WestminsterResearch

http://www.westminster.ac.uk/westminsterresearch

Airline Schedule and Network Competitiveness: a consumer-

centric approach for business travel
Nenum, S., Graham, A. and Dennis, N.

NOTICE: this is the authors’ version of a work that was accepted for publication in
Annals of Tourism Research . Changes resulting from the publishing process, such as
peer review, editing, corrections, structural formatting, and other quality control
mechanisms may not be reflected in this document. Changes may have been made to
this work since it was submitted for publication. A definitive version was subsequently
published in Annals of Tourism Research, 80, 102822, 2020.

The final definitive version in Annals of Tourism Research is available online at:

https://dx.doi.org/10.1016/j.annals.2019.102822

© 2020. This manuscript version is made available under the CC-BY-NC-ND 4.0 license
https://creativecommons.org/licenses/by-nc-nd/4.0/

The WestminsterResearch online digital archive at the University of Westminster aims to make the
research output of the University available to a wider audience. Copyright and Moral Rights remain
with the authors and/or copyright owners.

Whilst further distribution of specific materials from within this archive is forbidden, you may freely
distribute the URL of WestminsterResearch: ((http://westminsterresearch.wmin.ac.uk/).

In case of abuse or copyright appearing without permission e-mail [email protected]

 AIRLINE SCHEDULE AND NETWORK COMPETITIVENESS:
A CONSUMER-CENTRIC APPROACH FOR BUSINESS TRAVEL

ABSTRACT

The aim of this paper is to undertake a comparative analysis of the competitiveness of airline
schedules and networks from a business traveller perspective with a particular focus on
Europe and travel to and from this continent. A unique and innovative customer-centric
model is developed using a passenger survey and airline schedule data to overcome the
shortcomings of traditional models that lack the passenger viewpoint. The results show that
Austrian Airlines/Vienna airport and Swiss/Zurich airport have the highest quality
connections, whilst the top five competing European hubs are served by the Lufthansa group.
The findings provide a significant opportunity to airlines to grow their knowledge and
understanding of their competitive position and their ability to offer scheduling convenience
to passengers.

Keywords: Airlines; schedules; networks; hubs; business travellers

1. INTRODUCTION

Traveller choice of air services is a complex issue. It is affected by numerous factors

including fare, scheduling convenience and other influences such as airline reputation,
reliability, safety, in-flight services, seat availability, airline competition and frequent flyer
programmes. Business passengers, being generally more journey time sensitive and less price
sensitive than leisure passengers, are usually assumed to be particularly influenced by
schedule factors and the design of airline networks. Therefore this paper focuses on this
important aspect of air service choice for business travellers, namely scheduling convenience,
whilst acknowledging that a number of the other factors undoubtedly also play a role. As a
consequence, it is not the purpose of the research to produce an econometrically driven all-
inclusive choice model for air passengers.

Traditionally to understand and model the attractiveness of the scheduling or ‘quality’ aspects
of the airline product, airlines have applied the so-called Quality of Service Index (QSI)
which uses parameters such as weekly frequency, seat capacity, aircraft type and total
journey time to estimate individual airline market share. Academic research in this area has
also tended to focus on these parameters. However such an approach is primarily a
mathematical computation process that lacks the passenger perspective. This paper
overcomes this shortcoming by adding in passenger views as regards scheduling convenience
that have been obtained from an extensive passenger survey. Hence it has been possible to
develop an innovative consumer-centric approach with new index measures to demonstrate
passenger preferences which has been called the realistic market share estimation (REMSET)
model. The overall specific aim here in using the REMSET model is to undertake a
comparative analysis of the competitiveness of airline schedules and networks from a

1
business traveller perspective with a particular focus on Europe and travel to and from this
continent.

2. LITERATURE REVIEW

Analysing and modelling passenger choice of air services and airports is an area that has
received considerable attention in the literature (e.g. Adler et al., 2005; Başar and Bhat, 2004;
Coldren and Koppelman, 2005; Gao and Koo, 2014; Hess et al., 2007; Moreno, 2006;
Oyewole, 2007; Pels et al., 2001), especially by using multivariate probit or discrete choice
logit models With the latter, there are mutually exclusive alternatives to be considered, each
with an associated utility or attractiveness (Garrow, 2010). This general methodology has
also been used within the wider tourism industry for other purposes such as tourism
expenditure (e.g. Wu et al. 2013; Rashidi and Koo, 2016). For air travel, the choice factors
investigated typically include price, schedule features (e.g. flight time, direct or connecting
flight, frequency, routing, aircraft type) and many others such as access time/cost, on-time
performance, frequent flyer programmes, seat availability, in-flight service, airline/airport
reputation, airline competition, security and safety (Abdelghany and Abdelghany, 2009). For
example Coldren and Koppelman (2005) used level-of-service, connection quality, carrier
attributes, aircraft type, and departure time in their discrete choice model for markets in the
United States and Canada.

Meanwhile industry sources, such as the global passenger survey undertaken by the
International Air Transport Association (IATA), have found that the top three factors
influencing ticket purchase with a particular airline are price (43%), schedule and convenient
flight time (21%) and frequent flyer programme (13%) (IATA, 2015). Similar findings have
been recorded elsewhere, for example in the UK, where it was observed that the top three
influences were price, flight times and location of departure/destination airports (Mintel,
2013). This shows how important flight times, which are the focus of this research, are to
passengers.

The passenger’s choice is also influenced by their specific characteristics such as socio-
economic profiles (Castillo-Manzano and Marchena-Gómez, 2010; Milioti et al., 2015) and
income (Gallet and Doucouliagos, 2014). Moreover trip purpose is very important. At the
broadest level, even just differentiating between business and leisure passengers can be
insightful as these two groups have different travel preferences and required experiences
(Unger et al., 2016) - so separate choice models can be developed (e.g. see Wu and So,
2018). As an example in the UK, the Civil Aviation Authority (CAA) found that the ‘flight
time or route was a particular consideration for business flyers’ (CAA, 2015: 37), whilst
Mason (2001) found that for UK business travellers the most significant factors were
punctuality and frequency, with price/ticket flexibility only being the third most important.
As regards fares, in an analysis of 21 different research papers, Gillen et al. (2002) observed
median price elasticity values of -0.26 for long-haul international business travel and -0.73
for short to medium-haul business, whilst the values for leisure demonstrated a more price
sensitive demand (-0.99 for long-haul international leisure and - 1.52 short to medium-haul

2
leisure). InterVISTAS (2007) in their comparative study of 23 papers also confirmed that
business travellers were less sensitive to fare changes than leisure travellers. Moreover
Seetaram et al. (2018) confirmed that passengers using business class are less price sensitive
when it comes to paying passenger taxes. In addition with their discrete logit model
Proussaloglou and Koppelman (1999) found a higher price sensitivity of leisure travellers,
with greater importance of convenient schedules being given to business models. Milioti et
al. (2015) with their multivariate probit model concluded that people who travel for business
purposes are less price sensitive and more likely to consider the flight schedule (and frequent
flyer program) as being important choice factors.

As a consequence this means that for business travel schedule convenience attributes can be
very significant hence the focus of this research on business travel. This is the case for both
direct flights, and also connecting flights where there may well be many options on offer.
This has led to growing research in this area and particularly in the development of airport
connectivity models (Burghouwt and Redondi, 2013). Different approaches have been
adopted, especially when assessing hub airports, that not only take into account the number of
connections, but also the perceived ‘quality’ in terms of transfer and in-flight times (Doganis
and Dennis, 1989; Veldhuis 1997; Burghouwt and De Wit, 2005; Burghout and Veldhius,
2006). The so-called Netscan model, which has been developed in some of this research, is
also used by the industry organisation ACI-Europe (2017). Meanwhile Allroggen et al.
(2015) have established a Global Connectivity Index that includes a destination quality
variable as well. In addition more detailed hub performance of certain individual airlines has
been examined. For example Li et al. (2012) analysed All Nippon Airways’ dual-hub strategy
at Toyko’s Haneda and Narita airports, while Logotheris and Miyoshi (2018) used what they
defined as a Hub Connectivity Performance Analyser (HCPA) to evaluate the hub
connectivity of Turkish Airlines at Istanbul airport and Emirates Airline at Dubai airport.
Within this literature on connectivity models, various approaches to determining key
connectivity variables such as the minimum and maximum connection times, and maximum
detour distances for connections, are frequently discussed and have been used to inform the
development of the REMSET model.

As already discussed, within the airline industry passenger choice factors are commonly
taken into account by using the Quality of Service Index (QSI). This was initially used pre-
airline deregulation by the US government to predict traffic changes due to changes in airline
service. It has evolved to take into account factors such as aircraft types, number of stops and
direct versus indirect services (Belobaba et al., 2016). A coefficient is applied to each of these
factors which can have a relative value (e.g. non-stop flight = 1, single connection = 0.25) or
absolute value (e.g. number of frequencies). By imputing data related to the airport pair
market size and airline schedules, the index can then calculate market share. It has thus
become more sophisticated but has still been subject to much criticism (Jacob et. al., 2012;
Transportation Research Board, 2002), particularly due to its simplistic methodology, and
being based on data from airline schedules which cannot take into account the passenger
perspective.

3
In essence, there are currently two main methods to assess the competitiveness of airline
schedules and networks. One is a ‘top-down’ approach to developing a network connectivity
index by considering current airline schedules simultaneously and making assumptions about
key connectivity variables such as the minimum and maximum connection times, and detour
distances. The other ‘bottom-up’ approach involves devising a scoring system with the QSI
index for individual routes, taking into account factors such as aircraft type, frequency and
connections, which can be used to determine market share on each individual route.

However, both these approaches do not effectively take into account the passenger viewpoint.
This is primarily due to two reasons. Firstly, the assumptions that are made are rarely built
around actual up-to-date evidence about passenger preferences obtained directly from
passengers. Secondly they ignore other important factors that affect passenger attitudes, for
example, in relation to operating or code-share flights, scheduling preferences for connecting
flights and ideal time allocations at airports and between sectors of multi-leg itineraries. The
methodology adopted here and described in the next section aims to overcome these
significant shortcomings by adopting a more passenger-centric approach. Moreover this
method brings greater flexibility, as it can be used just as effectively on a route-by-route,
country-by-country or region-by-region basis, with a focus on either comparative airline or
airport performance.

3. METHODOLOGY

3.1 Data Sources

The methodology for this research involved combining information obtained directly from
the schedules of airlines and a passenger survey to inform the schedule convenience
measures. For the schedule data, IATA’s Schedule Reference Service (SRS) database, which
is designed to fulfil the data requirements of airlines, global distribution systems (GDSs) and
airports, was used. This contains information of passenger and cargo schedules of more than
900 airlines worldwide, representing more than 99.9% of the air segments flown every day.
This specific analysis is based on 115,711 flights (52,867 operating flights and 62,844
codeshare flights) for a final week in June 2016 as detailed in SRS.

The passenger survey was undertaken in 2015 at nine global airports (New York, Delhi,
London, Istanbul, Dubai, Geneva, Frankfurt, Hong Kong and Johannesburg) with a total
sample size of 962. The number of respondents was evenly split between the nine airports
with the smallest sample of 98 at Delhi airport and the highest sample size of 114 at Hong
Kong airport. The survey was undertaken in the departure areas of the airports, with a paper
self-completion questionnaire, at different time intervals and different days of the week over
the January-March period to encourage diversity within the respondent set.

Following on from the literature review, the basic attributes of a business trip itinerary with
regards to schedule convenience were identified and grouped into three categories, namely
core, time- and product- related factors. The core attributes involved gathering basic schedule

4
information obtained directly from SRS such as direct or connecting flights, and operating or
codeshare flights. However, in order to determine the quantitative relationship concerning
preferences between a direct and connecting flight as well as an operating and codeshare
flight, questions regarding the passenger’s willingness to pay different levels of fare for each
option were asked in the passenger survey. As expected, this found that travellers were
willing to pay more for direct flights than for connecting services, with their willingness to
take a codeshare flight was lower than for an operating service. The time-related measures,
which were also directly obtained from airline schedules, were journey time, flight time and
connecting time (ttotal, tflight, tconn), with the journey time being equal to the sum of the other
two measures.

The product-related factors were mostly measured with the survey of passenger preferences.
In order for the responses to be included into the accepted respondent set, two criteria were
tested: (i) The respondents had to have flown within the past 12 months and (ii) they had to
report at least some interest in schedule convenience (e.g. listing this as a factor that
influenced their choice of flight). Overall within the sample there was a broad spread of
return trip frequency in the last 12 months (1=19%; 2=23%; 3-5=36%; 6-9=15%; 10 or
more=7%) with an average trip number of 4.06.

However, data for a few product-related factors were obtained directly from the SRS
database. This included each airport’s minimum connection time (MCT), which is the
minimum time required for a passenger to leave the aircraft of the incoming flight at the
connecting airport, complete the formalities and catch the next flight. Another measure, taken
directly from the schedule data was the detour factor. This is the ratio between the direct and
indirect connecting flights in terms of distance. The detour factor has a minimum value of 1
and as the detour ratio rises, the inconvenience of the flight increases compared to a direct
flight alternative. Any flights above a detour factor of 1.75 were excluded from the analysis
to avoid using remote connections. Data related to the ratio between the flight time of the two
connecting flights were also gathered.

The first of the product-related attributes obtained directly from the passenger survey were
the departure and arrival time quality measures (qdep, qarr), when passengers were asked about
the attractiveness of departure time and arrival time (in terms of worst, poor, good, best) for
two hour time periods in a 24 hour day/night. For example for departing flights, the best
periods were 08:00 – 09:59, 18:00 – 19:59 and 20:00 – 21:59, whereas the worst were 02:00
– 03:59 and 04:00 – 05:59. Then there was the Maximum Connection Time (MaxCT). This
was needed because not all connections meeting the MCT criterion are attractive for
passengers and there was a need to eliminate the unattractive very long connections. The
survey asked about the maximum tolerance to wait for the next flight and the average value
was 290 minutes.

The survey also asked for preferences about the Time Split Ratio (qsplit). For instance, for two
competing connecting itineraries having the same flight time of 9 hours, passengers may have
a preference for having 8 hours for the first leg and 1 hour for the second leg, compared to the
alternative which could be an equal 4.5 hours of flight time for both legs. Flights split either

5
at the very early or at the very late phases of the journey appeared to be the most popular.
This could be more convenient as it can provide, for example, more uninterrupted sleep and
in-flight-entertainment time.

Another measure is the Stress Time Factor (tstress) which was computed with reference to the
MCT. This is needed because this minimum time is not always favoured by passengers, as
any delays with the incoming flight or any complication or inconvenience experienced at the
connecting airport can increase the risk of missing the second flight. For this reason, it was
assumed that passengers prefer a buffer time (tbuffer) (on top of the MCT at the connecting
airport). This assumption was tested by the passenger survey and an average buffer value of
29.2 minutes was obtained. If the connecting time of the journey was less than the MCT plus
the buffer time preferred by the passengers it was deemed stressful. The Stress Time Factor
was calculated to be the difference between the connecting time and the sum of the MCT and
buffer. It will be zero for direct flights (as the passenger does not use a connecting airport)
and when the connecting time is greater than the MCT plus buffer time. There was also the
Wasted Time Factor (twaste) which, like the Stress Time Factor, was calculated with reference
to the MCT and was tested through the passenger survey. Any connecting time, which is
more than the MCT plus buffer time, is deemed to be wasted, but this has a value of zero if
the connecting time is less, or if it is a direct flight. Therefore these definitions imply that if
the connecting time at the connecting airport is not equal to the sum of the MCT and buffer
time demanded by passengers, then there is either a wasted or stress time for the passengers,
called the inconvenient time (%inconvenient_time). If the stress time of a journey is greater than
zero, the wasted time is equal to zero. Conversely, when a journey includes wasted time, then
it lacks stress time.

Table 1 summarises the details about all the attributes considered in the REMSET model, the
survey main findings (when appropriate) and the data sources used.

Table 1: REMSET Measures and Data Sources

Attribute Detail and results of survey (when appropriate) Source

(scale/unit)
Core Attributes
Direct or Direct flight from the origin to destination or connecting Schedule
connecting flight where passengers change aircraft at an intermediary data
flight airport.
(dichotomous
scale - no unit)
Operating or Flight is provided by the operating airline or on behalf of Schedule
codeshare another marketing carrier (i.e. a codeshare flight). data
flight
(dichotomous
scale - no unit)
Value of direct Passenger preferences (measured by willingness to pay) Survey
vs connecting Survey findings (varies by schedule

6
and operating inconvenience/convenience):
vs codeshare • Direct and operating service (udo): 1.230 – 1.333
flight • Direct and codeshare service (udc): 0.864 – 0.990
(numerical • Connecting and operating service (uco): 1.000 – 1.131
continuous • Connecting and codeshare service (ucc): 0.796 – 0.854
scale – no
unit)
Time - Related Attributes
Flight time Time spent on the flight excluding any connecting time. Schedule
(tflight) (mins) data
Connecting Time spent at the connecting airport (zero for direct flights). Schedule
time (tconn) data
(mins)
Journey time Total travel time between the departure from the origin Schedule
(ttotal) (mins) airport and landing to the destination airport (flight time + data
connecting time).
Product - Related Attributes
Minimum Minimum time required to leave the aircraft of the incoming Schedule
connection flight at the connecting airport, complete the formalities and data
time (MCT) catch the next flight.
(mins)
Maximum Passenger view of maximum acceptable connection time. Survey
connect time Survey findings: Average value of 290 minutes
(MaxCT)
(mins)
Detour factor Ratio between the direct and connecting flight in terms of Schedule
(numerical distance (minimum value is 1 for direct flights and data
continuous maximum value is set at 1.75).
scale)
Flight time Ratio between the flight time of the two connecting flights Schedule
split ratio data
(fsplit)
(numerical
continuous
scale)
Departure time Passenger view of attractiveness of departure time for each Survey
quality (qdep) time interval of the 24 hour day/night.
(ordinal scale Survey findings:
with four time 1 (Worst) A 02:00 – 03:59, 04:00 – 05:59
periods) 2 (Poor) A 00:00 – 01:59, 12:00 – 13:59, 14:00 – 15:59
3 (Good) A 06:00 – 07:59, 10:00 – 11:59, 16:00 – 17:59,
22:00 – 23:59
4 (Best) A 08:00 – 09:59, 18:00 – 19:59, 20:00 – 21:59

7
Arrival time Passenger view of attractiveness of arrival time for each time Survey
quality (qarr) interval of the 24 hour day/night.
(ordinal scale Survey findings:
depending on 1 (Worst) A 00:00 – 01:59, 02:00 – 03:59, 04:00 – 05:59
four time 2 (Poor) A 06:00 – 07:59, 20:00 – 21:59, 22:00 – 23:59
periods) 3 (Good) A 10:00 – 11:59, 14:00 – 15:59, 18:00 – 19:59
4 (Best) A 08:00 – 09:59, 12:00 – 13:59, 16:00 – 17:59
Time split Passenger preferred time split ratio between the two Survey
ratio quality connecting flights of one journey.
(qsplit) Survey findings: Passengers preferred their journey to be
(numerical interrupted by the connection either at the very early or late
discrete scale stages of the flight -
with five (For nine hours flying time 41% preferred first leg=8
options) hour/second leg = 1 hour and 35% preferred first=1
hour/second leg = 8 hours)
Buffer time Passenger preferred additional connecting time to the MCT. Survey
(tbuffer) (mins) Survey findings: Average value of 29.2 minutes
Stress time For connecting flights, the stress time factor is the difference Calculated
factor (tstress) between (MCT + tbuffer) and tconn: from other
(mins)
{
Stress Time = MCT + t
0, if tconn ≥ (MCT + tbuffer
buffer - tconn
)
measures
, if tconn < (MCT + tbuffer)
For direct flights, the stress time factor is zero.
Wasted time For connecting flights, the wasted time factor is the Calculated
factor (twaste) difference between tconn and (MCT + tbuffer): from other
(mins)
Wasted Time = t { 0, if tconn ≤ (MCT + tbuffer
conn - (MCT + tbuffer)
)
measures
, if tconn > (MCT + tbuffer)

For direct flights, the wasted time factor is zero.

%inconvenient_time For connecting flights, the share of inconvenience time (i.e. Calculated
(%) stress or waste) within the total journey time from other
measures

3.2 Building the REMSET Model

This section now describes how the REMSET model was built, using the information from
Table 1. Firstly combining the values of qdep, qarr and qsplit can produce an overall schedule
time convenience score for a given itinerary:

qconvenience = qdep + qarr + qsplit

The minimum possible value of qconvenience defines the worst possible time convenience,
whereas at the other extreme, the highest value refers to the best timing. As shown in Table 1,
for qdep and qarr the values were defined to be between 1 and 4, where 1 implies the least

8
preferred option, whilst qsplit has a minimum value set at zero. Therefore, the minimum value
of qconvenience can be 2 (1 for qdep, 1 for qarr and 0 for qsplit), while the maximum can be 8 + qsplit
(4 for qdep, 4 for qarr).

After identifying the routing and the flight type (operating vs codeshare) of an itinerary, an
index score qindex can be obtained. For example (as can be seen from Table 1) with a direct
and operating flight, the qindex value will range from 1.230 to 1.333 depending on the time
convenience factor, qconvenience. If the itinerary has the lowest qconvenience score of 2, the quality
value of the product would be 1.230. On the other hand, in the reverse case, if the qconvenience is
reported to be the highest, which is 8 for a direct service, then the quality value for the
product would be 1.333. For itineraries with qconvenience scores in between the lower and upper
bound, the qindex value would result in a value more than 1.230 and less than 1.333. These
values were obtained from the willingness to pay part of the passenger survey which looked
at the perceived passenger value of connecting vs direct flights, and operating vs codeshare
flights. The willingness to pay was indexed to 1 for an inconvenient time on an operating
connecting flight.

In essence qindex is a function of an itinerary’s routing, type and qconvenience scores. However
the qindex score is not a leg based metric assessing the standalone quality index of the
individual segments within a journey. Instead it refers to a factor that measures the quality
value of the entire directional itinerary as a whole. This implies that for connecting journeys,
qindex does not reflect the impact of an inconvenient time period within the total journey time
due to stress or wasted time. So qindex needs to be discounted by %inconvenient. The factor that is
obtained after the discounting is called qindex_normalised

Finally although qindex_normalised is a parameter measuring an itinerary’s schedule related

quality, it does not take into account the relative performance in total journey time. ttotal is a
definite factor of choice for travellers, as they do not appreciate longer journey times.
Therefore, the relative advantage or disadvantage of the itinerary’s ttotal needs to be
incorporated into the final schedule convenience score, which is performed by adjusting the
qindex_normalised. In order to include this ttotal into the final quality score calculation, firstly if an
airline offers more than one combination of connections, total journey times of each
combination are weighted with reference to the corresponding seat supply. Then the
normalised quality indexes (qindex_normalised) are discounted with reference to the shortest
available product per service type to find the qa_index_normalised. For each service type, the best
performing airline gets qa_index_normalised equal to qindex_normalised, where the remaining
qindex_normalised scores are discounted with reference to ttotal, by the percentage deviation from
that minimum. Through this discounting mechanism, shorter ttotals are rewarded. While the
deviation from the minimum ttotal increases, the adjusted normalised quality score decrease as
the appreciation of the product descends in comparison to the competitors available in the
market. This produces the final adjusted normalised quality score (qa_index_normalised) which is a
key measure of the REMSET model.

9
For the research in this paper, the REMSET model was used to compare the hub performance
of hub airlines and their codeshare partners by using their schedule convenience scores (i.e.
qa_index_normalised) and frequency shares (%f) for connecting itineraries at a region-to-region level.
The qa_index_normalised score of the airport can be regarded as the hub’s connection quality score as
it indicates the average quality score of the itineraries using the airport as a transfer point. %f
in the general REMSET model is the frequency shares connecting itineraries. It is based on
all possible flights, calculated from scratch from the SRS database, which can be used for
connections that fall between the MCT and MaxCT and have a detour factor of less or equal
to 1.75. For this specific analysis the %f is calculated by rebasing these general percentages
with reference to the sum of connecting frequencies available at only the selected hub airports
in the paper at a region to region level. %f and qa_index_normalised include both the operating and
codeshare itineraries and all flights were considered, including multiple stop ones. Since
these factors are examined to assess the connectivity efficiency of the competing airline hub
airports only connecting flights are included in the analysis, eliminating direct services. The
connectivity index (CI), which is the capacity blended hub quality performance of the
competing hub airports is calculated as follows:

CI = %f x qa_index_normalised

So CI is a numerical index that gives an overall connectivity score for airports by taking into
account both the schedule convenience scores, based on passenger preferences, and the
frequency shares. For this analysis, regional origin and destination (O&D) pairs associated
just with Europe (Europe/Middle East, Europe/America, Europe/Africa, Europe/South Asia,
the Far East and the Commonwealth of Independent States, Europe/Europe) reported in the
2016 schedules (both directions) are considered for selected hubs, as a global coverage would
be too extensive for this paper. As Europe/Australia is only served one-stop in a limited
manner, the Australasian continent is not included. In terms of total O&D and connecting
revenue passenger kilometres (RPKs), in 2017 these five regional flows accounted for around
a third of global RPKs. The Europe/Europe traffic was the largest market (37% of RPKs)
followed by Europe/America (26%) and Asia/CIS (19%). The share for Europe/Middle East
was 11% and Europe/Africa 7%. (Boeing, 2018) The airlines, their hubs and country location
that were used for the analysis are shown in the Appendix. While determining the selected
airports, major hubs in the Gulf region were chosen. Likewise, historically larger European
hubs (i.e. Amsterdam, London Heathrow, Frankfurt, Munich and Paris Charles De Gaulle)
served by the major three global alliances and emerging transfer points (i.e. Istanbul Ataturk
and Sabiha Gokcen) were included in the sample.

4. RESULTS

4.1 Europe/Middle East

The frequency share, quality and CI (equal to frequency share x quality) scores of the
competing airline hubs are reported in Table 2 for the Europe/Middle East market. The best
schedule convenience score for the connecting itineraries are achieved at VIE by Austrian

10
Airline flights followed by ZRH which is used as the hub airport by Swiss. The central
location in Europe of these airports, their compact layout giving short MCTs, and the
scheduling strategy of the two airlines, helps explain why the airports achieve the highest
qa_index_normalised scores. However, the frequency share is relatively disadvantaged in comparison
to rival hubs in the region.

Table 2: Europe/Middle East Connectivity Scores

Airport %f qa_index_normalised CI
IST 30.3% 0.7685 23.29
FRA 11.9% 0.8128 9.67
DXB 8.5% 0.7354 6.25
CAI 6.9% 0.7223 4.98
VIE 4.9% 0.9093 4.46
MUC 5.2% 0.7945 4.13
CDG 6.1% 0.6747 4.12
SAW 6.2% 0.6549 4.06
AMS 4.6% 0.7786 3.58
DOH 4.5% 0.7867 3.54
ZRH 3.6% 0.8709 3.14
AUH 2.8% 0.6079 1.70
BRU 1.6% 0.8023 1.28
LHR 1.8% 0.6471 1.16
MAD 0.8% 0.6102 0.49
LIS 0.3% 0.5873 0.18

The qa_index_normalised is lowest at AUH with the geographical location of Abu Dhabi being
relatively unattractive in comparison to other hubs located in Central or Eastern Europe for
the itineraries in the Europe/Middle East region. Likewise, LHR, LIS and MAD are also
disadvantaged with this market as their locations are not ideal for connections between
Central (or Eastern) Europe and the Middle East. Moreover, the network structure and
scheduling of BA together with high MCTs at LHR, results in this carrier being ranked at the
bottom of the table with all three measures, together with Iberia and TAP.

The frequency share and the CI score are highest at IST, as Turkish Airlines has a significant
share of the Middle East/Europe connections. Furthermore, SAW (the other airport of
Istanbul and a base of the low cost carrier Pegasus) reports a considerable CI score, despite
its poor qa_index_normalised score due to its high frequency score. This confirms that the
geographical location of Istanbul which is located in between the two continents can be
regarded as a crucial factor of connectivity within the Europe/Middle East axis. Comparing
the CI scores at IST (23.29) and FRA (9.67) implies that almost 2.5 times more passengers

11
would transfer via IST than FRA on the hub airlines and their code share partners if only
flight frequency and passenger scheduling preferences are taken into account when
determining flight choice.

4.2 Europe/America

The North America/Europe and South America/Europe markets are reported separately in
Tables 3 and 4. Regarding the adjusted index normalised quality scores, VIE again has the
best performance for North America followed by ZRH (no connections were identified for
VIE for South America). LHR, MAD and CDG are found to have the best qa_index_normalised
scores in the Europe/South America market.

Table 3: Europe/North America Connectivity Scores

Airport %f qa_index_normalised CI
LHR 19.0% 0.8002 15.20
FRA 15.6% 0.8318 12.98
AMS 16.1% 0.7174 11.55
CDG 16.0% 0.7055 11.29
MUC 7.3% 0.8444 6.16
ZRH 6.9% 0.8548 5.90
MAD 6.4% 0.7512 4.81
VIE 4.5% 0.9421 4.24
BRU 3.8% 0.8401 3.19
IST 1.9% 0.7259 1.38
LIS 1.7% 0.7144 1.21
DOH 0.8% 0.6179 0.49
DXB 0.0% 0 0
AUH 0.0% 0 0
CAI 0.0% 0 0
SAW 0.0% 0 0

Table 4: Europe/South America Connectivity Scores

Airport %f qa_index_normalised CI
LHR 19.9% 0.8152 16.22
MAD 17.3% 0.8005 13.85
AMS 16.2% 0.7922 12.83
CDG 14.5% 0.8012 11.62
LIS 9.4% 0.7775 7.31
FRA 7.6% 0.7421 5.64

12
ZRH 4.8% 0.7232 3.47
IST 2.9% 0.6883 2.00
AUH 3.0% 0.5982 1.79
DOH 2.4% 0.6223 1.49
DXB 2.0% 0.6122 1.22
MUC 0.0% 0 0
VIE 0.0% 0 0
BRU 0.0% 0 0
CAI 0.0% 0 0
SAW 0.0% 0 0

Between Europe and America (North and South), the CI score of LHR ranks the highest. The
hubs in the Gulf region namely, DXB, DOH and AUH, are relatively disadvantaged in the
Europe/America market due to their geographical location. It would be inconvenient for
passengers to travel to the Middle East for a connection from Europe to America and vice
versa. Furthermore, unlike its performance in the Europe/Middle East market, IST does not
have a competitive connectivity pattern - also due to its location. Moreover, the supply
advantage of British Airways, Lufthansa and Air France/KLM assist LHR, FRA, AMS, MUC
and CDG in achieving higher CI scores. British Airways BA via LHR offers the highest
number of connecting frequencies in the North America and, thanks to its partnership with
Iberia, BA via LHR also reports the highest frequency share to/from South America. The
location of London between the edge of Europe and America is a significant advantage for
LHR when considering connectivity. MAD and LIS have relatively high scores, primarily
explained by their location and the strong historic and cultural links with Spain and Portugal
and the South American continent.

4.3 Europe/Africa

The results for the Europe/North Africa and Europe/Sub Saharan Africa markets are shown
separately in the Tables 5 and 6. Lufthansa via its hubs FRA and MUC offers the largest
number of connecting frequencies from/to North Africa while Air France via CDG has the
highest connecting frequency share from/to Sub Saharan Africa followed by its partner KLM
via AMS. Concerning the adjusted index normalised quality scores, ZRH (the hub of Swiss)
has the best performance despite its lower frequency share in the entire Europe/Africa
market. BRU’s qa_index_normalised ranks second following ZRH in both North and Sub Saharan
Africa. This finding is in line with the business strategy of SN Brussels which involves a
strong presence in the African continent.

Table 5: Europe/North Africa Connectivity Scores

Airport %f qa_index_normalised CI
FRA 13.6% 0.8171 11.11

13
CDG 15.1% 0.6856 10.35
LHR 12.2% 0.7065 8.62
IST 9.1% 0.6688 6.09
CAI 9.5% 0.6212 5.90
AMS 7.7% 0.6144 4.73
ZRH 5.3% 0.8831 4.68
DOH 5.7% 0.5998 3.42
AUH 5.4% 0.5648 3.05
MUC 3.3% 0.8005 2.64
VIE 2.9% 0.8611 2.50
BRU 2.0% 0.8651 1.73
MAD 2.4% 0.6612 1.59
SAW 2.7% 0.5544 1.50
DXB 2.3% 0.5212 1.20
LIS 0.8% 0.6713 0.54

Table 6: Europe/Sub Saharan Africa Connectivity Scores

Airport %f qa_index_normalised CI
CDG 20.4% 0.6968 14.21
AMS 21.3% 0.6504 13.85
BRU 12.2% 0.7402 9.03
IST 11.3% 0.6159 6.96
FRA 7.6% 0.7095 5.39
LHR 5.0% 0.6903 3.45
CAI 4.6% 0.6829 3.14
DOH 5.2% 0.5984 3.11
AUH 4.1% 0.6146 2.52
MUC 2.8% 0.6922 1.94
DXB 2.4% 0.5407 1.30
ZRH 1.4% 0.7499 1.05
MAD 1.1% 0.5894 0.65
VIE 0.4% 0.5225 0.21
LIS 0.2% 0.5357 0.11
SAW 0.0% 0 0

The superiority of Turkish Airlines over Emirates in transfer opportunities is verified in

Tables 5 and 6 as IST’s %f scores are greater than those of DXB. Additionally, in the
Europe/Sub Saharan Africa market, DXB’s qa_index_normalised score is inferior to the connectivity
index of AUH although both airports are located in the same country and are only 123

14
kilometres away from each other. Therefore, although the qa_index_normalised is influenced by the
geography of the airports, it is clearly not the sole parameter that shapes the connection
quality of the hub airport, which will also be clearly affected by factors such as the layout and
design of the terminal, its utilisation, and the scheduling strategies of the airlines.

4.4 Europe/South Asia, the Far East and the Commonwealth of Independent States (CIS)

In the Europe/South Asia and Far East market, Austrian Airlines via VIE has the best
qa_index_normalised score again, whilst Turkish Airlines via IST has the highest %f share and CI
score as shown in Table 7. The CI scores of the top-ranking airports are not far from each
other, and the dominance of the Gulf hubs (DOH, DXB) and IST in the market is apparent.
DOH has a higher score than DXB which could perhaps be due to more optimal scheduling
by Qatar Airlines and/or more congested facilities at DXB. ZRH again has a relatively
superior qa_index_normalised score in comparison to many other European hubs while CDG and
BRU do not report a competitive quality score. The quality scores of FRA, MUC and AMS
are found to be very similar to each other. Although CAI has an attractive qa_index_normalised
score like VIE, the airport has a very low CI score due to its inadequate level of traffic in
comparison to the rival hubs.

Table 7: Europe/South Asia and Far East Connectivity Scores

Airport %f qa_index_normalised CI
IST 13.7% 0.8670 11.88
DOH 12.7% 0.9089 11.54
DXB 13.5% 0.8112 10.95
FRA 12.7% 0.8191 10.40
AMS 12.2% 0.8150 9.94
LHR 10.4% 0.8008 8.33
AUH 8.6% 0.8056 6.93
CDG 7.4% 0.7669 5.68
ZRH 5.0% 0.8583 4.29
MUC 1.8% 0.8022 1.44
MAD 1.1% 0.6204 0.68
BRU 0.9% 0.6500 0.59
VIE 0.6% 0.9666 0.58
CAI 0.5% 0.9128 0.46
LIS 0.2% 0.4974 0.10
SAW 0.0% 0 0.00

The results for the Europe/CIS route are shown in Table 8. Turkish Airlines via IST and
Lufthansa via FRA have a significant advantage in the market due to their large frequency
share in comparison to rivals. Additionally, IST has the highest qa_index_normalised in the market

15
followed by VIE. Primarily due to their relative geographical inconvenience, MAD and LIS
and the hubs in the Gulf region do not have competitive connectivity indexes.

Table 8: Europe/CIS Connectivity Scores

Airline %f qa_index_normalised CI
IST 33.1% 0.9086 30.07
FRA 24.7% 0.8765 21.65
CDG 6.7% 0.8007 5.36
AMS 6.4% 0.7763 4.97
SAW 6.3% 0.7321 4.61
MUC 4.2% 0.7654 3.21
ZRH 3.8% 0.7965 3.03
VIE 2.8% 0.8128 2.28
LHR 3.0% 0.7569 2.27
CAI 2.8% 0.5609 1.57
BRU 1.7% 0.7075 1.20
DXB 1.6% 0.6769 1.08
AUH 1.0% 0.6534 0.65
MAD 0.7% 0.6412 0.45
DOH 0.6% 0.6888 0.41
LIS 0.2% 0.5976 0.12

4.5 Europe/Europe
With the intra-European connections, the results are overwhelmingly influenced by the
domestic flights as shown in Table 9. Indeed the connections are dominated by the Lufthansa
group due to its relatively larger domestic market. Additionally, the central location of FRA
and MUC in the European continent is another advantage for those airports in achieving
comparably higher CI scores. As expected, the hubs in the Gulf region cannot report a CI
score due to the geographical inconvenience. In terms of the qa_index_normalised scores, ZRH
ranked top, followed by VIE and MUC.

Table 9: Europe/Europe Connectivity Scores

Airline %f qa_index_normalised CI
FRA 31.2% 0.8043 25.09
MUC 14.0% 0.8432 11.80
CDG 12.3% 0.7421 9.13
ZRH 8.8% 0.8821 7.76
AMS 9.2% 0.7766 7.14

16
LHR 8.1% 0.7224 5.85
VIE 5.0% 0.8511 4.26
IST 4.6% 0.6761 3.11
MAD 3.1% 0.6912 2.14
BRU 1.7% 0.7021 1.19
LIS 1.1% 0.6222 0.68
SAW 0.5% 0.5831 0.29
CAI 0.4% 0.6652 0.27
DXB 0.0.% 0 0
AUH 0.0% 0 0
DOH 0.0% 0 0

4.6 Europe/Worldwide

By using the calculated qa_index_normalised scores referred in the tables above, it is possible to
develop the aggregate Europe/worldwide quality index score of the competing hubs. This has
been achieved by simply averaging the quality index (Figure 1). VIE and ZRH have the best
average qa_index_normalised scores while the hubs of Lufthansa, namely FRA and MUC, rank third
and fourth respectively. LHR, AMS, IST and CDG are placed at the middle rankings as the
competitiveness of these hubs in qa_index_normalised scores vary depending on the specific market.
Interestingly, all the five top hubs are served by the Lufthansa group. The Gulf hubs rank
lower primarily due to their geographical location. These relatively lower qa_index_normalised scores
for the Gulf hubs do not necessarily imply that these airports lack good quality connections
everywhere. Indeed whilst they are not in an ideal location for connections with
Europe/America or Europe/Africa, they can, however, offer better connection quality in other
markets. MAD and LIS also have low values, reflecting their relatively weak competitive
position with the exception of the American markets. In addition, SAW ranks second to last
which demonstrates the difficulty that the predominantly point-to-point low cost carrier
Pegasus has in competing for transfer traffic with Turkish Airlines even though they serve the
same location.

Figure 1: Worldwide (Europe/World Market) Average qa_index_normalised Scores

17
 0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
S

IE
CA
LI

M
IS
X

LH

U
U

CD

BR

FR

ZR
SA

V
D

M
A

In order to observe the effect of frequency on connectivity of the hubs in the Europe/World
market, it is possible to calculate the percentage CI (denoted as %CI) of each airport. %CI is
calculated by summing each airport’s CI values in Tables 2 to 9 and then dividing this with
the total sum of all CIs. Therefore, the %CI of an airport refers to its share of the CI scores in
the selected markets and is shown in Figure 2. FRA has the highest share of %CI followed by
IST and CDG. Despite their advantage with the qa_index_normalised scores VIE and ZRH cannot
report higher %CI shares as these airports are relatively disadvantaged in terms of their
frequency supply. As expected, the Gulf hubs have a lower %CI below 4%, primarily due to
their geographical position.

Figure 2: Worldwide (Europe/World Market) %CI Scores

18%
16%
14%
12%
10%
8%
6%
4%
2%
0%
A

IE

S
CA
M

LI
IS

LH

X
FR

CD

ZR

BR

U
V

SA
M

D
A

5. CONCLUSIONS

18
The overall aim of this paper has been to undertake a comparative analysis of the
competitiveness of airline schedules and networks from a business traveller perspective with
a particular focus on Europe and travel to and from this continent. In achieving this aim the
analysis has reflected not only on comparative network or hub airport performance, but also
on the connection effectiveness of the airlines using these airports as their hub location.
These are crucial factors that business travellers take into account when choosing specific air
services. Using the REMSET model, the qa_index_normalised score of the itineraries has been
used as a metric of connectivity quality, while %f was used to reflect the volume of traffic in
terms of frequency at the hub airport. Together these are used to produce the connectivity
index (CI) of the airports which assesses the capacity blended hub quality performance.

Overall, although VIE and ZRH appear to offer ‘high-quality’ connections, the carriers using
these airports as hubs (Austrian and Swiss respectively) are disadvantaged in their volume of
supply which results in their CI scores being relatively small in comparison to their rivals.
IST is found to be a competitive hub for Europe/Middle East and Europe/Far East
connections whereas the hubs in the Gulf region are found to be attractive for the
Europe/South Asia and Far East markets. Furthermore LHR is relatively competitive for both
Europe/North America and Europe/South America connections. Moreover, the Air
France/KLM group is found to be relatively active in the Sub Saharan Africa market although
BRU’s performance is notable in this region too. Geographic location of the hubs is an
essential component of their qa_index_normalised and CI scores. For instance, the hubs in the Gulf
region such as DXB, AUH and DOH are relatively disadvantaged in the Europe/America
market due to the higher detour ratios, as are LIS and MAD with the regions east of Europe.

Although the methodology used in this paper is different from the Netscan model used by
ACI Europe (2018), the overall worldwide CI results are fairly similar. Amongst
European/Middle Eastern hubs, the Netscan model ranks Frankurt first followed by AMS,
CDG, IST, LHR, MUC, DOH, DXB, SVO (Moscow – not included) MAD, MXP (Milan –
not included) and ZRH. The comparative ranking with this research are FRA, IST, CDG,
AMS, LHR, ZRH, MUC, MAD, DOH and DXB. Likewise the research here agrees with
Logothetis and Miyoshi (2018) who find that Turkish Airlines with their Istanbul hub provide
more transfer opportunities between Europe and Africa and America than Emirates.
However, by contrast these authors find Emirates/DXB to be superior to Turkish/IST in the
Asia-Pacific market whereas with this research IST is ranked top in CI terms (albeit that
DXB comes a close third).

These similarities arise primarily because of the frequency share component in the
connectivity measure. However what this research shows, which is missing in other research,
is the consumer-centric perspective regarding schedule convenience, namely the qa_index_normalised
factor. This unique measure identifies the connection and schedule quality of an airline for
any given route or routes taking into account passenger priorities and preferences. The
research has thus conceptualised the comparative and competitive advantages of airline
schedule and network design from the passenger viewpoint and made a significant and
distinctive contribution to knowledge and theory in this area.

19
However, a potential limitation with the application of the REMSET model in this specific
case is the possibility of any bias within the survey. To minimise any bias the wording of the
questions were carefully designed, as was the survey implementation plan at the airports.
The sample was considered adequate in terms of size and diversity even though it was not
possible for practical reasons to use a totally representative sample with a random selection
method. However the issue of any bias within the sample could be addressed with additional
surveys, with the value of certain variables revised if necessary. Nevertheless this does not
undermine the major significance of this research and its role in developing a new and
innovative consumer-centric methodology to assess passenger schedule convenience.

Clearly, the research has considerable practical implications. The results can be used by
individual airlines to assess their competitive position in different regions, analyse the
efficiency of their network investments, and as an input into future scheduling decisions. In
particular, the reasons for the top performance of ZUR (Swiss) and VIE (Austrian Airlines) in
terms of schedule convenience could be investigated further and potentially used as an
example of best practice. One possible general influencing factor might be the level of
competition as it could be argued that if there is less competition (e.g. in terms of a higher
frequency/seat share for certain airlines, or less airlines serving the market with connections)
then there would be less incentives for the airlines in more limited competitive situations to
design or fine tune their network/schedules to produce ‘quality’ connections. However a
visual examination of the relationship between %f (as a measure of competition) and quality
scores shows no obvious link between the two, although this could be investigated further.

The flexibility of the REMSET model means that the findings that have been produced on a
region to region basis can be expanded to include more airports or airlines or further
investigated at a more disaggregate level, such as country to country basis or airport to airport
basis, for example to measure the impacts of additional frequency, changing departure times
or code-sharing with partner airlines, albeit that this is beyond the scope of this paper here
due to length constraints. In fact the REMSET model can be used to assess schedule
preferences for any itinerary as long as the schedule information has been included in the
specific schedule database that is being considered. The research also has implications for
airport management who can assess their comparative position for connections in different
markets which can have consequences, for example, in how they market themselves to
airlines and passengers, or plan for connecting flights.

Having a focus on business passenger, this paper has concentrated on schedule convenience
which is an important choice factor for such passengers. For pure leisure travellers, the
influence of fare is likely to be far more significant and so adopting this approach for these
travellers needs to be undertaken with much more caution. Whether dealing with either
business or leisure travel, it needs to be recognised that to achieve a total picture concerning
choice factors (including other drivers such as fares, load factors, seat availability, airline
competition) would be more appropriately dealt with techniques such as passenger choice
models, but this was not the purpose of this research. Meanwhile the REMSET model could

20
be suitably developed to provide even further insight into the consumer’s view of schedule
convenience by considering factors such as airport connecting facilities and airport
preferences. It could also be used more extensively by considering different years to assess
how the airline networks are evolving and in addition network centrality measures (e.g.
degree, closeness, betweenness, eigenvector) could be calculated to provide further insightful
connectivity comparisons.

REFERENCES

ACI-Europe (2018). Airport Industry Connectivity Report 2017. Brussels: ACI-Europe.

Adler, T., Falzarano, C., and Spitz, G. (2005). Modeling service trade-offs in air itinerary
choices. Transportation Research Record: Journal of the Transportation Research Board,
(1915), 20-26.

Allroggen, F., Wittman, M. D., and Malina, R. (2015). How air transport connects the world–
A new metric of air connectivity and its evolution between 1990 and 2012. Transportation
Research Part E: Logistics and Transportation Review, 80, 184-201.

Abdelghany, A. and Abdelghany, K. (2009). Modeling Applications in the Airline Industry,

Farnham: Ashgate

Başar, G. and Bhat, C. (2004). A parameterized consideration set model for airport choice:
An application to the San Francisco Bay area. Transportation Research Part B:
Methodological, 38(10), 889-904.

Belobaba, P., Odoni, A.R. and Barnhart, C. (eds) (2016). The Global Airline Industry, 2nd
Edition, Chichester: John Wiley and Sons.

Boeing (2018). Current Market Outlook 2018-2037, Seattle: Boeing.

Burghouwt, G., and De Wit, J. (2005). Temporal configurations of European airline

networks. Journal of Air Transport Management, 11(3), 185-198.

Burghouwt, G. and Veldhuis, J. (2006). The competitive position of hub airports in the
transatlantic market., Journal of Air Transportation, 11(1), 106-130.

Burghouwt, G. and Redondi, R., (2013). Connectivity in air transport networks. An

assessment of models and applications, Journal of Transport Economics and Policy, 47, 35–
53.

CAA (2015). Consumer Research for the UK Aviation Sector – Final Report, CAP 1303.
London: CAA.

21
Castillo-Manzano, J. I., & Marchena-Gómez, M. (2010). Analysis of determinants of airline
choice: profiling the LCC passenger. Applied Economics Letters, 18(1), 49-53.

Coldren, G. and Koppelman, F. (2005). Modeling the competition among air-travel itinerary
shares: GEV model development. Transportation Research Part A: Policy and Practice,
39(4), 345-365.

Doganis, R. and Dennis, N. (1989). Lessons in Hubbing. Airline Business, 3, 42-47.

Gallet, C.A. and Doucouliagos, H., (2014). The income elasticity of air travel: A meta-
analysis. Annals of Tourism Research, 49, 141-155.

Gao, Y. and Koo, T. (2014). Flying Australia–Europe via China: A qualitative analysis of the
factors affecting travelers' choice of Chinese carriers using online comments data. Journal of
Air Transport Management, 39, 23-29.

Garrow, L.(2010). Discrete Choice Modelling and Air Travel Demand: Theory and
Application. Farnham: Ashgate.

Gillen, D., Morrison,W. and Stewart, C. (2002). Air Travel Demand Elasticities: Concepts,
Issues and Measurement. Available at:
http://www.fin.gc.ca/consultresp/airtravel/airtravstdy_-eng.asp (Assessed 30 September
2016).

Hess, S., Adler, T. and Polak, J.W. (2007). Modelling airport and airline choice behaviour
with the use of stated preference survey data, Transportation Research Part E: Logistics and
Transportation Review, 43(3), 221-233.

IATA (2015). 2015 IATA Global Passenger Survey. Geneva: IATA

InterVISTAS (2007). Estimating Air Travel Demand Elasticities: Final Report, Vancouver,
InterVISTAS.

Jacobs, T., Garrow, L, Lohatepanont, M., Koppelman, F., Coldren, G., & Purnomo, H.
(2012). Airline Planning and Schedule Development. In Quantitative Problem Solving
Methods in Airline Industry. Boston, MA: Springer.

Li, W. K., Miyoshi, C., & Pagliari, R. (2012). Dual-hub network connectivity: An analysis of
all Nippon Airways’ use of Tokyo’s Haneda and Narita airports. Journal of Air Transport
Management, 23, 12-16.

Logothetis, M., & Miyoshi, C. (2018). Network performance and competitive impact of the
single hub–A case study on Turkish Airlines and Emirates. Journal of Air Transport
Management. 69, 215-223.

22
Mason, K. J. (2001). Marketing low-cost airline services to business travellers. Journal of Air
Transport Management, 7(2), 103-109.

Milioti, C. P., Karlaftis, M. G., & Akkogiounoglou, E. (2015). Traveler perceptions and
airline choice: A multivariate probit approach. Journal of Air Transport Management, 49, 46-
52.

Mintel (2013) Airlines – UK, London: Mintel.

Moreno, M. B. (2006). Airline choice for domestic flights in São Paulo metropolitan area: an
application of the conditional logit model, Journal of Air Transportation, 11(2), 22-42.

Oyewole, P., Sankaran, M., & Choudhury, P. (2007). Consumer choice of airlines in
Malaysia: A synthesis of perspectives from participants in the air travel market. Journal of
International Consumer Marketing, 20(1), 19-31.

Pels, E., Nijkamp, P. and Rietveld, P. (2001). Airport and airline choice in a multiple airport
region: an empirical analysis for the San Francisco Bay Area, Regional Studies, 35(1), 1-9.

Proussaloglou, K. and Koppelman, F.S., (1999). The choice of air carrier, flight, and fare
class. Journal of Air Transport Management, 5(4), 193-201.

Rashidi, T.H. and Koo, T.T. (2016). An analysis on travel party composition and expenditure:
a discrete-continuous model. Annals of Tourism Research, 56, 48-64.

Seetaram, N., Song, H., Ye, S. and Page, S., 2018. Estimating willingness to pay air
passenger duty. Annals of Tourism Research, 72, 85-97.

Transportation Research Board (2002). Aviation Demand Forecasting. Washington D.C.:

National Research Council.

Unger, O., Uriely, N. and Fuchs, G., (2016). The business travel experience. Annals of
Tourism Research, 61, 142-156.

Veldhuis, J. (1997). The competitive position of airline networks. Journal of air transport
management, 3(4), 181-188.

Wu, C. L., & So, T. H. (2018). On the flight choice behaviour of business-purpose passengers
in the Australian domestic air market. Journal of Air Transport Management, 72, 56-67.

Wu, L., Zhang, J. and Fujiwara, A. (2013). Tourism participation and expenditure behaviour:
Analysis using a scobit based discrete–continuous choice model. Annals of Tourism
Research, 40, 1-17.

23
Appendix: Airports and their Hub Airlines used in the Research

Airline Airport Country

Air France Paris Charles De Gaulle France
(CDG)
KLM Amsterdam/Schiphol (AMS) The Netherlands
British Airways London Heathrow (LHR) United Kingdom
Lufthansa Frankfurt International Germany
(FRA)
Lufthansa Munich International (FRA) Germany
Swiss Zurich Kloten (ZRH) Switzerland
Austrian Airlines Vienna Schwechat (VIE) Austria
Iberia Madrid Barajas (MAD) Spain
TAP Air Portugal Lisbon Humberto Delgado Portugal
(LIS)
Turkish Airlines Istanbul Ataturk (IST) Turkey
Pegasus Airlines Istanbul Sabiha Gokcen Turkey
(SAW)
Emirates Airline Dubai International (DXB) United Arab Emirates
Etihad Airways Abu Dhabi International United Arab Emirates
(AUH)
Qatar Airways Doha Hamed (DOH) United Arab Emirates
EgyptAir Cairo International (CAI) Egypt

24