Choo Etal 2013 Exploring Characteristics of Airport Access Mode Choice A Case Study of Korea

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

This article was downloaded by: [UQ Library]

On: 24 May 2013, At: 02:55


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Transportation Planning and Technology


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gtpt20

Exploring characteristics of airport


access mode choice: a case study of
Korea
a b c
Sangho Choo , Soyoung (Iris) You & Hyangsook Lee
a
Department of Urban Design & Planning , Hongik University ,
72-1 Sangsu-Dong, Mapo-Gu , Seoul , 121–791 , Korea
b
Institute of Transportation Studies , University of California ,
Irvine , CA , 92697–3600 , USA
c
Department of Civil and Environmental Engineering Rutgers ,
The State University of New Jersey , 100 Brett Road, Piscataway ,
NJ , 08854 , USA
Published online: 23 May 2013.

To cite this article: Sangho Choo , Soyoung (Iris) You & Hyangsook Lee (2013): Exploring
characteristics of airport access mode choice: a case study of Korea, Transportation Planning and
Technology, DOI:10.1080/03081060.2013.798484

To link to this article: http://dx.doi.org/10.1080/03081060.2013.798484

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-


conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
Transportation Planning and Technology, 2013
http://dx.doi.org/10.1080/03081060.2013.798484

Exploring characteristics of airport access mode choice:


a case study of Korea
Sangho Chooa*, Soyoung (Iris) Youb and Hyangsook Leec
a
Department of Urban Design & Planning, Hongik University, 72-1 Sangsu-Dong, Mapo-Gu,
Seoul 121791, Korea; bInstitute of Transportation Studies, University of California, Irvine, CA
926973600, USA; cDepartment of Civil and Environmental Engineering Rutgers, The State
University of New Jersey, 100 Brett Road, Piscataway, NJ 08854, USA
(Received 9 March 2011; accepted 12 October 2012)

This paper explores air passengers’ behavior of airport access mode choice and
Downloaded by [UQ Library] at 02:55 24 May 2013

develops airport access mode choice models for two major domestic airports in
Korea, Gimpo Airport and Daegu Airport. First, descriptive analyses are
conducted to identify whether the key explanatory variables such as trip purpose,
travel time, and travel cost are statistically different among airport access modes.
Second, logistic regression models are proposed for three types of airport access
mode choices: auto and public transit, auto and bus, and auto and subway. The
model results indicate that the choice of airport access mode is significantly
affected by travel characteristics and sociodemographic variables. This study also
estimates access mode choice models for business and nonbusiness travel,
indicating that access mode choice is significantly more affected by demographic
characteristics of nonbusiness travel air passengers than those for business travel
passengers.
Keywords: airport access mode; logistic regression model; mode choice behavior;
nonbusiness travel; Korea

1. Introduction
In the 1990s, Korea built several international airports to attract foreign travelers and
to support Korean international travel demand resulting from regional growth.
Unlike other forms of transportation, air travel is preferred for long distance trips,
and the journeys follow relatively planned schedules. Air travelers usually choose
airports after considering various factors such as air fare, frequency of flights or
flight schedule, departure time, access distance, and modes to the airport. Thus, new
airports failed and closed as a result of being uncompetitive, leaving investors with
huge amounts of debt. On the other hand, Gimpo Airport, in the capital city of
Seoul, experiences high demand and suffers from auto traffic congestion on major
access roads to the airport.
The prediction of air passenger mode choice decisions for travel to and from
airports is essential in designing and operating airport facilities, as well as managing
airport access traffic. In some cases, we want to understand the consequence of
decisions such as changes in the price or service pattern of ground transportation

*Corresponding author. Email: shchoo@hongik.ac.kr


# 2013 Taylor & Francis
2 S. Choo et al.

service. Mode decisions could include changes in parking rates or the introduction of
new transportation systems and routes. For example, we had difficulty in predicting
airport access mode demand when Incheon Airport Rail and Subway 9 were opened
in 2009 and 2010, because characteristics affecting demand for airport travel mode
choice were unknown. Finally, escalating interests in environmental issues have been
leading research on how to predict air quality impacts of airport access modes. It is
necessary to be able to control or mitigate total emissions in airport regions.
However, there is no general approach to model how air passengers will choose their
access or egress mode in response to changes in the airport ground transportation
system.
The development of airport access mode choice models has been ongoing for
more than 30 years. Multinomial Logit (MNL) models and more complex Nested
Logit (NL) models have been commonly employed. However, it is not clear which
model is more appropriate. Selecting appropriate explanatory variables to account
for the role of traveler income, age, gender, occupation, and trip purpose in the mode
choice process have not yet been addressed. In addition, the factors affecting airport
Downloaded by [UQ Library] at 02:55 24 May 2013

travel are significantly different from those for other transportation models,
considering travel cost and travel time as key factors. For instance, assume that an
air traveler has excess luggage and wants to enjoy a comfortable journey; s/he would
most likely choose a car or a taxi rather than public transit. That is, other factors
such as comfort and amount of luggage can influence airport access mode choice.
Given the importance of understanding airport ground access mode choice, it is
not surprising that a vast body of literature on ground access mode choice began as
early as the 1970s. As shown in Table 1, most conventional studies have focused on
analyzing the choices made for airports and travel to and from airports.
In one of the earliest efforts, Ellis, Bennett, and Rassam (1974) attempted to
model airport access mode choice using a MNL model, which was followed by
similar studies over the next ten years (e.g., Sobieniak et al. 1979; Gosling 1984;
Harvey 1986; Skinner 1976). Sobieniak et al. (1979) developed the MNL model for
access mode choice for intercity terminals, including airports, using the 1978 travel
survey data in Ottawa-Hull and vicinity, Canada. The set of airport access modes
considered in this study are automobile driver, automobile passenger, taxi, and
limousine. The model results indicate that convenience-related variables such as
walking time and baggage handling are dominant factors that affect access mode
choice. Skinner (1976) also employed the MNL model for airport choice in the three
airports: Friendship International Airport, Washington National Airport, and
Dulles International Airport. The study concluded that flight frequency and airport
accessibility were essential determinants of airport choice. Similar investigation was
conducted by Harvey (1986), who developed a MNL model of airport access mode
choice (drive, drop-off, transit, airporter, and taxi) using data from the 1980 survey
of departing air passengers in the San Francisco Bay Area. His study indicates that
travel time and travel costs have a significant effect on airport access mode choice.
To address the limitations of MNL models, some studies starting in the mid-
1980s proposed NL models (Ben-Akiva and Lerman 1985). One of the first
applications of NL models to airport access mode choice was conducted by Harvey
(1988), in which an integrated model of airport choice and ground access mode
choice for San Francisco Bay Area was developed. Subsequent air traveler mode
choice models used a nested structure: Boston, Massachusetts (Harrington 2003;
Table 1. Summary of airport access mode choice models.

Ellis, Bennett, and


Downloaded by [UQ Library] at 02:55 24 May 2013

Reference Rassam (1974) Skinner (1976) Sobieniak et al. (1979) Gosling (1984) Harvey (1986)

Transportation Planning and Technology


Data location US (1966, Baltimore  US (Baltimore and Canada (Ottawa-Hull US (1980, San US (1980, San
(year) Washington airport) Maryland area) and vicinity) Francisco Bay Area) Francisco Bay Area)
Sample size Unspecified 16,000 1742 Unspecified 1850
Model type MNL model MNL model MNL model MNL model MNL model
Airport Private car Automobile driver Auto (park) Drive
access mode Rented car Only consideration of Automobile passenger Auto (drop-off) Drop-off
Taxi access time Taxi Limousine Transit
Limousine Limousine Airporter Airporter
BART/taxi Taxi
Level of Service
Cost Cost - Travel time
Line-haul time Driving time - Auto access to transit
Significant Purpose (business) Waiting time Access time Travel Cost
explanatory Travel cost Flight frequency Walking time Line-haul time - Drop-off passenger time
variables Travel time Airport accessibility Baggage Waiting time Traveler Characteristics
Female Air party size - Luggage
Purpose (business) Household income - Household size
Household income - Departure from home

3
Table 1 (Continued)

4
Pels, Nijkamp, and Rietveld Gupta, Vovsha, and Donnelly
Reference (1998, 2003) Basar and Bhat (2004) Tam and Lam (2005) (2008)

Data location (year) US (1995, San Francisco US (1995, San Francisco Hong Kong (2004, HKIA) US (2005, New York and New
Bay Area) Bay Area) Jersey Area)
Sample size 21,124 1918 498 19,127
Model type Two level NL model PCMNL model Structural equation model NL model
Downloaded by [UQ Library] at 02:55 24 May 2013

- Airport (upper level)


- Access mode (lower level)
Airport Upper level:
access - San Francisco Airport Public modes Auto drop-off
mode - San Jose Airport - Franchised bus Auto-Park
- Oakland Airport - AEL Taxi and limos
Lower level: Only consideration of Nonpublic modes Shared vans, shared limos,

S. Choo et al.
- Private car, rental car access time - Taxi and hotel courtesy vehicle
- Private schedule - Private car Rental cars
- Public transit - Tourist bus Rail
- Door-to-door - Hotel car Local buses
- Hotel courtesy, taxi - Other Chartered buses
- Limousine
Male Travel time
Age Travel cost including transit
fare
Significant explanatory Average fare Education Daily airport parking rate
variables
Access time Access time Long-haul flight Resident or visitor
Access cost Flight frequency No. of transfers Gender
Cost Age group
Air passengers’ expectations Household income group
and perceptions
Travel party size
Transportation Planning and Technology 5

Harrington et al. 1996), Portland, Oregon (Portland Metro 1998), and the South
East and East of England Regional Air Service (SERAS) (Halcrow Group Ltd.
2002). The Boston Logan International Airport model developed separate sub-
models for resident business trips, resident nonbusiness trips, nonresident business
trips, and nonresident nonbusiness trips. Among the sub-models, the two resident
sub-models used a NL model, with four shared-ride public modes at the top level and
separate nests for door-to-door modes (taxi and limousine) and automobile modes
(drop off, short-term parking, long-term parking, and off-airport parking). Soon
after developing the Boston Logan model, Portland updated the initial MNL models
with estimated NL models. The SERAS airport access model used NL models with a
multilevel tree to account for the complex pattern of 12 predefined ground access
modes.
In addition, a number of studies have used two-level NL models to represent
airport choice as a high-level nest with airport access transportation choice as a lower-
level nest. Pels, Nijkamp, and Rietveld (1998) introduced a combined airport-airline-
access mode choice model for residents who lived in the Bay area. Two trip purposes,
Downloaded by [UQ Library] at 02:55 24 May 2013

business and leisure, were considered. The proposed two-level NL model included
airport and access mode choice at the upper level and airline choice at the lower level.
In another study, Pels, Nijkamp, and Rietveld (2003) suggest a two-level NL model of
airport choice (SFO, SJC, and OAK) at the upper level and access mode choice
(private car, rental car, private scheduled, public transit, door-to-door van, hotel
courtesy, taxi, and limousine) at the lower level, using the 1995 Bay Area Airline
Passenger Survey data. This study implies that leisure travel is more influenced by
(access) travel cost, while business travel is more affected by (access) travel time. Both
studies show that airport access mode choice is more sensitive to access travel time
than access travel cost. Besides the abovementioned models, several other studies
employed two-level NL models (Bondzio 1996; Monteiro and Hansen 1996). However,
these models are equivalent to MNL models from the perspective of ground access
mode choice, since they have a single-level nest for the mode choice process (Gosling
2006). The current Korean airport system consists of only a few major airports and
minor airports. Therefore, the airport choice of air travelers in Korea is limited,
making a two-level NL model inappropriate for the Korean airport system.
Although NL models have been successfully applied in a number of studies
throughout the world, some studies continue to use MNL models to address airport
ground access mode choice (Tambi and Falcocchio 1991; Dowling Associates Inc.
2002; Psaraki and Abacoumkin 2002). The San Jose model was developed for light
rail line ridership using air passenger surveys in 1995. Similar to the Boston model,
the same four market segment (resident business trips, resident nonbusiness trips,
nonresident business trips, and nonresident nonbusiness trips) and MNL model with
seven modes (private car, rental car, scheduled airport bus, door-to-door shuttle van,
taxi, public transit bus, and light-rail access) were adopted. Explanatory variables
consisted of travel time by each mode, waiting time, walking distance, and cost.
Basar and Bhat (2004) proposed a Probabilistic Choice Set Multinomial Logit
(PCMNL) model for airport choice studies, which they applied to the San Francisco
Bay Area. The statistical evaluation indicated that the PCMNL model outperforms
the MNL model for an estimation sample and a validation sample. Gupta, Vovsha,
and Donnelly (2008) applied the two-level NL models of Pels, Nijkamp, and Rietveld
(1998, 2003) to New York and New Jersey areas. This study covered nine airports
6 S. Choo et al.

and 28 county flight lines based on 19,127 observations in 2005. Although the two-
level NL model was formulated in this study, a MNL model was found to be
statistically preferable. Overall, their finding was that traveler behavior was
significantly different for business versus nonbusiness purposes.
In the most recent studies, Tam and Lam (2005) developed access mode choice
models for the Hong Kong International Airport using Structural Equation
Modeling (SEM). They found that the choice of the access mode was determined
by several factors including level of service for each mode (travel time and
punctuality), gender, age, the level of education, frequency of airport visit, and
travel cost. In addition, Tam, Lam, and Lo (2010) attempted to incorporate
passengers’ perceived service quality in their airport ground access mode choice
model. In the study, a latent variable model is incorporated by the use of a Multiple
Indicators and Multiples Causes (MIMIC) model.
Considering the many different factors affecting airport passenger mode choice,
it is unrealistic to determine the impacts of airport ground transport improvements
or to create effective management strategies, without an appropriate mode choice
Downloaded by [UQ Library] at 02:55 24 May 2013

model that reflects regional characteristics. This study explores air passengers’
behavior of airport access mode choice, and develops airport access mode choice
models for two major domestic airports, Gimpo Airport and Daegu Airport, using
data from the 2004 Airline Passenger Pilot Survey in Korea.
As shown in the reviewed literature, identifying explanatory variables is a critical
process in developing airport ground access mode choice. First, Chi-squared and
analysis-of-variance tests were conducted to determine the key explanatory variables
among trip purposes, travel time, and travel cost. Test results show whether or not,
trip purpose, travel distance, travel time, travel cost, age, and income are significantly
different across airport access modes. Second, logistic regression models are
proposed for three types of airport access mode choices: (1) auto and public transit,
(2) auto and bus, and (3) auto and subway. Logistic regression analysis is employed
to investigate the relationship between binary or discrete responses and a set of
explanatory variables (Agresti 1996). Several characteristics of logistic regression are
important to mention in the context; (1) the independent variables do not have to be
normally distributed, (2) the regression can handle nonlinear relationships, (3)
explicit interaction and power terms can be added, and (4) factor scores are
estimated by assigning a value of zero to the dummy variable when the corresponding
factor score was negative, otherwise the dummy has a value of one. Logistic
regression models contain the necessary and significant variables for access mode
choice for nonbusiness travel and for business travel.
The paper is organized as follows: Section 2 discusses the characteristics of the
data and Section 3 presents the descriptive analyses relating access mode choice to
trip purpose, travel time, travel distances, travel costs, and demographic variables
individually, using Chi-squared tests and one-way analysis of variance. The fourth
section presents final empirical results for the specified models. Conclusions are
discussed in the final section.

2. Data characteristics
The 2004 Airline Passenger Pilot Survey in Korea was used as the data source for this
study. The survey focused on departing air travelers in November 2004 at six major
Transportation Planning and Technology 7

international and domestic airports. Two of the major domestic airports, Gimpo
Airport and Daegu Airport, were selected for this study, because both these airports
are accessible by subway. Gimpo Airport is located in west of Seoul, the capital city,
and Daegu Airport is in the center of Northern Gyeongsang Province in the
southeastern part of Korea. The annual number of departing passengers of Gimpo
Airport and Daegu Airport is around seven million and 0.7 million, respectively.
A total of 1588 and 702 effective samples from Gimpo Airport and Daegu
Airport were obtained from the survey. Table 2 illustrates the key sociodemographic
and trip characteristics of the sample: gender, age, occupation, income, trip purpose
and airport access mode. For the two airports, over 60% of the respondents were
male, with Gimpo Airport showing a slightly higher percentage (65.8%) than Daegu
Airport (60.8%), as depicted in Figure 1.
Air travelers were mainly between the ages of 20 and 40 years, which comprises
86.0% of the traveling population at Gimpo Airport and 82.2% at Daegu Airport. At
both airports, the largest age group of travelers was around 30 years of age. Figure 2
shows the age distribution of respondents.
Downloaded by [UQ Library] at 02:55 24 May 2013

The majority of the respondents were in professional/technical or administrative/


managerial occupations. At Gimpo Airport, 31.2% of the respondents had
professional/technical jobs and 23.4% had administrative/managerial jobs, showing
a distinctive difference from other occupations which exhibited percentages ranging
from 0.9% to 13.9%. In the case of Daegu Airport, professional/technical jobs
(26.2%) made up the largest portion, followed by administrative/managerial jobs
(18.7%), housekeepers (15.7%), and students (15.5%). In terms of annual household
income, 68.1%70.0% of travelers were in the range of US$20,000US$50,000.
Hence, the sociodemographic distributions of both airports seem to be similar.
As for trip purpose, 35.8% (Daegu Airport) and 41.1% (Gimpo Airport) of
respondents were traveling for business, and 27.5% (Gimpo Airport) and 43.4%
(Daegu Airport) were traveling for leisure. These percentages indicate that more than
two-thirds of air travel is generated for business and leisure purposes. Travelers at
Gimpo Airport showed a relatively high percentage of returning or homecoming
trips (17.4%). Figure 3 depicts the trip propose distribution by airport.
To access Gimpo Airport, regular buses and limousine buses were used most
frequently, at a rate of 31.7%. Car (27.4%) and subway (25.1%) modes were also used
as main access transportation modes. On the other hand, most passengers who
depart from Daegu Airport used cars (43.2%) or taxi (33.5%) to get to the airport.
This is not surprising, since the subway system in the Seoul region where Gimpo
Airport is located (total length is around 280 km and the subway is directly
connected to the airport) is much bigger and better connected than in the Daegu
region (total length is around 26 km and the nearest station is about 1km away from
the airport). Figure 4 shows the airport access mode distribution.

3. Descriptive analysis
This study examines if key variables such as trip purpose, travel time, travel cost, age
and income are significantly related to airport access mode choice. Analysis of
variance (ANOVA) and Chi-squared tests were conducted in order to identify
statistical differences among access modes. The ANOVA test was used for continuous
8 S. Choo et al.

Table 2. Sample characteristics.

Gimpo Airport (Unit: %) Daegu Airport (Unit: %)

Characteristics
Number of respondents 1588 702
Gender N1581 N 702
Male 65.8 60.8
Female 34.2 39.2
Age N1576 N 702
19 or younger 1.3 10.7
2029 28.8 30.1
3039 34.5 30.2
4049 22.7 21.9
5064 11.7 14.1
65 or older 1.0 3.0
Occupation N1580 N 701
Student 8.9 15.5
Downloaded by [UQ Library] at 02:55 24 May 2013

Housekeeper 11.8 15.7


Professional/technical 31.2 26.2
Manager/administrator 23.4 18.7
Sales/service 8.4 7.7
Agricultural 0.9 2.4
Manufacture/transportation 1.7 2.7
Other 13.9 11.0
Annual household income (US$) N1579 N 699
Less than $20,000 7.5 0.7
$20,000$29,999 20.0 30.1
$30,000$39,999 27.8 30.2
$40,000$49,999 20.3 21.9
$50,000$59,999 11.0 14.1
$60,000 or more 13.4 3.0
Trip purpose N1588 N 702
Leisure 27.5 43.4
Return (homecoming) 17.4 8.4
Business 41.1 35.8
Commute/school 0.6 0.1
Visit relatives/entertainment 9.7 9.5
Shopping 0.0 0.1
Other 3.7 2.6
Airport access mode N1588 N 702
Car 27.4 43.2
Taxi 15.9 33.5
Regular bus 12.0 13.7
Limousine bus 19.7 1.3
Subway 25.1 7.5
Other 0.0 0.9

or quasi-continuous variables such as travel distance, time and cost, age and income,
while the Chi-squared test was employed for the categorical variable of trip purpose.
Airport access modes were classified into four types (car, taxi, bus, and subway),
combining ‘regular bus’ and ‘limousine bus’, and excluding ‘other’ due to the
Transportation Planning and Technology 9

Figure 1. Distribution of respondents by gender.

insufficient sample size. To appropriately perform the Chi-squared test, the cross-
tabulation table must be reliable (that is, each cell in the cross-tabulation table should
have an expected count of five or more). To ensure reliability, two categories of trip
Downloaded by [UQ Library] at 02:55 24 May 2013

purpose (commute/school and shopping) are combined into ‘other’.


The highest access mode percentages are: leisure and return trips, 32.5% and
39.1% of the respondents in Gimpo Airport used buses, and for business, visiting
relatives and other purposes, 28.8%, 31.8%, and 33.8% used cars, respectively. Bus,
car, and subway users represent more than 20.0% for most trip purposes. However,
return trips by car represented only 16.7% of the total trips. On the other hand, air
travelers using Daegu Airport appear to have different mode choice patterns. The
percentages of car use were significantly high for all purposes with a range from
42.1% to 60.0%, except for return trips, which showed more frequent use of taxis
(39.7%). Table 3 provides a cross-tabulation table for access mode distribution by trip
purpose.
ANOVA tests were used for comparing the means of travel distance, travel time,
travel cost, age, and income across airport access modes. As shown in Table 4, the
means of all variables result significant differences among modes in both airports at
a level of a 0.1. Pearson Chi-squared test results in Table 4 indicate that there are
significant differences in trip purpose distribution by access modes at Gimpo Airport
(p-value 0.007); however, this is not the case at Daegu Airport (p-value 0.243).
That is, access mode choice of air travelers at Gimpo Airport is associated with trip

Figure 2. Distribution of respondents by age.


10 S. Choo et al.

Figure 3. Distribution of respondents by trip purpose.

purpose, whereas at Daegu Airport trip purpose is not a significant indicator of


access mode choice. For example, those who travel for leisure tend to take buses, and
for business prefer cars, in order to access the airport.
Downloaded by [UQ Library] at 02:55 24 May 2013

Table 5 presents the mean travel distance, cost, and time by airport access mode.
Passengers who access the airports by bus have the longest mean distance traveled,
whereas those who access the airports by taxi have the shortest. This is because taxi
fare increases with distance. Public transit riders are likely to have greater mean travel
time. Subway and bus access modes have the longest travel time for access to Gimpo
Airport and Dague Airport, respectively. These results are directly related to the
regional characteristics of these modes around the airport locations. Considering the
travel cost, the most expensive mode was taxi service and the most cost-efficient
mode was the subway for the two airports. Subway and taxi service exhibit distinctive
cost differences. However, there is a definite trade-off between time savings and high
fares for taxi service.

4. Access mode choice modeling


4.1. Model specification
As mentioned above, airport access modes were classified into four categories (car,
taxi, bus, and subway), and therefore, MNL models or NL models can be utilized for
developing airport access mode choice models. However, the data for ground access

Figure 4. Distribution of respondents by airport access mode.


Downloaded by [UQ Library] at 02:55 24 May 2013

Table 3. Cross-tabulation of airport access mode and trip purpose.

Transportation Planning and Technology


Gimpo Airport (N 1588) Daegu Airport (N696)

Trip purpose Car (%) Taxi (%) Bus (%) Subway (%) Total (%) Car (%) Taxi (%) Bus (%) Subway (%) Total (%)

Leisure 129 66 142 100 437 128 99 47 30 304


(29.5) (15.1) (32.5) (22.9) (100.0) (42.1) (32.6) (15.5) (9.9) (100.0)
Return 46 44 108 78 276 18 23 13 4 58
(16.7) (15.9) (39.1) (28.3) (100.0) (31.0) (39.7) (22.4) (6.9) (100.0)
Business 188 106 187 172 653 111 90 33 13 247
(28.8) (16.2) (28.6) (26.3) (100.0) (44.9) (36.4) (13.4) (5.3) (100.0)
Visit relatives/entertainment 49 28 46 31 154 34 20 9 4 67
(31.8) (18.2) (29.9) (20.1) (100.0) (50.7) (29.9) (13.4) (6.0) (100.0)
Other 23 8 20 17 68 12 3 3 2 20
(33.8) (11.8) (29.4) (25.0) (100.0) (60.0) (15.0) (15.0) (10.0) (100.0)
Total 435 252 503 398 1588 303 235 105 53 696
(27.4) (15.9) (31.7) (25.1) (100.0) (43.5) (15.1) (15.1) (7.6) (100.0)

11
12 S. Choo et al.

Table 4. Descriptive analyses of key variables by airport access mode.

Gimpo Airport Daegu Airport

One-way ANOVA test F value (P-value) F value (P-value)


- Travel distance 11.8 (0.000) 10.8 (0.000)
- Travel time 46.1 (0.000) 31.8 (0.000)
- Travel cost 207.2 (0.000) 35.3 (0.000)
- Age 10.2 (0.000) 6.9 (0.000)
- Income 5.5 (0.001) 3.9 (0.008)
Chi-squared test x2 value (P-value) x2 value (P-value)
- Trip purpose 27.3 (0.007) 15.0 (0.243)

modes were so limited that the MNL model could not be applied directly. In fact, the
data contains only travel time and travel cost information for the chosen mode. To
resolve the limitation of the data, logistic regression models are formulated after
regrouping the access mode categories into auto (car and taxi) and public transit (bus
Downloaded by [UQ Library] at 02:55 24 May 2013

and subway).
If Pi is the probability of choosing an auto, the general form of the logistic
regression model can be written as:
!
Pi
Y ¼ log ¼ b0 X ;
1  Pi

where a dependent variable Y is binary (if the respondent chooses an auto, Y 1,
otherwise Y 0), X is the vector of explanatory variables which influence Pi, and b is
the vector of parameters of the explanatory variables. If that formula is transformed
into a probability model, it is consistent with a binary logit model:

expðb0 X Þ
Pi ¼
1 þ expðb0 X Þ
Generally, the logistic regression model can be estimated by the maximum likelihood
method. In addition, the elasticity of Pi can be calculated by eb. This study employed
statistical modeling software, SPSS, to estimate logistic regression models.
The next section develops logistic regression models with three different mode
pairs for each airport and compares their results. The three pairs are auto/public

Table 5. Mean values of key variables by airport access mode.

Airport access mode

Car Taxi Bus Subway Total mean

Gimpo Airport Travel distance (km) 25.8 20.3 28.7 25.2 25.6
Travel time (min) 45 33 45 54 45
Travel cost (Won) 7734 13,764 4044 1116 5864
Daegu Airport Travel distance (km) 23.6 15.4 37.5 19.8 22.4
Travel time (min) 39 28 56 41 38
Travel cost (Won) 3523 8085 3262 2132 4918
Note: US$ 1 1100 (Korean) won.
Transportation Planning and Technology 13

Table 6. Model estimation of airport access mode choice.

Auto/ PT Auto/Bus Auto/Subway


Coefficient (t value) Coefficient (t value) Coefficient (t value)

Gimpo Airport
Constant 0.266 (0.83)
Travel time 0.588 (2.78) 0.0249 ( 7.52)
Return (purpose)* 0.693 (5.25) 0.107 (3.73) 0.480 (2.29)
Business (purpose)* 0.0177 ( 6.70) 0.633 (3.84) 0.414 (2.46)
Age 0.594 (4.06) 0.114 (2.53) 0.0447 (5.69)
Annual household income 0.399 (2.42)
Professional job*
Number of observations 1321 986 926
L(0) 915.6 683.4 641.9
L(c) 908.3 663.8 606.0
L(b) 874.5 644.7 550.0
x2 (P-value) 67.6 (0.000) 38.3 (0.000) 111.9 (0.0)
Downloaded by [UQ Library] at 02:55 24 May 2013

r20 1 L(b)/L(0) 0.04 0.06 0.14


r2c 1 L(b)/L(c) 0.04 0.03 0.09
Daegu Airport
Constant 1.036 (2.12)
Travel time 0.0376 ( 6.34) 2.341 (4.05)
Travel distance 0.634 (2.32) 0.0404 ( 6.44) 1.113 (1.64)
Business* 0.0323 (2.80) 0.00562 (1.69) 0.0349 (1.82)
Age 0.615 (2.26) 0.0308 (2.24) 1.271 (1.71
Female* 0.664 (1.84)
Administrative job*
Number of observations 489 458 417
L(0) 338.9 317.5 289.0
L(c) 251.7 199.2 110.4
L(b) 215.6 164.8 106.1
x2 (P-value) 72.3 (0.000) 68.9 (0.000) 8.6 (0.018)
r20 1 L(b)/L(0) 0.36 0.48 0.63
r2c 1 L(b)/L(c) 0.14 0.17 0.04
Note: Asterisk (*) indicates dummy variables. L(0) is an equally likelihood, L(c) is a likelihood of the
model having only constants, and L(b) is a likelihood value of an estimated model.

transit (PT), auto/bus (excluding cases with subway choice), and auto/subway
(excluding cases with bus choice). Additional logistic regression models for business
and nonbusiness travelers are also developed.

4.2. Model results


Logistic regression models for airport access mode choice were estimated with the
key variables discussed in descriptive analysis, as well as sociodemographic variables.
Table 6 shows the estimated model results for the three pairs of access mode choice.
All explanatory variables in the models were statistically significant at a 0.1 and
conceptually interpretable. In addition, as a goodness-of-fit test statistic, the x2 value
of each model showed that the estimated model significantly differs from the market
share model (the model containing only constant terms) at a BB 0.05. However, r2
14 S. Choo et al.

values appeared relatively low. Daegu Airport models had a better fit than Gimpo
Airport models.
Unlike the previous studies, travel cost was not statistically significant in some
models and had a counterintuitive sign in other models. This implies that the choice
of access mode is directly related to travel time rather than travel cost, because air
travelers tend to have higher income and consider less travel cost of airport access
modes. In the following section, we interpret the models by airport, and then discuss
the auto/public transit models for business and nonbusiness travelers.

4.3. Gimpo airport


Turning to the explanatory variables in the models of Gimpo Airport, travel time has
a negative coefficient for all models. It makes sense that the longer the travel time, the
lower will be the likelihood of selecting an auto for airport access. Interestingly, air
travelers with return or homecoming trip purposes tend to take public transit.
Perhaps, they are less likely to have time constraints compared to those with other
Downloaded by [UQ Library] at 02:55 24 May 2013

purposes. In the auto/bus model, the annual household income variable is


statistically significant and has a positive sign. As expected, high income travelers
are more likely to choose an auto as their airport access mode.
In the auto/subway model, three variables are also significant: business purpose,
age, and profession. It is interesting that air passengers for business travel prefer the
subway to an auto for airport access. Elderly travelers are more likely to choose an
auto, because they feel uncomfortable in accessing and taking a subway. People who
engage in professional and technical jobs tend to choose an auto. Professionals tend
to have higher incomes and their value of time is relatively high, therefore they use an
auto more frequently due to its speed and convenience.

4.4. Daegu airport


The auto/PT model of Daegu Airport has more significant variables than that of
Gimpo Airport. Similarly, travel time is significant with a negative coefficient in all
models except the auto/subway model. Additionally, travel distance has a negative
effect on the probability of choosing an auto in the auto/bus model. This suggests
that a bus is more competitive than an auto with respect to time and cost for longer
distance travel. It is reasonable that air travelers with business purposes tend to
choose an auto, since their companies would pay the travel cost. This result is
opposed to that of the auto/subway model of Gimpo Airport due to a different
choice set. Females are more likely to choose an auto. Similar to professionals,
administrative/managerial workers seem likely to choose an auto.

4.5. Business versus nonbusiness travel


To investigate access mode choice for business and nonbusiness travel, the data were
split into business and nonbusiness travel purposes, and then access mode choice
model for each travel purpose were developed. Only the auto/PT models were
estimated due to the small sample size. As shown in Table 7, all models have travel
time variables. Business travel models include annual household income and
administrative job variables, while nonbusiness travel models contain travel distance,
Transportation Planning and Technology 15

Table 7. Airport access model estimation of business and nonbusiness travel (auto/PT).

Business Nonbusiness
Coefficient (t value) Coefficient (t value)

Gimpo Airport
Constant 0.446 ( 1.52)
Travel time 0.116 (0.39) 0.0159 (4.57)
Travel distance 0.0221 (5.30) 0.000678 (4.02)
Age 0.197 (3.26) 0.0325 (4.47)
Annual household income
Number of observations 575 746
L(0) 398.6 517.1
L(c) 397.4 510.1
L(b) 376.7 480.5
x2 (P-value) 41.4 (0.000) 59.2 (0.000)
r20 1 L(b)/L(0) 0.05 0.07
r2c 1 L(b)/L(c) 0.05 0.06
Downloaded by [UQ Library] at 02:55 24 May 2013

Daegu Airport
Constant 0.298 (0.53)
Travel time 2.659 (5.96) 0.0404 (5.24)
Age 0.0337 (3.59) 0.0586 (4.00)
Female* 2.213 (2.13) 0.744 (2.36)
Administrative job*
Number of observations 190 299
L(0) 131.7 207.3
L(c) 86.1 163.9
L(b) 72.5 135.7
x2 (P-value) 27.2 (0.000) 56.4 (0.000)
r20 1 L(b)/L(0) 0.45 0.35
r2c 1 L(b)/L(c) 0.16 0.17
Note: Asterisk (*) indicates dummy variables. L(0) is an equally likelihood, L(c) is a likelihood of the
model having only constants, and L(b) is a likelihood value of an estimated model.

age, and female variables. This means that demographic characteristics of air
passengers significantly affect access mode choice for nonbusiness travel. Similarly,
elderly air passengers at Gimpo Airport and female passengers at Daegu Airport
show a tendency to choose an auto to access the airports for nonbusiness travel. As
expected, air passengers with high incomes at Gimpo Airport are likely to use an
auto as an access mode for business travel.

5. Conclusions
This study has explored air passengers’ airport access mode choice and develops
airport access mode choice models for two major domestic airports, Gimpo Airport
and Daegu Airport, using data from the 2004 Airline Passenger Pilot Survey in
Korea. The authors first conducted Chi-squared and ANOVA tests to identify
whether the key explanatory variables such as trip purpose, travel distance, travel
time, travel cost, age, and income are statistically different among airport access
modes. The test results showed that all variables were significantly different across
16 S. Choo et al.

access modes at both airports (except trip purpose at Daegu Airport). Logistic
regression models were proposed for three choice sets  auto and public transit, auto
and bus, and auto and subway  for both airports, considering the key explanatory
variables, as well as other sociodemographic variables. The model results indicate
that the choice of airport access mode is significantly affected by travel time, travel
distance, trip purpose, age, gender, occupation, and income. Our findings are
consistent with those of the previous studies. It was also found that there is a
difference in access mode choice between the two airports due to their geographical
differences. This study also estimated access mode choice models for business and
nonbusiness travel. The model results showed that demographic characteristics of air
passengers more significantly affect access mode choice for nonbusiness travel than
for business travel. Overall, key factors to affect airport access mode choice are
different by airport or region, even by country. The results of this study can provide
useful insight for airport planners and policy-makers designing and operating airport
facilities, as well as managing airport access traffic, particularly in a region with a
subway system.
Downloaded by [UQ Library] at 02:55 24 May 2013

One recent industry research study has predicted world air passenger traffic to
grow by 4.7% annually up to 2031 (Airbus 2013). As air travel grows, not only does
the number of air passengers increase, but the number of airport employees and
visitors would also increase (Budd, Ison, and Ryley 2011). The preference of access
modes and the resulting traffic impacts would be significantly different by airport
user types, since passenger demand levels are irregular, while airport employees
demand levels are fixed. In future work, the airport survey could not only focus on
passengers, but also be extended to airport employees and commuting mode choice
of employees. This would be useful in understanding and supporting airport ground
access strategies aimed at resolving traffic congestion and moving toward ‘green’
transportation policies.

References
Agresti, A. 1996. An Introduction to Categorical Data Analysis. New York: John Wiley & Sons.
Airbus. 2013. ‘‘Navigating the Future: Global Market Forecast 20122031.’’ Accessed March
26. http://www.airbus.com/company/market/forecast/?eIDdam_frontend_push&docID
27599.
Basar, G., and C. R. Bhat. 2004. ‘‘A Parameterized Consideration Set Model for Airport
Choice: An Application to the San Francisco Bay Area.’’ Transportation Research Part B 38
(10): 889904. doi:10.1016/j.trb.2004.01.001.
Ben-Akiva, M., and S. R. Lerman. 1985. Discrete Choice Analysis: Theory and Applications to
Travel Demand. Cambridge, MA: The MIT Press.
Bondzio, L. 1996. ‘‘Study of Airport Choice and Airport Access Mode Choice in Southern
Germany, Airport Planning Issues.’’ In Proceedings of Seminar K, PTRC European
Transport Forum, Vol. 409. Uxbridge: Brunel University.
Budd, T., S. Ison, and T. Ryley. 2011. ‘‘Airport Ground Access: Issues and Policies.’’ Presented
at the 2011 Annual Meeting of Transportation Research Board, Paper Number 11-0780,
January 24. Washington, DC: TRB.
Dowling Associates, Inc. 2002. San Jose International Airport Transit Connection Ridership.
Final Report. Oakland, CA: San Jose International Airport, Lea Elliott and Walker
Parking.
Ellis, R. H., J. C. Bennett, and P. R. Rassam. 1974. ‘‘Approaches for Improving Airport
Access.’’ Journal of Transportation Engineering 100: 661673. http://cedb.asce.org/cgi/
WWWdisplay.cgi?21990.
Transportation Planning and Technology 17

Gosling, G. D. 1984. An Airport Ground Access Mode Choice Model. Technical Document,
UCB-ITS-TD-84-6. Berkeley, CA: Institute of Transportation Studies, University of
California.
Gosling, G. 2006. ‘‘Predictive Reliability of Airport Ground Access Mode Choice Models.’’
Transportation Research Record 1951 (1): 6975. doi:10.3141/1951-09.
Gupta, S., P. Vovsha, and R. Donnelly. 2008. ‘‘Air Passenger Preferences for Choice of Airport
and Ground Access Mode in the New York City Metropolitan Region.’’ Transportation
Research Record: Journal of the Transportation Research Board 2042 (1): 311. doi:10.3141/
2042-01.
Halcrow Group Ltd. 2002. SERAS Surface Access Modelling, South East and East of
England Regional Air Services Study. London: Department of Transport, Local Govern-
ment and the Regions.
Harrington, I. E. 2003. The Logan Airport Passenger Ground Access Mode Choice Model,
Draft Memorandum. Boston, MA: Central Transportation Planning Staff.
Harrington, I. E., J. McClennen, E. Pereira, and C. Y. Wang. 1996. Summary of People Mover
Study Passenger Mode Choice Models, Draft Memorandum. Boston, MA: Central
Transportation Planning Staff.
Harvey, G. 1986. ‘‘Study of Airport Access Mode Choice.’’ Journal of Transportation
Engineering 112 (5): 525545. doi:10.1061/(ASCE)0733-947X(1986)112:5(525).
Downloaded by [UQ Library] at 02:55 24 May 2013

Harvey, G. 1988. ACCESS: Models of Airport Access and Airport Choice for the San
Francisco Bay Region-Version 1.2. Berkeley, CA: Metropolitan Transportation Commission.
Monteiro, A. B., and M. Hansen. 1996. ‘‘Improvements to Airport Ground Access and the
Behavior of a Multiple Airport System: BART Extension to San Francisco International
Airport.’’ Transportation Research Record 1562 (1): 3847. doi:10.3141/1562-05.
Portland Metro. 1998. PDX Ground Access Study Model Summary. Portland, OR: Travel
Forecasting Staff.
Pels, E., P. Nijkamp, and P. Rietveld. 1998. ‘‘Access to Airports: A Case Study for the San
Francisco Bay Area.’’ Presented at 38th Congress of the European Regional Science
Association, Vienna, August 26.
Pels, E., P. Nijkamp, and P. Rietveld. 2003. ‘‘Access to and Competition between Airports: A
Case Study for the San Francisco Bay area.’’ Transportation Research Part A 37: 7183.
http://www.sciencedirect.com/science/article/pii/S0965856402000071.
Psaraki, V., and C. Abacoumkin. 2002. ‘‘Access Mode Choice for Relocated Airports: The
New Athens International Airport.’’ Journal of Air Transport Management 8 (2): 8998.
doi:10.1016/S0969-6997(01)00033-3.
Skinner, R. E., Jr. 1976. ‘‘Airport Choice: An Empirical Study.’’ Transportation Engineering
Journal 102: 871883. http://cedb.asce.org/cgi/WWWdisplay.cgi?7045.
Sobieniak, J., R. Westin, T. Rosapep, and T. Shin. 1979. ‘‘Choice of Access Mode to Intercity
Terminal.’’ Transportation Research Record 728: 4753. http://trid.trb.org/view.aspx?id=
153291.
Tam, M. L., and W. H. K. Lam. 2005. ‘‘Analysis of Airport Access Mode Choice: A Case
Study in Hong Kong.’’ Journal of the Eastern Asia Society for Transportation Studies 6:
708723. http://www.easts.info/on-line/journal_06/708.pdf.
Tam, M. L., W. H. K. Lam, and H. P. Lo. 2010. ‘‘Incorporating Passenger Perceived Service
Quality in Airport Ground Access Mode Choice Model.’’ Transportmetrica 6 (1): 317.
doi:10.1080/18128600902929583.
Tambi, J. E., and J. Falcocchio. 1991. ‘‘Implications of Parking Policy for Airport Access
Mode Choice.’’ Presented at the 1991 Annual Meeting of the Transportation Research
Board, January 1317. Washington, DC: TRB.

You might also like