensions for an overview. Through
this, we resolved the RQ1. From the quantified comparisons, we learned that
there was a reasonable difference in traveler perceptions between the FSCs and
LCCs. In terms of dimension comparisons, the most significant dimensions for
the FSCs and LCCs were tangibles and reliability, respectively. On the other
hand, the least significant dimensions were assurance and empathy,
respectively. To differentiate more specifically between services, we compared
the results in terms of unique features. There were a number of unique features
(15 of 27 for the FSCs and five of 19 for the LCCs) for service quality that
simultaneously did not belong to either type of air carrier. We showed how each
type of air carrier should focus on or improve specific features of their service to
sharpen service differentiation through discussions of unique features and
reviews from the literature. The conclusions were able to provide answers for
RQs 2 and 3.
With respect to academic research perspectives, this also suggested that the
service features in survey forms used for FSCs and LCCs, which have
resembled one another thus far, needed to be modified to incorporate specific
features that might be revealed in the unique topics of this study. Moreover, a
couple of primary service items (inflight meal services and seats), which have
typically been regarded as part of a specific dimension (tangibles), were
subdivided and matched to multiple dimensions simultaneously. This finding also
needs to be considered for future research designs and covers RQ4.
Using the brief analysis of sentiments to provide an answer for RQ5, we found
that LCC customers generally held less negative emotions than FSC customers,
although the difference was not significant. Because the reliability was negatively
perceived and the most significant, it was concluded that LCCs should hone their
features to avoid negative sentiments. On the contrary, FSCs need to sharpen
tangibles features with respect to providing more positive sentiments since the
dimension is positively perceived and the most significant.
This study is limited in that the LDA modeling itself cannot provide the causes of
differentiation between FSCs and LCCs. For example, the results do not explain
why such topics occur and how they are specifically related. To do this, additional
investigations, including co-occurrence and trend analyses, might be employed.
Through such additional analyses, we would be able to expand the knowledge
regarding service quality differentiation in the air transport industry.
Purchase answer to see full
attachment
Why Choose Us
- 100% non-plagiarized Papers
- 24/7 /365 Service Available
- Affordable Prices
- Any Paper, Urgency, and Subject
- Will complete your papers in 6 hours
- On-time Delivery
- Money-back and Privacy guarantees
- Unlimited Amendments upon request
- Satisfaction guarantee
How it Works
- Click on the “Place Order” tab at the top menu or “Order Now” icon at the bottom and a new page will appear with an order form to be filled.
- Fill in your paper’s requirements in the "PAPER DETAILS" section.
- Fill in your paper’s academic level, deadline, and the required number of pages from the drop-down menus.
- Click “CREATE ACCOUNT & SIGN IN” to enter your registration details and get an account with us for record-keeping and then, click on “PROCEED TO CHECKOUT” at the bottom of the page.
- From there, the payment sections will show, follow the guided payment process and your order will be available for our writing team to work on it.