Airline choice for domestic flights in Vietnam: application of multinomial logit model

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM VIETNAM –THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS AIRLINE CHOICE FOR DOMESTIC FLIGHTS IN VIETNAM: APPLICATION OF MU

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM

VIETNAM –THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

AIRLINE CHOICE FOR DOMESTIC FLIGHTS IN VIETNAM: APPLICATION OF MULTINOMIAL LOGIT MODEL

BY

TRAN PHUOC THO

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, December 2016

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

HO CHI MINH CITY THE HAGUE

VIETNAM THE NETHERLANDS

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

AIRLINE CHOICE FOR DOMESTIC FLIGHTS IN VIETNAM: APPLICATION OF MULTINOMIAL LOGIT MODEL

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

TRAN PHUOC THO

Academic Supervisor:

TRUONG DANG THUY

HO CHI MINH CITY, December 2016

ACKNOWLEDGEMENT

First of all, I would like to express my gratitude supervisor Dr Truong Dang Thuy of the Vietnam – The Netherlands Programme (VNP) at Ho Chi Minh City University of Economics for his patience, enthusiasm, and immense knowledge He not only guided me to the right direction but also continuously supported in overcoming a lot of obstabcles in my research

Second, I would like to thank all of the respondents for spending their time to answer the questions in my survey They contribute significantly in collecting data for my study Without their participation, I am sure that the survey could not be conducted successfully

Finally, my sincere thanks also go to my family and my friends for encouraging me throughout two years

of study as wel as throughout the process of researching and writing this thesis

Thank you

Tran Phuoc Tho

December, 2016

ABSTRACT

In 2015, Vietnam witnessed the booming of airline industry The participation of low cost carriers makes the airline market more and more competitive Understanding the behavior of passengers is essential for any carriers to make their strategic policies

This study employs the multinomial logit model with the data of 122 respondents to investigate the impacts of characteristics of passengers as well as attributes of airlines on the airline choice The characteristics of passengers include age, gender, marital status, education, and income whereas the attributes of airlines consist of price, number of flights of airlines, punctuality, comfort of seat space, and quality of check in service

A stated preference survey is conducted online from 16th to 23rd of October 2016 to collect the data of 122 respondents, who used to travel by air at least one time before They are required to finish three tasks The first task is providing their information, such as age, gender, marital status, education, and income The second one is evaluating about the quality of services of the three airlines, including Vietnam Airline, Vietjet, and Jetstar The final part is hypothetical scenarios of fifteen domestic routes given along with the prices of airlines for the respondents to choose one of the three airlines

Jetstar is chosen as the base outcome, the results of multinomial logit model suggest that characteristics of airlines have relationships with the ratios of probability of chosing Vietnam Airline or Vietjet over probability of chosing Jetstar, except for the satisfaction of customers about staff at the check in counter When comparing one airline and the based airline (Jetstar), the attributes of the third airline is also necessary to be taken into consideration In general, a good judgment of service of an airline makes the odds ratios of that airline and the base increased In contrast, a good evaluation of the based carrier or of the other airline makes the odds ratios declined Besides that, income has positive association with probability of choice Vietnam Airline and Vietjet but negative relation with Jetstar, holding other variables constantly

TABLE OF CONTENTS

Contents

ACKNOWLEDGEMENT I ABSTRACT B TABLE OF CONTENTS: IV LIST OF TABLES VI LIST OF FIGURES VII

INTRODUCTION 1

1.1 Problem statement 1

a Overview of airline industry 1

b Airline industry in Vietnam 1

1.2 Research objectives 3

1.3 Research questions 4

1.4 Scope of the thesis 4

1.5 Structure of thesis 4

LITERATURE REVIEW 5

2.1 Theoretical review 5

a Random Utility Model (RUM) 5

b Reveal Preference & Stated Preference survey 7

2.2 Empirical review 8

RESEARCH METHODOLOGY 13

3.1 Stated preference method 13

3.2 Questionnaire and survey process 14

3.3 Attributes of airlines 16

3.4 Model specification 18

DATA & EMPIRICAL RESULTS 23

4.1 Data 23

4.2 Empirical results 31

a Controlling variables 35

b Attributes of airline 37

c Effect of different routes 38

CONCLUSION 41

REFERENCES i

APPENDIX v

LIST OF TABLES

Table 3.1 Summary of hypothetical scenarios in survey: 15

Table 3.2 Attributes of airline: 17

Table 3.3 Prices and numbers of flights by routes of carriers 20

Table 3.4 Description of variables: 21

Table 4.1 Social demographic characteristics 27

Table 4.2 Estimation results of multinomial logit model 32

LIST OF FIGURES

Figure 3.1 The screen of the online survey 16

Figure 4.1 Airline Choice for Destinations 24

Figure 4.2 Frequency Of Income 25

Figure 4.3 Willingness to pay for routes 26

Figure 4.4 Check-In Service Evaluation 28

Figure 4.5 Cabin Crew Service Evaluation 28

Figure 4.6 Food & Drink Onboard Evaluation 29

Figure 4.7 Inflight Seat Space Evaluation 29

Figure 4.8 On-time Performance Evaluation 30

Figure 4.9 Schedules Delay Evaluation 30

Figure 4.10 Predicted probability of airline choice and income 35

Figure 4.11 Predicted probability of airline choice and age 36

CHAPTER 1

INTRODUCTION

1.1 Problem statement

a Overview of airline industry

In 2015, the world’s aviation industry achieved the highest net profit in history, 33 billion dollars It is nearly double when compared to a net profit of 17.4 billion dollars in 2014 Particularly, the aviation industry in Asia Pacific obtained net profit of more than 5.8 billion dollars In addition, region of Asia Pacific accounted for 31% of global passengers, while Europe and North America is 30% and 26%, respectively It is noted that low cost carrier has transported over 950 million passengers, approximately 28% of those who are scheduled passengers (IATA report, 2016)

According to The International Air Transport Association (IATA), number of air travelers is forecasted to increase nearly double, from 3.8 billion in 2016 to 7.2 billion in 2035 IATA also announces the five fastest growing markets that have the most additional passengers per year for over the next 20 years, including China, US, India, Indonesia, and Vietnam In detail, Vietnam may have 112 million new passengers for a total of 150 million Moreover, IATA also stated that Vietnam is one of the seven countries which have fastest growth in aviation industry Besides that, Vietnam Government pays much attention to infrastructure which is one of the most critical components of air transport sector Vietnam’s planning is to have 26 airports by 2020; particularly Long Thanh International Airport will be ready by 2020

b Airline industry in Vietnam

The Vietnam airline industry, which was administered by Ministry of Transport and Civil Aviation Authority of Vietnam, has witnessed rapid growth in 2015 compared to the figures in

2014 The whole market served 40.1 million of passengers and transported 771 thousand tons of cargo In particular, transportation of domestic carriers is 31.1 million passengers, increased by 21% This positive sign with the falling of crude oil price of 30% in 2015 are stimulus for airline carriers to continue reducing fares in order to meet the demand of transportation of passengers

It could be said that airline industry in Vietnam has a potential market due to many reasons First, population of Vietnam is more than 90 million Thus, demand of traveling is very high Moreover, in the recent years, income of Vietnamese is increasing so that the demand of transit

of people is also higher day by day People have many options to choose means of transports not only faster but also safer Although there are some disasters of airline in 2014 in the world, it seems that traveling by air is the safest way According to IATA Safety Report, there were 12 fatal accidents in the total of 73 accidents, which caused to 641 fatalities on over the world in

2014 This is not a high proportion when comparing to about 33 billion passengers in 2014 (IATA Annual Review 2015) Moreover, air travel helps people save much time For examples,

it takes about two days to transit by train from Ho Chi Minh City to Ha Noi while only two hours

by air Finally, thanks to internet, e-commerce is more and more popular People can stay at home, and buy tickets with the cheap price at the time of promotion of carriers

In 1956, the Government established the Vietnam Civil Aviation Department At that time, there were only five aircrafts to serve some domestic flights In 1993, Vietnam Airlines was set up as a national carrier Until 1995, by gathering 20 aviation enterprises, Vietnam Airlines Corporation was born and the airline itself is the core business Now, Vietnam Airlines is operating an extensive network of domestic and international services to Southeast and North Asia, Europe and Australia In July 2016, ANA Holding Inc became a strategic shareholder after purchasing of

an 8.77% stake Vietnam Airlines claimed that, under the restructure plan, it will keep on to divest the shareholding of state to 75% Skytrax, organization of the leading airline and airport rating of the world, certified that Vietnam Airlines is a 4-star airline

Vietjet Air, an international low cost carrier, was the first privately owned airline in Vietnam Although Vietjet Air was approved to operate in November 2007, it launched the first flight in December 2011, with only 3 aircrafts Up to 2015, Vietjet had 29 aircrafts with 28 domestic routes and 12 international routes As planning of Vietjet in 2016, it will have 42 aircrafts to meet the demand of travel and open more 3 domestic and 5 international fleets

Another airline is Jetstar Pacific Airlines JSC This airline was founded in 1990 as Pacific Airlines and commenced operations in 1991 with charter cargo services under control of Vietnam Airlines Corporation In 2005, it began to operate in passenger service In 2007, Qantas Airway Limited bought a portion of Pacific Airlines’ shares and changed it as model of low cost

carrier It officially became a part of Jetstar network in 2008, named Jetstar Pacific In 2012, Vietnam Airlines purchased a 70% stake, so up to now Qantas is having only 30% stake in the company

Vietnam Air Services Company (VASCO) is one of a subsidiary of Vietnam Airlines Corporation From 2004 to now, VASCO has transported passengers from Tan Son Nhat Airport

to Southern airport such as Ca Mau, Con Dao, Rach Gia, Can Tho and many other routes Besides of service flight, VASCO also plays a role as a multi functioning airline and providing maintenance service for private aircrafts

In summary, there are four domestic carriers are operating in Vietnam at present, including Vietnam Airlines, Vietjet, Jetstar, and VASCO In the past, there were another two airlines used

to operate: Indochina Airlines and Air Mekong Due to difficulty in finances, Indochina Airlines claimed to stop all of the flights after one year in operation in 2009 Similarly, because of loss in business, Air Mekong had to halt commercial flights in 2013 Until January in 2015, it is officially revoked by The Ministry of Transport

There are many literatures about the theory of customer behavior and empirical studies about airline choice of passengers The annual report of IATA (The International Air Transport Association) in 2015 shows the answers of the passengers with the question “What is the first reason for choosing an airline?” It is found that nonstop flight (15%) and lowest fare (14%) are the reasons why customers choose an airline while recommended by travel agent and in-flight service is just accounted for 4% and 3%, respectively However, in Vietnam, airline industry has just been booming in the recent years so there are not many researches focus on this topic Knowing the preference of passengers is necessary for both aviation firms and foreign investors

It helps not only the three carriers have policies that are suitable for Vietnamese people but also investors in evaluate the airline market to make decision in investing or not

1.2 Research objectives

This study uses stated preference survey and employs the multinomial logit model to identify the factors that have impacts on airline choice of passengers These factors include the characteristics of both airline and air travelers This study is expected to provide information on

factors affecting the choice of passengers, and thus provide information for carriers in identifying their target market segments and efficiently improving their services

1.3 Research questions

There are two questions are proposed First, what are attributes of airlines that giving impacts on travelers in deciding which airline to fly? Second, what are demographic factors of air travelers that have influence on their airline choice?

1.4 Scope of the thesis

Although there are four carriers in Vietnam airline market, this research examines the airline choice of three carriers, including Vietnam Airline (VNA), Vietjet (VJ), and Jetstar (BL) VASCO is excluded from the choice set since VASCO just operate in the Southest with short flight, for example from Sai Gon to Ca Mau, Rach Gia, Con Dao Moreover, the main business

of VASCO is providing maintenance service for aircrafts, not transporting passengers Therefore, the market share of VASCO is very small so the elimination of VASCO is not a severe problem

CHAPTER 2

LITERATURE REVIEW

This chapter first introduces the economic literature of individual choice, which is the foundation for empirical studies in analyzing choices of economic agents, including air travellers The chapter then provides a review of empirical studies that analyzed choice of passengers among carriers Based on these reviews, a model is set up to analyze the choices of air travelers among the three airlines: Vietnam Airlines, Vietjet, and Jetstar

2.1 Theoretical review

a Random Utility Model (RUM)

Random Utility Model is commonly used to represent individual choice behavior Thurstone (1927) first introduced a law of comparative judgment and originally developed the terms of psychological stimuli, which leads to the result of binary probit model now This is a model of whether the respondents could get the different level of stimulus The stimuli concept was further developed as utility by Marschak (1960) The random utility model implies that the decision maker may know the utility of each choice alternative but the researcher may not know it fully Therefore, it is necessary to take uncertainty into account This leads to the result that the model

of utility consists of two parts, deterministic and random components Deterministic components could be observed and interpreted by the analyst while random components are unknown There are four main causes of uncertainty that Manski (1977) identified, including measurement errors, the use of proxy variables, unobserved of attributes of the choicer and unobserved attributes of the alternatives

Discrete choice models are based on the random utility theory and other assumptions It is assumed that the decision-makers choose among a finite choice set, which are collectively exhaustive and mutually exclusive alternatives and they select the alternative that brings the highest utility With every alternative, the deterministic factors of utility are stated as a function

of attributes, for example a linear function The probability of selecting an alternative of an individual is the outcome of the choice model Besides that, random components are also the key

factors The difference of assumptions of the distribution of the error terms causes many forms of choice models According to Train (2009), the main models include logit, GEV, probit and mixed logit model

First, logit model is assumed that the error terms is iid extreme value The term of iid means independent, identically distributed (Train, 2009) It is assumed that the unobserved factors are not correlated and have the same variance with alternatives This assumption, on the one hand, is restrictive, on the other hand, makes the choice probability have a very convenient form This convenience makes the logit model used popularly; however, in some situations, the assumption

of un-correlation over alternatives could be not appropriated The sequences of choices over time are also derived under the independence assumption This means that each choice does not depend on the others Thanks to the convenient form, most of the researchers utilize this model

to examine many aspects of air choice behavior (Escobari & Mellado (2014); Warburg (2005); Yoo & Ashford (1996))

Second, to avoid the assumption of independence in logit model, generalized extreme value models or GEV which imply a generalization of the distribution of extreme value were developed (Train, 2009) The generalization allows the relationship of unobserved factors and alternative It could be seen as a special case of logit model when this correlation does not exist The less or more flexibility of the correlations depend on the kinds of GEV model For instance,

a comparatively simple GEV classifies the alternatives in many groups, called nests The unobserved factors are assumed to have the same correlation with alternatives in the same nest but no correlation with ones in the others nests Hess (2008) employs nested logit model to establish model of air travel behavior Pels et al (2001) also use nested logit model to describe the passenger concerning in airports and airlines

Third, probit can deal with three limitation of logit model Train (2009) shows the restrictions of logit model, including not representing random taste variation, IIA property and correlation between unobserved components and alternatives However, probit model assumes that errors terms are normally distributed Therefore, the only limitation of probit model is that, in some cases, unobserved factors may not have normal distribution

Finally, mixed logit permit the unobserved factors to have any distribution In this model, unobserved factors could be divided in two parts One part includes all of the heteroskedasticity and correlation while the other part is iid extreme value It is noted that the first part could obey any distribution, not excluding non-normal distribution Adler et al (2005) apply mixed logit model to develop itinerary choice model The research of Warburg (2005) employs both of multinomial logit model and mixed logit to understand the flight choice behavior of passengers

In reality, there are many other discrete choice models specified for specific purposes by researchers These models are often established by incorporate the concepts of other models For example, a mixed probit could be obtained by breaking down the observed components as in mixed logit, yet, the second part is normal distributed in lieu of extreme value distributed By acknowledging the motivation and derivation of these models, researchers are able to determine the model that is suitable for a specific situation to achieve the goals of their studies

b Reveal Preference & Stated Preference survey

There are two main kinds of surveys which are conducted to analyze the behavior of customers, including revealed preference (RP) and stated preference (SP) survey RP data provide information about the preferences in a real choice environment This brings the primary advantage of RP data, actual behavior of respondent However, it is difficult to do trade-off analysis with RP data (Bhat & Sardesai, 2004) Moreover, for new alternatives introduced in the new market, it could not handle the models with RP data (Whitaker et al, 2005) According to Yoo and Ashford (1996), there are three practical limitations of RP data First, it is not enough variation for some interesting variables to calibrate a statistical model Second, researchers face

to difficulty with estimating model that reflects the trade-off ratios due to the correlations of explanatory variables Finally, to calibrate statistical models, it is necessary to carried out very large surveys to obtain enough observations Therefore, not many researchers employ this method of survey in modeling choice behavior of customers Carrier (2008) use RP data of a booking data so that the study does not include the non-booked travel alternatives, such as income, purpose of travel,…Escobari and Mellado (2014) collect data from the online travel agency and use posted priced and the changes of inventory to explain the demand of flights

In contrast, in SP survey, the hypothetical scenarios are designed to understand the stated responses of the interviewers Thus, SP data could reduce the limitation of RP data According to Collins et al (2012), with SP data, it is possible to reproduce the output of behavior, such as willingness to pay In addition, by conducting SP survey, it is able to explore the choice behavior

of consumers regarding the alternatives that do not exist Nevertheless, SP data has limitation that the respondents may be uninterested or careless in a survey, or may express their own opinions about the context of survey rather than give information about a new product usage (Warburg, 2006) Besides that, decision making in hypothetical situation easily leads to the result

of bias because people may not do as what they say In practical, most of the researchers use SP survey for modeling choice behavior Adler et al (2005) do SP survey to analysis trade-offs in air itinerary choice while Collins et al (2012) use the interactive stated choice survey to investigate the behavior of air travelers Wen and Lai (2010) and Proussaloglou and Koppelman (1999) also use SP data to examine air carrier choice of passengers

In general, due to the full complement of RP and SP data, there are estimation techniques to be developed to combine these data sources to deal with limitation of each type of data It is suggested that the most effective way is to use both of method RP is useful for forecasting demand or realistic purposes while SP is useful for system planning purpose (Yoo & Ashford, 1996) Similarly, to present model of itinerary choice, Atasoy and Bierlaire (2012) use mixed dataset of RP and SP The mixed data enable the study to succeed in estimating elasticity of price

in demand model

2.2 Empirical review

There are several studies that examine all the different aspects of airline choice behavior For instances, the researches of Basar and Bhat (2004), Hess and Polak (2005), and Pathomsiri and Haghani (2005) investigate the airport choice in multi-airport regions Besides that, some papers focus on not only airport choice but also other aspects of travel Ndoh et al (1990) study airport choice and route choice of passengers whereas Furiuchi and Koppelman (1994) examine the passengers’ destination choice and airport choice In addition, there are a few studies pay attention to air traveler choice rather than airport choice, such as the research of Chin (2002), Algers and Beser (2001), Proussaloglou and Koppelman (1999), and Yoo and Ashford (1996)

The multinomial logit model of choice is utilized in most of the studies mentioned above Other studies, such as Ndoh et al (1990), Furiuchi and Koppelman (1994), and Pels et al (2001) use the nested logit model to estimate the multidimensional and spatial choices of air travelers However, the papers that attempt to consider the issues of behavior or effects in air travel choices employ the mixed multinomial logit model (Hess & Polak, 2005; Pathomsiri & Haghani, 2005) Moreno (2006) uses the multinomial logit model to address airline choice for domestic flights in São Paulo There were 1,923 passengers interviewed at the departing lounges of São Paulo-Guarulhos International Airport (GRU) and São Paulo-Congonhas Airport (CGH) It is believed that airline choice is the result of the tradeoff due passengers have to face with flight cost, flight frequency, and performance of airline Thus, three types of variables are tested First, variables associated with cost are the lowest and highest fare The second type of variables is those associated with flight frequency, including the existence of connections or stops, travel period, and the day of the week Finally, age of airline is used to be proxy of performance of airline This study finds that the lowest fare is the best explained variable of airline choice Besides that, senior passengers seem to pay more attention to airline age than junior passengers In the same way, Nason (1981) conducts a stated preference survey to ask respondents to make a choice of airline among a list of airlines By employing multinomial logit model, the research considers airline choice as a function of attributes of airline service as well as characteristics of passengers With revealed preference survey, Prossaloglou and Koppelman (1995) examine airline choice of passengers who depart from Dallas and Chicago in the US In multinomial logit model, independent variables are schedule convenience, reliability, fares, city pair presence, market presence, and frequently flyer program of membership The results show that the attractiveness

of carriers and its market share are positively associated with program of frequently flyer Similarly, Nako (1992) explores the choice of airlines of business travelers as a function of the frequently flyer program of airlines It is concluded that frequently flyer programs affect positively on demand of airline Similarly, Prossaloglou and Koppelman (1999) investigate the passengers’ choice of airline, flight, and fare class by using logit model The authors consider that air travelers are rational decision makers, who tend to choose the alternative brings the highest utility The explanatory variables include fare class, fare price, presence of carrier market, service quality, frequent flyer participation of travelers, and flight schedules Moreover,

the study use separate models to estimate for different groups, such as business and leisure passengers These models are based on stated preference data, which is collected by a two-tier survey First, initial data involving in the characteristics of passengers, such as previous trip, purpose of trip, address, membership of frequent flyer are collected via mail survey Second, a sample of mail survey respondents is chosen randomly to be interviewed by phone The questionnaire is designed to simulate the search of individual for air travel options and their selection among alternatives like during a real process of booking air tickets The results suggest that behavior of leisure and business travelers are significant different Leisure travelers are more price-sensitive but less time-sensitive than business travelers Furthermore, businessmen pay more attention to frequent flyer programs and they are also willing to pay more to fly with their most preferred airlines

In contrast, Pels et al (2001) also use separate models for business and leisure traveler but the results suggest that the difference between two groups is very small The authors utilize the nested logit model to examine the preferences of passengers in concerning airports and airlines

In this research, the nests defined by airports as well as the nest defined by airlines are considered in detail This empirical study use data of the survey in San Francisco Bay Area in

1995 Furthermore, it is implied that an airline has two types of competitors: ones operate in the same airport and the others operate in other airports because access time to the airport are significant for both leisure and business travelers

Besides that, Warburg (2005) says that it is valuable to understand the passengers’ flight choice behavior and predict air travel demand The study helps the carriers give appropriate pricing policy and predict air travel demand in new routes In 2001, Warburg conducted stated preference survey which consists of 119 business and 521 non-business passengers The respondents were passengers who reported their most recent domestic flight They had to make

10 binary choices between the actual flight and hypothesis flight, which was 10 itinerary alternatives with the same departure and arrival place Therefore, Warburg (2005) claims that there is not existence of universal choice set This could be explained that travelers have different flight itineraries so there is ability of different choice set for each passenger The study employs both multinomial logit model and mixed logit model to examine the behavior of two groups of passengers: business and non-business travelers Similarly to the results of Prossaloglou and

Koppelman (1999), in multinomial logit model, the business people seem to be more sensitive to time while non-business ones are more sensitive to fare and men are more sensitive to fare than women

Moreover, the study of Yoo and Ashford (1996) investigates the flight choice behavior of Korean The respondents were who had long distance international air trips, which took more than 10 hours air journey time By employing logit model for both RP and SP data, the researchers also want to do comparative analysis of RP and SP survey Surveys were conducted

at the passenger terminal of Kimpo International Airport in Seoul, in Oct 1993 for RP Survey and in August 1994 for SP survey Total number of samples was equal in RP and SP data The research gives the result that passengers paid more for Korean airline than foreign airline and Korean residents paid more than foreign residents Likely, Escobari and Mellado (2014) estimate the demand of international flights by using a unique dataset with information of flight choices, prices, and characteristics of non-booked flights The data collected from the online agency

“expedia.com”, consist of 317 flights from 6 carriers between 19 and 24 Dec, from New York to Toronto and vice versa The prices and inventory changes for the flights departed from 19 to 24 Dec, 2008 were recorded The study focuses on one way and non-stop flights The findings show that if the price increases 10% in 100 seat aircraft, the quantity demand decrease by 7.7 seats For revealed preference survey, Ukpere et al (2012) investigate the determinants of airline choice making in the Nigerian domestic air transport With the questionnaire follows Likert scale

of ranking, data are collected to obtain both socio-economic characteristics and attributes of airline The socio-economic characteristics include sex, age, marital status whereas the airline attributes consist of comfort, on-board service, fare, frequency, behavior of crew, and power of monopoly These determinants could have effects on passengers in choosing airlines at the selected airports By using the nested logit model, the findings show that all of these variables are significant, that means they effect on making decision of customers The authors also recommend that airline should charge competitive fares and make their products distinct from others to attract more air travelers In constrast, Adler et al (2005) do stated preference survey

on the internet in 2003 to collect the detail information of about 600 individuals who have just paid for domestic air trip The aim of this study is to understand the tradeoffs that an individual faces to when choosing itinerary choices The characteristics of itineraries in the survey include

airline carrier, airport, fare, flight times, on-time performance of carrier, the time difference between the expected arrival time and schedule arrival time By employing mixed multinomial logit model, it is found that the effects of these characteristics of the service are statistically significant However, the limitation of the study of Adler et al (2005) is that it does not examine the effects of characteristics of demographics and trip on the choice of individuals

CHAPTER 3

RESEARCH METHODOLOGY

This chapter presents the research methods, including the identification of the airlines attributes that may affect traveler’s choice, the data collection methods, and the analytical model

3.1 Stated preference method

According to Whitaker et al (2005), stated preference (SP) method, a technique widely used to understand behavior of decision makers, plays an important role in forecasting demand of different types of airlines services Traditionally, analysis of demand employs revealed preference (RP) data, which is choices and decision making take place in actual environment Yet, approaching of RP has some practical limitations First, it is largely associated with costs of survey Moreover, based on RP data, new alternatives, which may be proposed in the market in future, could not be handled in the models

Wen and Lai (2010) suggest that although approaching of reveal preference collects data in real choice, it may be inappropriate since passengers often do not consider carefully all of attributes

of all possible airlines However, the stated choice approach could analyze how individual would respond to hypothetical choice situations, which are comprised a set of alternatives and attributes with their levels Recently, this method has been applied generally in airline choice and other choice problems

It is said that in research of travel behavior, there are two types of stated response (Hensher, 1994) First, a respondent is asked to identify his or her preferences in alternatives This task usually aims to find out a scale of metric, which is a rating scale or a rank ordering scale A rating scale is scale designed to obtain information about both of quantitative and qualitative attributes Likert scale and 1 to 10 rating scale are commonly used in researches However, a rank ordering scale has a little bit of difference With a rating task, individuals are able to order alternatives that listed so it could give the view of their degrees of preferences The study of Warburg et al (2006), Adler el al (2005) are typical examples of using rank ordering scale in

survey of airline choice Second, a respondent is required to take one of the listed alternatives This is named as first preference choice task It is important to address types of response strategy

at the beginning of conducting an SP survey since it defines the outputs

The survey of this study applies the first preference choice task In this task, based on the airfares

of airlines for a specific route, each respondent is required to choose one of three airlines: Vietnam Airlines (VNA), Vietjet Air (VJ), and Jetstar (BL) According to Hensher (1994), SP data has an appealing feature that is ability to view the stated response as the counterpart of reveal preference It is because in reality, individuals decide to select one option after considering a set of alternatives carefully Many researchers utilize this method in their studies, such as Wen & Lai (2010), Hong (2010) In the SP survey of Wen & Lai (2010), air travelers face to a choice set of three carriers: China Airlines, EVA Airways, and JAA for Tapei – Tokyo route whereas four airlines: China Airlines, EVA Airway, Cathay Pacific, and Dragon for Tapei – Hong Kong route Similarly, Hong (2010) conducts an SP survey which the task of respondents is select one of three airlines: British Airways, Air France, and Easyjet

3.2 Questionnaire and survey process

The questionnaire of this survey that is showed detail in the Appendix consists of three parts The first section is the questions about social demographic information and primary purpose of trip In the second part, respondents evaluate the quality of services of airlines, including attitude

of staff at check in counter, attitude of flight attendants, in-flight food and drink, seat space, and on-time performance For carriers that they have never had experience, there is an available choice for them “I have never used this service before” Finally, fifteen hypothetical situations are presented Each case is a specific route that departs from Tan Son Nhat Airport to others 15 domestic airports, is presented in Table 3.1 The hypothetical scenario is that if an individual has travel by air, with the airfare as listed, which airline he or she could choose In addition, respondents also reveal their possible purpose of trip and the highest price that they willing to pay for a ticket of each route However, if respondents think that they would never go to one place in future, they could choose option as “I will never go there” and skip the remaining questions to move to the new situation

Table 3.1 Summary of hypothetical scenarios in survey:

Vietnam Airline Vietjet Jetstar

Note: x: having at least one flight in a day

At the beginning, a pilot-test was conducted at the air ticket agency to identify the determinants that have influence on decision of customers in purchasing air tickets There were 18 customers, who had just bought air ticket, were asked to list all of the possible factors affect their choice These factors consist of fare, schedules, on time performance, quality of staff service, and comfort of seat onboard After that, this survey was proceeded online from 16th to 23rd of October 2016 SurveyMonkey, online survey development software is employed to design the questionnaire Figure 3.1 is the print screen of the online survey The link to access this questionnaire is sent to air travelers via mail as well as posted public on social media network, such as Facebook and Zalo The target respondents are those who have traveled by air before and used to fly with at least one of airlines: VNA, VJ, and BL It is noted that they may not have experiment with all of three airlines Because of loyalty of customers, some people tend to use only service of their favorite airline Therefore, it is not easy to find out a person who has chances to fly with all of three airlines

Figure 3.1 The screen of the online survey

3.3 Attributes of airlines

In service industry, customer satisfaction is one of the most important determinants of customer’s retention Fornell et al (1994) suggest that customer satisfaction brings benefits because it means that the company gets back fewer complaint, thus it make the cost of dealing with failure also decreased According to Zeithaml and Bitner (1996), satisfaction of customers

is made up of a number of factors, including price, term and condition, quality of products and services, and personal characteristics Quality of service is not the only crucial factor of customer loyalty since customers usually have trend to consider trade-off relationship between costs and benefits (Lee & Cunningham, 1996) Hence, price is also a key factor of consumer satisfaction Furthermore, the research of Athanassopoulos et al (2001) imply that customer behavior responds to customer satisfaction in one of the three ways, including staying with the existing providers, participating in word-of-mouth communicating, or changing service providers In airline industry, beside of price and quality of service, there are many determinants that affect on selection of airlines, including schedule time, frequency of flights, aircraft types, number of seats

of airline Many researchers investigate the effects of airline attributes on carrier choice, which are summarized as in Table 3.2

Table 3.2 Attributes of airline:

Attributes of

airline

Price Cost of a route (return fare) Continuos data Warburg (2005)

Cost of a route (one-way fare) AUD1600,

AUD1900, AUD2200, AUD2500

Collins & Hess (2012)

Average fare for each route Higher price;

Medium price; Lower price

Wen & Lai (2010)

Fare of the chosen flight Continuos data Adler et al (2005) Frequency of airline Number of flights/route/day Wen & Lai (2010)

Number of direct flights in the travel day

Moreno (2006)

Number of flights per week Yoo & Ashford

Wen & Lai (2010)

Percentage of on time flight itinerary 50%-99% Adler et al (2005)

On time service schedules Sometimes delay,

Always consistent

Hong (2010)

Seat space on board Seat pitch 31", 32", 34" Collins & Hess

(2012) Passenger's evaluation of seat Very uncomfortable

Comfortable enough Very comfortable

Wen & Lai (2010)

Check in service Passenger's evaluation of check in

service

Very uncomfortable Comfortable enough Very comfortable

Wen & Lai (2010)

Kindness of employees Not very polite and

friendly Very polite and friendly

Hong (2010)

3.4 Model specification

This study follows the framework of Random Utility Model of Manski (1977) since air travelers are assumed to be rational to maximize their utility Passengers tend to select the carrier that brings them the highest utility which has the form as below:

𝑈𝑖𝑛= 𝑉𝑖𝑛+ 𝜀𝑖𝑛= 𝛼𝑛+ 𝛽𝑛𝑋𝑖 + 𝜀𝑖𝑛Where U: Utility level of passenger

V: Portion of utility (observed utility), and 𝑉 = 𝛼 + 𝛽𝑋

𝜀 : Error terms (unobserved utility)

X: vector of explanatory variables

i : Passenger i

n = 1, 2, 3 denoted for Vietnam Airline, Vietjet, and Jetstar, respectively

It is reasonable to assume that the actual of choosing airline n is 𝑌𝑖𝑛, so:

(Gumbel distribution) According to Train (2009), the function of probability could be rewritten

as the base, and set α3 = 0 and β3 = 0, the probabilities of three airlines could be obtained:

In this model, it is noted that X is the vector of variables, including independent variables and controlling variables, are investigated whether they have the relationships with carrier choice or not Independent variables are composed of price, frequency of flights, and routes These factors are mentioned as hypothesis in the third part of the survey for respondents making decision Controlling variables which consist of information of respondents and their evaluation of airline service are collected through the first and the second section of the questionnaire The list of variables used in this study is described in Table 3.4

Notably, the price and frequency of flights are collected online base on the real business of airline This survey uses the price as the average price of each route of each airline in November

2016 whereas the frequency of flights is the actual number of flights that each carrier has in a day Table 3.3 is the summary of value of independent variables

Table 3.3 Prices and numbers of flights by routes of carriers

Route From Sai Gon To Price (100,000 VND) Number of flights

Vietnam Airline Vietjet Jetstar Vietnam Airline Vietjet Jetstar

Table 3.4 Description of variables:

Type of variable Variables Denotation Unit Description

freqvn Number of flights of a route in

Marital status single (Dummy) 1 = Single Education schoolyear Years Number of schooling years Income income Million VND Average income per month Occupation job_emp (Dummy) 1 = company employees

On time performance

ontvn_pun (Dummy) 1 = VNA previous flights

Type of variable Variables Denotation Unit Description

CHAPTER 4

DATA & EMPIRICAL RESULTS

This chapter presents the summary of data collected form the stated preference survey as well as the regression results of multinomial logit model

4.1 Data

The characteristics of social demography is describes in details as Table 4.1 Actually, there are135 respondents but 13 of them did not finish the questions so they are eliminated out of the data Each individual faces to 15 scenarios to select the airline Figure 4.1 shows the choice of airline in every hypothesis destination in the survey It seems that Ha Noi, Da Nang, and Phu Quoc are destinations that have the highest demand of travel because more than 98 percent of respondents think that they will go there in future VNA has the lowest percentage of choosing airline in all of routes Notably, there is no choice of VNA for the route of Sai Gon – Tuy Hoa and Sai Gon – Chu Lai since VNA does not have any flight in reality Therefore, in this survey, scenarios of Tuy Hoa and Chu Lai, VNA is excluded In general, VJ has the highest percentage

of choosing airline in most of scenarios Especially, for the route of Sai Gon to Ha Noi, and Da Nang, more than 60 percent of respondents select VJ However, in some cases such as Da Lat and Hue, BL is the airline that has the highest percentage of choice

Figure 4.1 Airline Choice for Destinations (Notes: HAN: Ha Noi, DAD: Da Nang, VII: Vinh, CXR: Nha Trang, DII: Da Lat, HUI: Hue, THD: Thanh Hoa, BMV: Buon Me Thuot, PXU:

Pleiku, PQC: Phu Quoc, HPH: Hai Phong, TBB: Phu Yen, UIH: Quy Nhon, VDH: Dong Hoi, VCL: Chu Lai)

There are 122 respondents in the data includes 84 females (68.85%) and 38 males (31.15%) (Table 4.1) Most of them are single, approximately 70% Besides that, this survey is conducted online so the respondents are young people who have ability to access internet easily Thus, the medium age of them is around 27 while the youngest is 20 and the oldest is 46 It is also noted that there are two people not reveal their age in this survey In addition, 83 respondents (68%) graduate from university and 32 (26.23%) have a master degree whereas only one person finished high school Moreover, it could be seen that a number of people in this survey are employees who are working for companies (83 respondents, about 68%) In contrast, just one person is freelancer or self-employee For the purpose of trip, more than 40 percent of them said that they usually travel by air for both of leisure and business Furthermore, the level of income that seems most of respondents reach in the period of survey is from 8 to 10 million per month (Figure 4.2)

Figure 4.2 Frequency Of Income

In the survey, people are asked to evaluate the quality of service of airlines, including check-in service, attitudes of cabin crew, quality of food and drink onboard (Figure 4.4 to 4.7) These services are rated with three scales: not good, good enough, and very good In addition, to judge the delay of schedules of flights, the respondents are asked two questions First, their previous flights of three airlines they ever traveled are punctual or not (Figure 4.8) Second, it is whether they can accept the on time performance of airlines (Figure 4.9) However, for those who have never used the services of any airline, there is an option for them: “I have not used the service before”

In general, approximately 36 percent of people in this survey said that they have not flied with Jetstar Pacific before, the highest rate among of three airlines This contrasts with Vietnam Airlines Besides that, most of people appraise that the quality of service of Vietnam Airlines is good and very good Specially, very few people judge that check-in and cabin crew service of Vietnam Airlines are unfriendly (just 1.64% and 0.82%, respectively) About punctuality, it seems that almost respondents agree that flights of Vietnam Airlines depart on time (about 60%) and around 74% think that the rate of delayed flights of Vietnam Airlines is acceptable Yet, Vietjet Air has the highest number of unpunctual flights (up to 62%) and more than 30% says that it is unacceptable for schedule changes of Vietjet Air

In addition, this survey also inquires the respondents to reveal how much they are willing to pay for an air ticket of each route, for example from Sai Gon to Ha Noi Figure 4.8 presents the mean

0 5 10 15 20 25

of willingness to pay for routes of the respondents It is clearly that people are pleased to pay more for long distance trips (Ho Chi Minh to Ha Noi, Hai Phong) and pay less for short distance trip (Ho Chi Minh to Nha Trang)

Figure 4.3 Willingness to pay for routes (Notes: HAN: Ha Noi, DAD: Da Nang, VII: Vinh, CXR: Nha Trang, DII: Da Lat, HUI: Hue, THD: Thanh Hoa, BMV: Buon Me Thuot, PXU: Pleiku, PQC: Phu Quoc, HPH: Hai Phong, TBB: Phu Yen, UIH: Quy Nhon, VDH: Dong Hoi, VCL: Chu Lai)

Table 4.1 Social demographic characteristics

Notes: 122 respondents, in which 2 not reveal their age

Demographic Characteristics Number of respondents Percentages (%)

Figure 4.4 Check-In Service Evaluation

Figure 4.5 Cabin Crew Service Evaluation

Figure 4.6 Food & Drink Onboard Evaluation

Figure 4.7 Inflight Seat Space Evaluation

Figure 4.8 On-time Performance Evaluation

Figure 4.9 Schedules Delay Evaluation

4.2 Empirical results

According to Gujarati (2011), the odds could tell how much this choice is preferred over that

choice The odds ratios are defined as the ratio of the probability of choosing alternative i and the probability of choosing alternative j, which is the base outcome Besides that, the positive value

of a coefficient suggests that a raise in this variable will increase the odds for choice i over choice j, when holding others variables constant This indicates that choosing i increase the utility of decision maker in comparing with choosing j Reversely, a negative coefficient of a regressor means that if this variable increases a unit, the odds for choice i over choice j will decrease, when holding others regressors constant It implies that choosing j is preferred than choosing i

In multinomial logit model, the relative risk ratios (rrr) could be obtained by exponentiating the multinomial logit coefficients, e coef The meaning of relative risk ratios is that for a unit change

in independent variable, the relative risk ratios of outcome j over the base outcome is expected to

change by a factor of that parameter

Table 4.2 reports the results of the multinomial logit models Model 1 is the regression for only controlling variables Model 2 adds two attributes of airlines including price and frequency of airline whereas mode 3 adds routes to see the effects of these factors In these models, the choice

of Jetstar Pacific (choice 3) is used as the reference It is noted that choice 1 and choice 2 are denoted for choice of Vietnam Airlines and Vietjet Air, respectively

It is noted that there are 122 respondents who are required to make decision in 15 scenarios Each choice in each scenario is considered as one observation In other words, if all of respondents answer all of 15 scenarios, there would be 1,830 observations However, in the survey, if the respondents say that they do not have demand to travel from Sai Gon to one place, they are not asked to make choice of airline in that scenarios Therefore, there are 605 observations when running the regression The results are presented in detail in the Appendix