Content uploaded by Bartosz Bursa
Author content
All content in this area was uploaded by Bartosz Bursa on Oct 20, 2021
Content may be subject to copyright.
Bartosz BURSA
MODELING THE INTRA-DESTINATION
TRAVEL BEHAVIOR OF TOURISTS
DISSERTATION
eingereicht an der
LEOPOLD-FRANZENS-UNIVERSITÄT INNSBRUCK
FAKULTÄT FÜR TECHNISCHE WISSENSCHAFTEN
zur Erlangung des akademischen Grades
DOKTOR DER TECHNISCHEN WISSENSCHAFTEN
DOCTOR OF TECHNICAL SCIENCES
Band 3
der Schriftenreihe des Arbeitsbereichs Intelligente Verkehrssysteme
an der Universität Innsbruck
Hrsg.: Markus Mailer, Innsbruck, 2021
Arbeitsbereich Intelligente Verkehrssysteme, Institut für Infrastruktur,
Universität Innsbruck – Technikerstraße 13, A-6020 Innsbruck
Studia Verlag 2021
ii
All rights reserved, in particular the right of reproduction, distribution, storage in
electronic data systems and translation.
Copyright © 2021 Bartosz Bursa
STUDIA Verlag
Herzog-Siegmund-Ufer 15, 6020 Innsbruck
verlag@studia.at
Printed in Austria 2021
ISBN 978-3-99105-014-8
This publication was printed with the financial sup-
port of the Vice-Rectorate for Research of the Leo-
pold-Franzens-University Innsbruck.
iii
Betreuungskomission
Supervising Committee
Hauptbetreuer
Principal advisor
Univ.-Prof. Dipl.-Ing. Dr. techn. Markus Mailer, Univer-
sity of Innsbruck, Department of Infrastructure, Unit of
Intelligent Transport Systems
Zweitbetreuer
Co-advisor
Prof. Dr.-Ing. Kay W. Axhausen, ETH Zurich, Institute for
Transport Planning and Systems
Prüfungskommission
Examination Committee
Vorsitzender
Chair
Univ.-Prof. Mag. Dr. Hans-Peter Schröcker, University of
Innsbruck, Department of Basic Sciences in Engineering
Sciences, Unit of Geometry and Surveying
Erster Beurteiler
First examiner
Univ.-Prof. Dipl.-Ing. Dr. techn. Markus Mailer, University
of Innsbruck, Department of Infrastructure, Unit of Intel-
ligent Transport Systems
Zweiter Beurteiler
Seconds examiner
Univ.-Prof. Dr.-Ing. Martin Fellendorf, Graz University of
Technology, Institute of Highway Engineering and
Transport Planning
iv
v
VORWORT DES HERAUSGEBERS UND
BETREUERS
Nachdem Band 1 der Schriftenreihe des Arbeitsbereichs Intelligente Verkehrssys-
teme eine Dissertation aus dem Eisenbahnwesen präsentierte, folgt mit der Doktor-
arbeit von Herrn Dr. Bursa, MEng in Band 3 nun innerhalb kurzer Zeit eine zweite
Veröffentlichung aus der Verkehrsplanung. Mit einem Thema zu Mobilität und
Tourismus ist sie einem der Forschungsschwerpunkte des Arbeitsbereichs zuzuord-
nen, der auch im interdisziplinären Forschungszent-rum Freizeit und Tourismus
der Uni Innsbruck beteiligt ist. Bartosz Bursa war mit seiner Dissertation im dort
eingerichteten Doktorratskolleg integriert. Dem Leitsatz des Arbeitsbereichs „Mo-
bilität der Zukunft erforschen und gestalten!“ entsprechend, steht die Untersuchung
und Entwicklung von Mobilitätsangeboten im Fokus, die eine nachhaltigere Mobi-
lität im Tourismus ermöglichen.
In diesem Kontext behandelt die vorliegende Arbeit von Bartosz Bursa einen
wichtigen Aspekt, der in der bisherigen Forschung weniger Beachtung gefunden
hat. Im Gegensatz zur An- und Abreise ist die Mobilität von Touristen an den Ur-
laubsdestinationen weniger erforscht. Dabei ist diese nicht nur für die Wahl des ei-
genen PKW als Verkehrsmittel für die Anreise gera-de im alpinen Tourismus oft
entscheidend. Zunehmend ergeben sich durch die Überlagerung der Fahrten der
Gäste mit jenen der Einheimischen auch an Werktagen kritische Verkehrsbelastun-
gen, wie sie früher oft nur an den An- und Abreisetagen beobachtet wurden. Doch
was ist für die Verkehrsmittelwahl und Zielwahl in der vor-Ort-Mobilität der Gäste
entscheidend? Wie unterscheiden sich die Einflussfaktoren von jenen in der Alltags-
mobilität? In seiner Arbeit ist Bartosz Bursa diesen Fragen nachgegangen und hat
mit einem Discrete Choice Ansatz Modelle zur Verkehrsmittelwahl in der vor-Ort-
Mobilität von Touristen entwickelt. Die Arbeit wurde von der Österreichischen For-
schungsgesellschaft Straße-Schiene-Verkehr und dem österreichischen Bundesmi-
nisterium für Klimaschutz, Umwelt, Energie, Mobilität, Innovation und Technolo-
gie (BMK) mit dem FSV-Preis ausgezeichnet.
Univ.-Prof. Dipl.-Ing. Dr. Markus Mailer, Herausgeber und Betreuer
vi
FOREWORD BY THE EDITOR AND
SUPERVISOR
Following Volume 1 of the publication series of the Unit Intelligent Transport Sys-
tems, which presented a dissertation from the field of railway engineering, the doc-
toral thesis of Dr. Bursa, MEng, in Volume 3 is now the second publication from the
field of transport planning within a short period of time. With a topic on mobility
and tourism, it can be assigned to one of the main research areas of the unit, which
is also actively engaged in the interdisciplinary Research Center Tourism and Rec-
reation at the University of Innsbruck. With his dissertation, Bartosz Bursa was in-
volved in the doctoral program established there. According to the mission state-
ment of the unit "Researching and designing mobility of the future!", the focus lies
on the investigation and development of mobility services that will allow for a more
sustainable mobility in tourism.
In this context, the present work by Bartosz Bursa addresses an important as-
pect that has received less attention in research to date. In contrast to long-distance
travel to and from destinations, the mobility of tourists at vacation resorts has re-
mained under-represented in academic studies. Yet this is often crucial not only for
the choice of one's private car as mode of transport for travel to the destination,
especially in alpine tourism. Increasingly, we are also observing tourists' on-site trips
overlapping with those of residents, resulting in critical traffic congestion on week-
days, which was previously seen only on arrival and departure days. But what are
the key determinants of guests' mode and destination choices in their on-site mo-
bility? How do the influencing factors differ from those in everyday mobility? In his
work, Bartosz Bursa pursued these questions and, using Discrete Choice Analysis,
developed models for transport mode choice for intra-destination mobility of tour-
ists. The work was awarded the FSV Prize by the Austrian Research Association for
Roads, Railways and Transport and the Austrian Federal Ministry for Climate Ac-
tion, Environment, Energy, Mobility, Innovation and Technology (BMK).
Univ.-Prof. Dipl.-Ing. Dr. Markus Mailer, editor and supervisor
vii
ACKNOWLEDGEMENTS
It is Markus Mailer who came up with the idea of looking specifically at the intra-
destination movements of tourists. Seemingly uninteresting at the beginning, the
topic turned out to be deeply intriguing, offering a great potential of underexplored
research niches located at the crossroads of transportation, travel behavior, tourism,
leisure and vacation studies. He has always had a good intuition and helped me
avoid mistakes (though not all!) and blind alleys of the research. He also cared about
optimal working conditions that allowed me to make constant progress during the
course of the doctoral studies. I am very grateful for funding the participation in
numerous conferences and subsidizing the courses on discrete choice at the EPFL
in Lausanne and at the University of Leeds, which opened my eyes to where I am on
the personal development axis and what still is to be done.
I truly appreciate Kay Axhausen allowing me to work with him and his team
during my short stays at the IVT on ETH in Zürich. These visits were always ex-
tremely productive, resulted in huge progress in my work and the discussion we had
were always an intellectual pleasure. It never ceases to amaze me how broad his
knowledge is. He is the Renaissance Man in transportation. Chapeau bas! I am also
thankful to Felix Becker who always offered advice to me, whenever I had doubts
concerning modeling in R.
I am grateful to Martin Fellendorf for undertaking the role of the second evalu-
ator. I hope we’ll have more occasions to work together in the future.
Furthermore, I would like to thank my working colleagues in Innsbruck for
friendly atmosphere over the last four years. Special thanks goes to Stephan Tischler
for introducing me in the world of backcountry skiing and for these few tours that
we made together. Hoping for more!
Emma Komarek receives my thanks for her laborious work on questionnaire dig-
italization and data cleaning.
The biggest appreciation deserves my beloved girlfriend Patrycja who suffered
at least as much as I did and survived even more. She took the responsibility of the
graphic design of the PAPI questionnaires and was always at hand with her creativity
viii
and excellent graphic design skills. She was a steadfast supporter in the tough time
and a peaceful soul alleviating the unnecessary stress, rush and chaos I used to no-
toriously produce. Your dedication in the last months was worth one's weight in
gold.
Me working on this thesis was surely not the only thing bothering my mum dur-
ing the past four years. I would not have reached this level of education, had she not
strived for better chances for me as a single parent many years ago. Thank you for
that.
This endeavor is also for my grandma, who passed away last winter, for keeping
her fingers crossed for me all her life.
It has been intensive time – hectic, exhausting and formidable but also chal-
lenging, stimulating and rewarding. The pandemics that swept over the world in the
spring of 2020 right at the moment of finishing the thesis, the precariousness of
working in the academia and the entire scientific world evolving uncontrollably in
different directions did not make it easier either. Fortunately, living in Innsbruck
compensates for these hardships with interest. I have never really expected that I
could live in the Alps, even though, as someone who loves mountaineering, climb-
ing, cycling, etc., it has always been running through my mind. I can still remember
the moment I came across the job posting of the University in Innsbruck and the
discussions we had at home before taking the bold decision to jump ship and leave
London.
All in all, given that I embarked upon this scientific journey having only vague
idea of how the academia works, being unaware of research methods and even hav-
ing very mediocre knowledge about statistics, the sole fact that I reached the finish
line, can be called a lifetime accomplishment.
ix
KURZFASSUNG
Angesichts der ständig steigenden touristischen Nachfrage in den Alpenländern, des
damit verbundenen Verkehrsaufkommens und der daraus resultierenden negativen
Externalitäten sowie der sozialen und ökologischen Kosten ist es dringend notwen-
dig, eine Verkehrspolitik zu entwerfen, die in der Lage ist, den Tourismusverkehr
effizient zu steuern und in Anbetracht der begrenzten finanziellen, räumlichen und
ökologischen Ressourcen umsichtig in die Verkehrssysteme und die Infrastruktur
zu investieren. Während es ein deutliches Forschungsinteresse an Fernreisen und
Ankunfts-/Abreisemustern von Touristen gibt, sind Forschungsarbeiten zur touris-
tischen Mobilität während des Aufenthalts in der Urlaubsdestination leider so gut
wie nicht vorhanden. Dies erschwert es den politischen Entscheidungsträgern, fun-
dierte Entscheidungen zu treffen, die durch wissenschaftliche Erkenntnisse gestützt
sind. Die vorliegende Dissertation versucht, diese Forschungslücke zu schließen
und ein "analytisches" Licht auf das Reiseverhalten von Touristen am Reiseziel zu
werfen.
Zunächst wird dazu der Stand des Wissens in der Verkehrs- und Tourismuslite-
ratur recherchiert und zusammengefasst, um Faktoren zu identifizieren, die poten-
ziell Einfluss auf Mobilitätsentscheidungen von Touristen haben könnten. Der
Überblick über den Forschungsstand in den drei elementaren Wahlkomponenten
im Reiseverhalten, der Ziel-, Verkehrsmittel- und Routenwahl sowie der Theorie der
gemeinsamen Entscheidungen und den Auswirkungen des Wetters dient als Grund-
lage für die Gestaltung einer mehrteiligen, maßgeschneiderten Befragung zum Ver-
kehrsverhalten. Weiteres wird über die durchgeführte Feldforschung basierend auf
einer Umfrage, die in den Jahren 2018 und 2019 in drei Tourismusregionen im öster-
reichischen Bundesland Tirol durchgeführt wurde, berichtet.
Nach der deskriptiven Auswertung der Befragungsdaten, die auch die Unter-
schiede zwischen Sommer- und Wintersaison hervorhebt, werden in der Disserta-
tion ökonometrische Wahlmodelle für die Analyse von Entscheidungen über die
Verkehrsmittelwahl von Touristen eingesetzt. Anhand der Wege und Aktivitäten
x
der Befragten, ergänzt durch Daten aus externen Quellen, werden mittels Multino-
mial- und Nested-Logit-Spezifikationen die Einflussfaktoren ermittelt und deren Ef-
fektgröße in der erhobenen Stichprobe geschätzt.
Darauf aufbauend werden die vorgeschlagenen Wahlmodelle zur Berechnung
der Indikatorwerte für politische Maßnahmen verwendet. Dabei werden für alle Al-
ternativen Elastizitäten auf Änderungen in der Reisezeit und den Reisekosten ge-
schätzt. Darüber hinaus wird der Wert der Reisezeitersparnis (VTTS) von Touristen
für Reisen mit dem Auto und mit dem öffentlichen Verkehr berechnet. Sowohl die
Elastizitäten als auch die VTTS von Touristen werden mit den in österreichischen
und internationalen Studien berichteten Werten für die Mobilität der ansässigen
Bevölkerung verglichen.
Abschließend fasst die Dissertation die Ergebnisse zusammen und diskutiert
ihre Implikationen für Wissenschaft, Wirtschaft und Politik. Sie resümiert die Leis-
tungen der entwickelten Modelle und gibt klare Empfehlungen für ihre Anwendung
unter Berücksichtigung der Grenzen aller Forschungsphasen. Zudem werden Lü-
cken in der Wissenschaft identifiziert und weitere Aufgaben formuliert, die die For-
schung zur touristischen Mobilität über den Rahmen dieser Arbeit hinaus voran-
bringen können.
xi
ABSTRACT
In the face of a continuous increase in tourism demand in the Alpine countries, the
associated traffic volumes, and the resulting negative externalities as well as social
and environmental costs, there is an urgent need to design policies capable of man-
aging tourist traffic efficiently and to invest in transport systems and infrastructure
wisely, given the limited financial, spatial and environmental resources. Unfortu-
nately, while there is a considerable research interest in long-distance travel and
arrival/departure patterns of tourists, research on tourist mobility during the stay at
the destination is almost non-existent. This prevents policy-makers from making
informed decisions backed by scientific evidence. The dissertation attempts to fill
this research gap and shed an “analytical” light on travel patterns of tourists at the
destinations.
In the first instance, the transportation and tourism literature are researched
and synthesized in order to identify factors that might be potentially influential on
tourist decisions. The overview of the state of research on the three elementary
choice components in travel behavior, destination, mode and route choice, the the-
ory of joint decisions and the impact of weather serves as a basis for the design of a
multipart bespoke travel-activity survey. A field report from the survey conducted
in 2018 and 2019 in three tourist regions in the Austrian province of Tyrol is pro-
vided.
Following the descriptive analysis of the survey data highlighting differences
between the summer and winter seasons, the thesis employs econometric models of
choice for the analysis of tourist transport mode decisions. Based on the trips and
activities of the respondents, and supplemented by data from external sources, Mul-
tinomial and Nested Logit specifications are used to find the impactful factors and
measure their effect size within the collected sample.
Next, the proposed choice models are used to calculate values of the indicators
for policy measures. Elasticities with respect to changes in travel time and travel cost
are estimated for all alternatives. Furthermore, the Value of Travel Time Savings
(VTTS) of tourist visitors are calculated for travel by car and on transit. Both the
xii
elasticities and VTTS of tourists are compared to values of local residents reported
in Austrian and international studies.
Finally, the thesis recapitulates the findings and discusses their implications for
science, economy and policy. It summarizes the performance of the models devel-
oped and provides clear recommendations for their application, taking into account
the limitations at all stages of the research. In addition, new gaps in science are
identified and further tasks are formulated that could advance the research on tour-
ist mobility beyond the scope of this thesis.
xiii
CONTENTS
VORWORT DES HERAUSGEBERS UND BETREUERS ............................................... V
FOREWORD BY THE EDITOR AND SUPERVISOR ...................................................VI
ACKNOWLEDGEMENTS ............................................................................................ VII
KURZFASSUNG .............................................................................................................. IX
ABSTRACT ....................................................................................................................... XI
CONTENTS .................................................................................................................. XIII
LIST OF FIGURES ....................................................................................................... XVII
LIST OF TABLES .......................................................................................................... XIX
GLOSSARY .................................................................................................................... XXI
INTRODUCTION .................................................................................................... 23
Motivation ...................................................................................................... 23
Research objective and research questions ................................................ 24
Outline ............................................................................................................ 25
STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS ..................... 26
Definitions ..................................................................................................... 26
Data collection methods ............................................................................... 27
Activity-based travel modeling .................................................................... 29
Discrete Choice Analysis ............................................................................... 31
2.4.1 Random utility theory ........................................................................ 31
2.4.2 Modeling approaches ......................................................................... 32
2.4.3 Discrete choice and tourist travel in large-scale transportation
models .................................................................................................. 33
Destination choice ......................................................................................... 35
2.5.1 Destination choice in daily travel ..................................................... 35
xiv
2.5.2 Destination choice on vacation ......................................................... 36
Mode choice ................................................................................................... 38
2.6.1 Mode choice in daily travel ................................................................ 38
2.6.2 Mode choice on vacation .................................................................. 40
Route choice ................................................................................................... 41
Impact of weather .......................................................................................... 43
Joint decisions ................................................................................................ 45
Undirected travel ........................................................................................... 47
THE SURVEY WORK ............................................................................................. 49
Travel pattern of a tourist ............................................................................ 49
Survey design .................................................................................................. 51
3.2.1 Survey location .................................................................................... 51
3.2.2 Survey methods ................................................................................... 53
Survey instrument .......................................................................................... 54
3.3.1 Personal questions .............................................................................. 55
3.3.2 Sojourn-related questions .................................................................. 55
3.3.3 Activity diary ....................................................................................... 56
3.3.4 Joint travel ........................................................................................... 57
Survey participation and response burden.................................................. 58
Complementary datasets ............................................................................... 59
Imputation of missing values....................................................................... 60
Descriptive analysis ....................................................................................... 61
3.7.1 Exclusion of responses ....................................................................... 61
3.7.2 Characteristics of the respondents ................................................... 61
3.7.3 Mode choice of the respondents ....................................................... 71
3.7.4 Joint travel ........................................................................................... 76
3.7.5 Impact of weather ............................................................................... 78
THE MODELING WORK ....................................................................................... 81
Data processing .............................................................................................. 81
4.1.1 Characteristics of the decision-makers and the sojourn ................ 81
4.1.2 Attributes of alternatives ................................................................... 83
4.1.3 Individual choice set of available modes .......................................... 85
4.1.4 Exclusion of observations ................................................................. 88
Model specification and estimation ............................................................ 89
Multinomial Logit models............................................................................ 90
4.3.1 Models with travel time and travel cost ...........................................92
4.3.2 Models with access, egress and in-vehicle travel time for transit 94
4.3.3 Models with service quality variables for transit ............................. 95
xv
4.3.4 Models with interactions of travel time and cost with
sociodemographic- and sojourn-related variables ......................... 96
4.3.5 Models with hotel-related variables ................................................ 98
4.3.6 Models with company variables ....................................................... 99
4.3.7 Models with peak-time variables ..................................................... 101
4.3.8 Models with duration of the following activity ............................. 102
4.3.9 Models with trip purpose ................................................................. 103
4.3.10 Models with number of trips ........................................................... 104
4.3.11 Models with average duration of activities .................................... 105
4.3.12 Models with weather-related variables .......................................... 106
4.3.13 Models with tourists’ information levels ........................................ 108
4.3.14 Models with sport frequency ........................................................... 109
4.3.15 Full models ......................................................................................... 110
Nested Logit models ..................................................................................... 118
Cross-Nested Logit models .......................................................................... 121
Indicators for policy measures .................................................................... 124
4.6.1 Elasticity to changes in attributes of alternatives .......................... 124
4.6.2 Value of Travel Time Savings (VTTS) ............................................. 130
CONCLUSION ........................................................................................................ 133
Contribution and findings............................................................................ 133
Discussion and implications for science, economy and policy ................ 135
Limitations ..................................................................................................... 137
OUTLOOK AND FUTURE WORK ...................................................................... 139
Detailed modes of transport ....................................................................... 139
Model for joint trips, tours and household members’ daily schedules .. 139
Model for joint choice of activity, destination and transport mode ....... 140
Model for route choice ................................................................................ 140
Model for tourists’ budget and time consumption at the destination .... 141
Tourist self-selection .................................................................................... 141
Integration with transport modeling software .......................................... 141
SUMMARY ..................................................................................................................... 143
BIBLIOGRAPHY ............................................................................................................ 145
A APPENDIX: SURVEY QUESTIONNAIRES .......................................................... 170
A.1 PAPI questionnaire ...................................................................................... 170
A.2 CAPI questionnaire ...................................................................................... 179
CURRICULUM VITAE .................................................................................................. 196
xvii
LIST OF FIGURES
Figure 3.1 Example of a tourist daily schedule with long-distance trips to a
destination and back home .................................................................... 50
Figure 3.2 Location of the study area on the map of Austria and its neighboring
countries. Red-colored rectangle is presented in detail in Figure 3.3 .52
Figure 3.3 Location of the tourist regions Ötztal, Zillertal and Hohe Salve (red
dotted areas) in the province of Tyrol (color map) in Austria .............52
Figure 3.4 Correlations of the decision-makers’ characteristics ........................... 64
Figure 3.5 Reasons for choosing particular transport mode for travel to a tourist
destination (multiple choice possible) .................................................. 68
Figure 3.6 Locations of the accommodations reported in the survey .................. 69
Figure 3.7 Location of the activities depending on survey location ..................... 70
Figure 3.8 Chosen mode depending on the sociodemographic characteristics .. 72
Figure 3.9 Chosen mode depending on the characteristics of the sojourn and
travel to the destination........................................................................... 73
Figure 3.10 Number of trips made by a given mode depending on time of day ... 74
Figure 3.11 Chosen mode depending on current trip purpose ................................75
Figure 3.12 Length of trips [km] depending on chosen mode ................................ 76
Figure 3.13 Duration of trips [min] depending on chosen mode ........................... 76
Figure 3.14 Chosen mode depending on family composition during the trip ...... 77
Figure 3.15 Chosen mode depending on company size (only household members)
................................................................................................................... 77
Figure 3.16 Distance travelled depending on family composition ......................... 78
Figure 3.17 Mode choice of tourists depending on precipitation........................... 79
Figure 3.18 Trip distance for each mode in summer depending on precipitation 80
Figure 4.1 Correlations of the attributes of alternatives........................................ 85
Figure 4.2 Structure of the NL model for summer ................................................. 119
Figure 4.3 Structure of the NL model for winter ................................................... 120
Figure 4.4 Structure of the CNL model for winter ................................................ 122
xviii
Figure 6.1 Cross-nested model for joint choice of activity, destination and
transport mode ....................................................................................... 140
Figure A.1 PAPI questionnaire – page 1 ................................................................... 171
Figure A.2 PAPI questionnaire – page 2.................................................................. 172
Figure A.3 PAPI questionnaire – page 3 .................................................................. 173
Figure A.4 PAPI questionnaire – page 4 ................................................................. 174
Figure A.5 PAPI questionnaire – page 5 .................................................................. 175
Figure A.6 PAPI questionnaire – page 6 ................................................................. 176
Figure A.7 PAPI questionnaire – page 7.................................................................. 177
Figure A.8 PAPI questionnaire – page 8 ................................................................. 178
Figure A.9 CAPI questionnaire – page 1 .................................................................. 180
Figure A.10 CAPI questionnaire – page 2 .................................................................. 181
Figure A.11 CAPI questionnaire – page 2 (continuation) ....................................... 182
Figure A.12 CAPI questionnaire – page 2 (continuation) ....................................... 183
Figure A.13 CAPI questionnaire – page 3 ................................................................. 184
Figure A.14 CAPI questionnaire – page 3 (continuation) ....................................... 185
Figure A.15 CAPI questionnaire – page 3 (continuation) ....................................... 186
Figure A.16 CAPI questionnaire – page 4 and 5 ....................................................... 187
Figure A.17 CAPI questionnaire – page 6 ................................................................. 188
Figure A.18 CAPI questionnaire – page 7 ................................................................. 189
Figure A.19 CAPI questionnaire – page 8 ................................................................. 190
Figure A.20 CAPI questionnaire – page 22 and 23 .................................................... 191
Figure A.21 CAPI questionnaire – page 24 and 25 ................................................... 192
Figure A.22 CAPI questionnaire – page 26 ............................................................... 193
Figure A.23 CAPI questionnaire – page 27 ............................................................... 194
Figure A.24 CAPI questionnaire – page 41 (last page) ............................................. 195
xix
LIST OF TABLES
Table 3.1 Characteristics of the survey regionsa .................................................... 51
Table 3.2 Summary of the survey protocol depending on survey region, wave,
method and language .............................................................................. 58
Table 3.3 Sociodemographic description of the sample ....................................... 61
Table 3.4 Description of the sojourn and the long-distance trip to the
destination ............................................................................................... 65
Table 3.5 Characteristics of the accommodations reported in the survey ......... 69
Table 3.6 Mobility rates of the surveyed sample of tourists and the
corresponding rates in countries from which most guests in Tyrol
originate. Values per day per person (mobile and not mobile persons
together) .................................................................................................... 71
Table 3.7 Impact of (perceived) weather on mode choice ................................... 79
Table 4.1 Choice alternatives – original alternatives reported in the survey and
aggregated alternatives used in the models .......................................... 86
Table 4.2 MNL models with travel time and cost ................................................. 93
Table 4.3 MNL models with access, egress and in-vehicle time for transit ....... 94
Table 4.4 MNL models with number of transfers on transit trips ...................... 95
Table 4.5 MNL models with headways between transit vehicles ........................ 96
Table 4.6 MNL models with interactions of travel time and cost with
sociodemographic- and sojourn-related variables ............................... 97
Table 4.7 MNL models with hotel-related variables ............................................ 99
Table 4.8 MNL models with company variables .................................................. 100
Table 4.9 MNL models with peak-time variables ................................................. 101
Table 4.10 MNL models with duration of the following activity ......................... 102
Table 4.11 MNL models with trip purpose ............................................................. 104
Table 4.12 MNL models with number of trips ....................................................... 105
Table 4.13 MNL models with average duration of activities ................................ 106
xx
Table 4.14 MNL models with weather-related variables ...................................... 107
Table 4.15 MNL models with tourists’ information levels about trip to
destination and mobility on-site .......................................................... 108
Table 4.16 MNL models with sport frequency ........................................................ 110
Table 4.17 Full MNL models for summer and winter ............................................. 111
Table 4.18 Comparison of all MNL models for summer. ....................................... 113
Table 4.19 Comparison of all MNL models for winter. .......................................... 116
Table 4.20 NL model for summer ............................................................................. 119
Table 4.21 NL model for winter ................................................................................ 121
Table 4.22 CNL model for winter ............................................................................ 122
Table 4.23 Structural parameters of the CNL model ............................................. 123
Table 4.24 Point cost elasticities for summer and winter – direct and cross ...... 126
Table 4.25 Point time elasticities for summer and winter – direct and cross ..... 127
Table 4.26 International comparison of direct and cross elasticities with respect
to cost and time ...................................................................................... 128
Table 4.27 Arc cost elasticities – direct and cross .................................................. 129
Table 4.28 Arc time elasticities – direct and cross ................................................. 129
Table 4.29 Value of Travel Time Savings (VTTS) [EUR/h] ................................... 130
Table 4.30 International comparison of Value of Travel Time Savings (VTTS)
[EUR/h]a ................................................................................................... 131
xxi
GLOSSARY
AIC: Akaike information criterion
ASC: Alternative Specific Constant
BIC: Bayesian information criterion
CHF: Swiss Franc currency
CNL: Cross-Nested Logit
DCA: Discrete Choice Analysis
EUR: Euro currency
GDP: Gross domestic product
GPS: Global Positioning System
LL: Log-likelihood
LR: Likelihood-ratio
MMNL: Mixed Multinomial Logit
MNL: Multinomial Logit
NL: Nested Logit
RP: Revealed preference
SD: Standard deviation
SP: Stated preference
UNWTO: United Nations World Tourism Organization
VOT: Value of Time
VTTS: Value of Travel Time Savings
xxii
23
INTRODUCTION
MOTIVAT I O N
Tourism industry is an important source of income in the Alpine areas of Austria,
Switzerland, Italy or France. Tourism accounts for 17.5% of direct Gross Domestic
Product (GDP) in the Austrian province of Tyrol (MCI, 2014). While the average stay
duration of tourists in Tyrol decreased from 5.1 nights in 2000 to 4.0 nights in 2019,
the number of arrivals increased by almost 60% from around 8 million to more than
12 million over the last two decades, despite no expansion on the supply side as the
number of beds dropped by 7% in this period (Statistics Austria, 2020). This is an
evidence of an accelerating trend of short yet more frequent holidays (Alegre and
Pou, 2006; Gössling et al., 2018; Martínez-Garcia and Raya, 2008), which unavoida-
bly results in an increase in tourism-related travel (Schlich et al., 2004). Given the
fact that almost 75% of inbound holiday trips to Austria are made by private car
(Austrian National Tourist Office, 2014), the effects of this trend on traffic conges-
tion and parking space management at the destinations can be substantial (Culli-
nane and Cullinane, 1999; Dickinson and Robbins, 2007, 2008; Regnerus et al., 2007).
Particularly in mountainous regions, where alternative routes are limited, tourist
traffic coinciding with daily commute, leisure and freight traffic leads to disturb-
ances to local communities in high season (Langer, 1995; Ogrin, 2012; Pechlaner and
Hamman, 2006; Scuttari et al., 2016; Scuttari et al., 2019; Scuttari and Isetti, 2019;
Tischler and Mailer, 2014) and deteriorates residents’ perception and attitudes to-
wards tourism development (Hudson, 2005; Lindberg et al., 1999; Mason and
Cheyne, 2000; McGehee and Andereck, 2004; Pegg et al., 2012).
Besides the effect on congestion and performance of transportation networks,
an increase in number of car trips in tourist regions inevitably implies a negative
environmental impact, which is clearly reflected in increased CO2 emissions
(Dolnicar et al., 2010; Filimonau et al., 2014; Gühnemann et al., 2021; Mailer et al.,
2019; Martín-Cejas, 2015; Rendeiro Martin-Cejas and Ramirez Sanchez, 2010; Unger
et al., 2016), but also other negative externalities such as higher noise levels (Barber
et al., 2011; Díez-Gutiérrez and Babri, 2020; Monz et al., 2016; Pickering and Barros,
24 INTRODUCTION
2013; Zhong et al., 2011) and higher number of accidents (Ball and Machin, 2006;
Bellos et al., 2020; Castillo-Manzano et al., 2020; Wang et al., 2016).
While the problems are recognized and present also in other non-urban desti-
nations, they have attracted only limited attention of researchers so far – a point
raised by Gronau (2017b) or (Dickinson and Dickinson, 2006). Local authorities still
do not have any quantitative evidence at their disposal. In effect, the policy
measures are often shots in the dark, which, despite entailing considerable expenses
(e.g. free transit services for tourists), lack proper evaluation and appraisal. This
work aspires to make a step in filling this gap.
In terms of vacation travel, we know much about travel decisions of people from
census data and studies on travel behavior conducted in origin countries. Moreover,
governments and international organizations collect aggregate data on tourism
economy, global markets and produce statistics on travelers moving between and
within countries. It is however rational and legitimate to assume that the travel be-
havior of tourists at the destination is not only different from how they behave on
daily basis at home (Guiver et al., 2008; Prillwitz and Barr, 2011), but also from the
behavior of residents in the regions they visit (Kinsella and Caulfield, 2011; Lumsdon,
2006). It also may not be in line with the data available at the aggregate level. Yet,
current research in this field is limited and concentrates merely on international
tourism demand and long-distance trips (Christensen and Nielsen, 2018; Gerike and
Schulz, 2018; Janzen et al., 2018). It still remains mostly unexplored how tourists
travel on-site at the destinations, which is of greater importance for local authorities
and communities than for central or federal governments. Moreover, as noticed by
LaMondia and Bhat (2013) most of research studies on tourists’ travel behavior up
to now are descriptive and therefore incapable of predicting.
RESEARCH O B JEC T I V E AN D R E S E A RC H Q U E ST I O N S
The fundamental goal of the thesis is to develop a comprehensive scientific ap-
proach to the analysis of tourist travel behavior at the destination. It will inform
tourism practitioners, transport planners and policy-makers working in tourist re-
gions about data collection procedures, modeling methods and implications for pol-
icy-making. With a focus on the transport mode choice, the thesis will deliver meth-
ods scientific in nature but capable of solving practical problems, where other ap-
proaches, successfully used for modeling daily travel, fail.
Driven by the above objective and based on a detailed review of the existing
literature, following detailed research questions have been defined:
1. What factors determinate travel decisions of tourists staying in alpine re-
gions in terms of mode choice?
2. Is there a substantial impact of the accompanying party size and composi-
tion?
1.3 OUTLINE 25
3. Is there a substantial impact of weather conditions?
4. How do tourists valuate their travel time savings depending on transport
mode?
5. How might tourists respond to policy measures aiming to change the modal
split in tourist regions?
OUTLINE
The thesis is comprised of six chapters with chapter 1 providing an introduction to
the topic, describing the motivation and setting the objectives for the research.
Chapter 2 provides a broad background for next chapters. It starts from defining
the terminology. Next, it reviews the literature on data collection methods and
travel decisions, with a particular focus on tourists in vacation setting. Finally, the
state of research in Discrete Choice Analysis (DCA) is described followed by appli-
cations of DCA in tourism and large-scale transportation models.
Chapter 3 covers the survey work. First, the conceptual framework of tourist
travel at the destination is presented. Next, a detailed description of survey meth-
odology and design is given, followed by response behavior statistics. The last part
of this chapter is a broad descriptive analysis of the collected data.
Chapter 4 deals with the second core part of the thesis – the modeling work.
The data preparation process is precisely described. Afterwards, model specifica-
tions and estimation results of Multinomial Logit (MNL), Nested Logit (NL) and
Cross-Nested Logit (CNL) models along with their interpretation are provided.
Chapter 5 synthesizes the results and formulates the answers to research ques-
tions stated in chapter 1. The findings are critically discussed and potential implica-
tions for science, economy and policy are proposed. Limitations of the research are
clearly highlighted.
Chapter 6 provides an outlook on future research and suggests prospective
study topics that could either be an extension of the research described in this thesis
or could resolve some of its limitations.
26
STATE OF RESEARCH ON TRAVEL
BEHAVIOR OF TOURISTS
DEFINITI O N S
As noticed by Arce and Pisarski (2009), there are many future challenges in describ-
ing tourists’ mobility that are caused by i.e. data unavailability, different levels of
analysis or inconsistencies in definitions. In particular, they highlight the following:
1. “Distinguishing between international versus domestic tourists and their
travel behaviour;
2. Distinguishing between visitors versus others who are not residents (immi-
grants, border workers, refugees, transit passengers, etc.) and their travel be-
haviour and impact on urban areas;
3. Distinguishing between overnight visitors (tourists per UNWTO) versus
same-day visitors and their travel patterns and impacts;
4. Distinguishing between business versus leisure travel and their travel pat-
terns and impacts;
5. Defining and investigating travel involving visiting former home (family vis-
its) in both domestic and international settings;
6. Inbound versus outbound direction of trip making and the dynamics of such
travel.” (Arce and Pisarski, 2009)
Also this thesis needs to cope with the above mentioned problems. In particular,
regarding point 1 and 3.
Therefore, several assumptions were made in the thesis to avoid ambiguities.
This thesis operates with the definitions of tourism and tourist as proposed by
United Nations and World Tourism Organization (1994), so as to avoid confusion
with traveler, vacationer or holidaymaker (Terrier, 2009). All these terms are used
in the thesis interchangeably though all meaning a tourist. The main restriction this
definition of tourist imposes, is that a person should be out of home (place of resi-
2.2 DATA COLLECTION METHODS 27
dence) for at least one night. It can be either a domestic (Austrian) or a foreign tour-
ist. The person must not be specifically on vacation, business purposes or family
visits are also allowed. It cannot be however a seasonal worker. Of interest are all
trips and activities performed during the stay (leisure and non-leisure).
DATA COLL E C T IO N M E TH OD S
This section reviews the relevant literature and examines traditional as well as more
recent data collection methods, keeping in mind that the thesis concentrates on the
intra-destination mobility of tourists in rural and alpine regions, i.e. their activities
and trips within mountain valleys and resorts.
The technological progress in recent years has provided academics with new
opportunities for measuring mobility by utilizing passively collected big data. Apart
from transport researchers also tourism researchers applied tracking technologies
in a number of studies (Shoval et al., 2014; Shoval and Ahas, 2016). However, these
deal with research questions relevant for tourism marketing, tourism demand or
tourism geography and overlook the transportation-related aspects of tourist travel
like traffic generated at destinations or transport mode choice.
Mobile positioning data have been widely utilized by tourism researchers in the
last decade (Ahas et al., 2008; Zhao et al., 2018). Yet, they proved useful only in ap-
plications limited to long-distance travel demand and tourism statistics. In trans-
portation, decisions strongly depend on characteristics of decision-makers (Lu and
Pas, 1999). However, mobile positioning data, for technical and ethical reasons, is
lacking this information. Only pure location data with time stamps is available, al-
beit in mountain regions the density of GSM transceiver stations is insufficient for
high-resolution analysis at the destination level. In addition, in alpine regions, cross-
border trips are very common, resulting in frequent changes of network provider.
Thus, only parts of these trips will appear in the dataset obtained from a national
provider.
GPS tracking can deliver very fine-grained data on tourist mobility allowing
analyses of specific activities or monitoring visitors to facilities, parks and venues (Li
et al., 2019). If complemented with additional questionnaires, GPS tracking can serve
as a superior alternative to traditional travel surveys among tourists. Currently, mo-
bile phones appear to be used more often in research than independent GPS trackers
since smartphone apps allow for correcting and annotating trips by the user and
answering supplementary questions (Prelipcean et al., 2018). Although the first stud-
ies reported on failed attempts of GPS tracking with mobile phones (McKercher and
Lau, 2009), the success rate increased over the last few years. So far, the most com-
plete and successful approach that combines an annotated travel diary and GPS
tracking in a smartphone app for tourist tracking was developed by Hardy et al.
(2017), who distributed 240 smartphones with a preinstalled tracking app among
visitors to Tasmania. However, besides high costs of such studies, there are practical
28 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
and organizational burdens. Tourists cannot be contacted before arrival to arrange
handing over the GPS units. In the case of a smartphone app, a communication
channel is necessary to make tourists aware of the app. Furthermore, battery con-
sumption and data roaming in the case of foreign visitors have to be considered.
Nonetheless, it is a promising approach and deserves further testing in the field.
Another alternative data source are social media services. Recent studies ap-
proximated tourist mobility patterns from geo-located Twitter data (Chua et al.,
2016; Provenzano et al., 2018), Flickr photos (Önder et al., 2016; Yang et al., 2017) or
Foursquare check-ins (Vu et al., 2018). However, in less populated areas, relevant
Points-of-Interest are underrepresented and geo-tagged tweets and photos are
scarce, which makes these methods applicable rather to city tourism (Sobolevsky et
al., 2015) or estimates of inter-destination tourist flows (Barchiesi et al., 2015). More-
over, even though the data can deliver valuable information on tourist activity for
the destination and park managers (Orsi and Geneletti, 2013), they are not of much
use for transport planners since a full reconstruction of all trips made is impossible.
Nevertheless, despite the expansion of big data, traditional surveys appear to be
still in use when investigating tourist populations. Big data on their own are not
capable of substituting traditional methods as they do not provide sociodemo-
graphic information, cannot measure unobserved variables or deliver strong causal
evidence (Chen et al., 2016; Mokhtarian, 2018). Unfortunately, as opposed to well-
established surveys on daily travel behavior (Brög, 2009), there is no consensus on
the design and methodology of such surveys in the tourism context that could lead
to a replicable approach. Also, very few researchers provide details on the survey
design and report on the fieldwork when applying travel diaries (Newmark, 2014;
Thornton et al., 1997; Tschopp et al., 2010). Author’s own experiences confirm many
weaknesses of diary-based surveys of tourists that are also known from surveys of
daily mobility, i.e. high costs, low response rate and high dropout rate. Besides, due
to high spatiotemporal dynamics of tourists on site, the sampling frame is unknown
and it is difficult to approach a representative sample when surveying outdoors. Sur-
veying visitors at their accommodations allows for more control over sampling (e.g.
indirect sampling through hotels) but requires a close cooperation with the accom-
modation providers, which is usually impossible without the support of local Desti-
nation Marketing Organizations. Even so, self-administered questionnaires distrib-
uted through tourism establishments prove very ineffective. It is therefore postu-
lated that only fully-assisted interviews can guarantee good quality results. Moreo-
ver, although travel diary data is detailed enough to model destination and mode
choice, it is usually insufficient to investigate route choice. Many of the above was
already noticed by Thornton et al. back in 1997 and is still up-to-date:
“Time-space diaries offer advantages over the other techniques, particularly ques-
tionnaires generating lists of 'places visited'. Diaries offer a more comprehensive
picture of tourist activities, including 'informal' ones such as relaxing. Anderson
2.3 ACTIVITY-BASED TRAVEL MODELING 29
argues: "The main quality of space - time diaries is perhaps that they record be-
haviour patterns which are not normally directly observable because of their spa-
tial and temporal extent" (1971, page 359). However, diaries also present problems.
Because diaries have been used for a wide variety of purposes they do not comprise
a uniform field of study. Therefore, there is considerable variation in underlying
methodologies, and important methodological and technical issues have not yet
been settled satisfactorily. Although diaries may be rich in detail on the patterning
of activities in space and time, there are still limitations on the amount of data
that can be recorded. For example, the day is usually divided into recording blocks
(to assist later analysis) but there is no clear guidance as to the appropriate length
of these blocks. There is also an unresolved debate as to how to record the spatial
coordinates of activities: whether in spatial zones, by precise named locations, or
by grid references. There is, of course, a danger that the approach taken may im-
pose an extraneous structure on the day or week that does not exist in reality.
Furthermore, the considerable effort required on the part of respondents for the
accurate recording of activities usually leads to low response rates. Similarly, the
quality of data obtained varies according to the enthusiasm of individual respond-
ents. In extreme cases, it is impossible to guarantee that uninterested respondents
do not complete the diary in retrospect, thus making it a recall document. Bell
also argues that diarists must be of a sufficient educational level to understand
often complex instructions, let alone complete the diary (1987, page 82). Bias can
also be seen to derive from the potential for accidental or willful misrepresentation
of data within self-completed diaries. Oppenheim, for example, claimed respond-
ents' particular interest in filling the diary will cause them to modify the very way
that behaviour is recorded, either through 'dutiful action' (that is, activities un-
dertaken in order to have something to record) or recording only those activities
likely to give a favourable impression (1966, page 215). However, it should be added
that many of these problems are common to other techniques, which also suffer
from the further disadvantages of the spatial and temporal limitations of the data
they obtain. Time - space diaries have been used in a number of social science
disciplines and are relatively well developed in retailing studies [for example, see
Wrigley and Guy (1983) for a review of this genre] compared with their relative
neglect in tourism studies. There are, however, some exceptions, the most notable
of which are Murphy and Rosenblood (1974), Gaviria (1975), Cooper (1978), Pearce
(1981), Pearce (1988), Debbage (1991), and Dietvorst (1994).” (Thornton et al.,
1997)
ACTIVITY -B A S E D T R A V E L M O D E L I N G
Following the industrial revolution and the emergence of private-use vehicles, the
20th century witnessed a rapid development of transport infrastructure. However,
30 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
increasing construction costs, spatial limitations and, as a result, decreasing effec-
tiveness of new investments in the last decades, have forced planners to switch from
the supply-oriented approach, concentrated on extension of transport networks, to
efficiently managing the growing demand for travel so that it suits the existing in-
frastructure (Bates, 2008; Pinjari and Bhat, 2011). Since 1970s, the modern transport
planning focuses therefore no more on aggregate demand produced by undefined
people masses but rather on actions of single individuals or households and hence
is very behavior-oriented. Transport modeling techniques, which serve as a tool
providing information for transport policy and demand management strategies,
have undergone the same transformation, from the trip-based approach to the ac-
tivity-based one. McNally (2000) and Pinjari and Bhat (2011) explain how and in
what aspects these two methods differ from each other.
First, in the activity-based approach, the demand for travel derives from the in-
dividuals’ needs to pursue activities, which is based on theoretically sound assump-
tions (Jones, 1979b). Secondly, travel is partitioned into tours, not trips. Tours is a
chain of trips that starts and ends at the same location. It is a lifelike approach that
can capture the interdependencies between subsequent trips (in terms of time, lo-
cation or transport mode) and is more consistent with people’s real behavior.
Thirdly, activity-based models can replicate how the individual allocates his or her
time, which is a constrained good, to activities and travel. Finally, activity-based
models operate at the disaggregate level of single individuals and thus can realisti-
cally respond to sociodemographic or infrastructural changes at a very high level of
detail, unachievable in trip-based models using average characteristics of arbitrarily
created traffic analysis zones (TAZ). On the whole, the flexibility of activity-based
modeling has made it possible to account for various dimensions of travel, e.g. inter-
personal and intra-household interactions, social networks, time use or activity
scheduling, resulting in a very powerful modeling instrument (Bhat and Koppelman,
1999).
Concurrent with the conceptual evolvement in the understanding of travel and
with the switch from trip-oriented to activity-oriented approach, the mathematical
apparatus available to researchers has undergone significant developments. The
emergence of discrete choice modeling provided researchers with a versatile tool for
reproducing travel behavior of individuals and operationalizing the activity-based
approach at a high level of detail. These techniques are briefly described in section
2.4.
The most current state-of-the-art direction in research as well as in practice is
to combine all the features mentioned above into one integrated model system that
uses activities and daily schedules of individuals and households to derive tours and
that models decisions with discrete choice methods and incorporates it all in a single
(microsimulation) platform (Bowman and Ben-Akiva, 2001; Davidson et al., 2007;
Miller et al., 2005). There is however still much to be done, especially in the days
2.4 DISCRETE CHOICE ANALYSIS 31
when the ICT technology blurs the distinction between travel and activity, flexible
work arrangements allow for working from different places at different times and
decisions are made under high uncertainty (Miller, 2020; Rasouli and Timmermans,
2014).
DISCRETE CH O ICE AN A L Y S I S
A distinctive feature of many decisions undertaken in travel and transport is that
they are discrete. As opposed to continuous regression models answering the ques-
tion of “how much”, the discrete choice models provide an answer to the question
“which one” (Train, 2009). In other words, an individual chooses one specific alter-
native out of a finite set of alternatives. Examples could be the choice of groceries,
mobile phone service providers, choice of university or number of cars in the house-
hold. In the transport context, typical discrete choices are the ones about transport
mode (car/bus/train or private/public transport), trip destination (which shopping
mall or restaurant) or about one of the possible routes leading to the destination.
Discrete Choice Analysis falls under a broad category of supervised machine
learning techniques, which is currently a rapidly evolving area, constantly extended
with new methods (Alpaydin, 2020). However, discrete choice itself is an already
established modeling system and has been used in research since 1970s. The foun-
dations of discrete choice analysis have been laid in mathematical psychology (Luce
and Suppes, 1965) and consumer theory. Since then, the subject of discrete choice
methods has been developing dynamically and expanding from econometrics and
marketing to other areas like urban planning, transportation or policy making (Ben-
Akiva and Lerman, 1985). The importance of discrete choice analysis has been
acknowledged in 2000 by awarding Daniel McFadden The Sveriges Riksbank Prize in
Economic Sciences in Memory of Alfred Nobel
1
for his contributions to discrete choice
analysis. Currently, the most advanced research in discrete choice models is being
conducted in the field of transportation.
The paramount assumption underlying the decision rules in discrete choice
models of travel behavior is that the decision-makers always try to maximize the
utility of their choices. This theory is called Random Utility Maximization (RUM)
(Marschak, 1959; McFadden, 1977a). Outside the discrete choice framework also al-
ternative mechanisms leading to a decision exist. These are not based on an opti-
mality criterion but rather on heuristics and elimination rules.
2.4.1 Random utility theory
The random utility theory has been a fundamental concept underlying many econ-
ometrical models since 1960s as it possesses several convenient features. The RUM
1
Which is often wrongly considered a Nobel Prize – according to NobelPrize.org (2020).
32 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
theory assumes the decision-maker behaves consistent with the concept of rational
behavior, that is, he or she make consistent choices following logical rules that are
in their best interests (Ben-Akiva and Lerman, 1985). At the same time, whilst the
RUM theory assumes the decision-maker makes deterministic choice, the observer
(analyst) is not capable of measuring the utilities perfectly and the model contains
an error (is stochastic) (Anderson et al., 1992). So, the individual does choose the
alternative with the highest utility but the utility is not known with absolute cer-
tainty and hence is random. Accordingly, the utility is comprised of the determinis-
tic (observable) component and random (unobservable) component denoted as
follows:
The deterministic part of the utility of is represented by a single objective function
that reflects the attractiveness of the alternative. This function can take different
forms but the linear-in-parameters additive formulation is the most common:
where is an explanatory variable (either attribute of the alternative or charac-
teristic of the decision-maker ) and is a vector of coefficients.
To be able to solve the model and obtain the choice probabilities, an assumption
about the distribution of the unknown error term is necessary. Different distribu-
tions lead to different choice models (described in the next section). A good example
is the logit model for which the error term is assumed to be independently and iden-
tically Extreme Value distributed.
2.4.2 Modeling approaches
Up to now, many variations and flavors of the basic logit model have been devel-
oped, the most important of which are described below. Since the choice sets in
transportation applications usually consist of more than two alternatives, the ele-
mentary binary models are not being considered in the thesis.
The Multinomial Logit Model (MNL) assumes that the random components of
the utilities are independent, identically distributed and Gumbel distributed (Ben-
Akiva and Lerman, 1985). An important property of MNL is the Independence from
Irrelevant Alternatives (IIA), which means that the choice between any two alterna-
tives cannot be affected by the existence of other alternatives. It has its conse-
quences, which are well-documented (red/blue bus paradox) (McFadden, 1973).
The Nested Logit (NL) model allows to relax the IIA condition, which is the
major shortcoming of the MNL model, and to model alternatives sharing some at-
2.4 DISCRETE CHOICE ANALYSIS 33
tributes within the so-called nests. For each nest a separate MNL model can be es-
timated, the inclusive value (logsum of estimated utilities) of the alternatives (low-
est level) is then transferred to the utilities of the nests (upper level) and the model
is estimated sequentially. This idea was first presented in the 1970s in works by Ben-
Akiva (1973), McFadden (1977a) and McFadden (1981). Daly (1987) presented a con-
venient and efficient procedure of simultaneous estimation, which is implemented
in most of modern software packages for discrete choice analysis.
The Cross-Nested Logit (CNL) belonging to the family of Generalized Extreme
Value (GEV) models (McFadden, 1977a) was proposed by Vovsha (1997) as a gener-
alization of the NL model. This flexible approach allows for correlations between
alternatives within different nests as opposed to arbitrary set nest-wise similarities
forced by the NL structure.
By combining different choice models, McFadden and Train (2000) proposed a
new group of Mixed Models (MMNL) that allow to capture the taste heterogeneity
among decision-makers and correlations between alternatives. Mixed Models can
take very flexible functional forms and approximate any discrete choice model.
However, they require to use the simulation methods for the estimation.
As for now, one of the most advanced development in the field of discrete choice
are Hybrid Choice Models (HCM) (Ben-Akiva et al., 2002), which incorporate the
effects of latent variables (e.g., attitudes, perceptions, social influences) into the dis-
crete choice modeling framework. Hybrid choice models are currently being inten-
sively researched in the transportation field (Abou-Zeid and Ben-Akiva, 2014; Vij and
Walker, 2014, 2016).
Apart from the above mentioned, in the recent years a few alternative ap-
proaches have emerged that are based on different assumptions than the classic util-
ity maximization principle. An example is the idea of Random Regret Minimization
(RRM) rooted in the Regret Theory (Chorus, 2012).
2.4.3 Discrete choice and tourist travel in large-scale transportation models
It has been confirmed that the discrete choice models outstrip the gravity models in
terms of accuracy, flexibility and robustness (Mishra et al., 2013), and that the activ-
ity-based models outperform the traditional four-step models and deliver more pre-
cise outcomes for the daily planning practice (Timmermans and Arentze, 2011). Ac-
cordingly, several implementations of discrete choice framework within the activity-
based transportation models have been developed in the recent years (as mentioned
in section 2.3). Besides the prominent European examples like the Swedish SIMS
model (Algers et al., 1996), the ALBATROSS model for the Netherlands (Arentze et
al., 2000) and the Belgian FEATHERS model (Bellemans et al., 2010), many imple-
mentations originate from the United States and Canada, for instance Portland
(Bowman et al., 1999), Florida (Chow et al., 2005), Southern California (Bhat et al.,
2013), Maryland (Maryland State Highway Administration, 2013), or Toronto and
34 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
Montreal (Miller et al., 2005; Yasmin et al., 2017). An example from Asia is Singapore,
where discrete choice modeling has been applied for the workplace choice in the
large-scale agent-based transportation model (Vitins et al., 2016).
Slowly, discrete choice models are being developed than can predict destination
and mode choice for inter- and intrastate long-distance travel demand (both busi-
ness and tourist/leisure), albeit they often suffer from problems related to data avail-
ability (Miller, 2004) or very coarse resolution (Rich and Mabit, 2012). There are sev-
eral examples of successful implementations of such models in large scenarios, such
as the Ohio statewide model (Erhardt et al., 2007), the model for the Canadian prov-
ince of Ontario (Llorca et al., 2018) or the Long-Distance Passenger Travel Demand
Modeling Framework (Outwater et al., 2015b; Outwater et al., 2015a) developed for
the FHWA (US Federal Highway Administration). However, as far as the short-dis-
tance tourist travel at the destination is concerned, the author is not aware of any
such implementations in large-scale transportation models.
If local tourist trips are modeled at all, it is done at the aggregate level, which
has been criticized by Lew and McKercher (2006). The few examples of such models
known to the author are: transportation model for the Austrian province of Salzburg
(Hofer, 2015), transportation model of the BVG Berliner Verkehrsbetriebe (Berlin
Transport Company) for Berlin in Germany (Franke, 2017) or the transportation
model for the Swiss canton of Bern (Vrtic et al., 2010). A conceptual framework that
integrates long- and short-haul travel demand into a single microscopic MATSim
model (Horni et al., 2016) and allows for visitors’ trips at the destination was recently
proposed by Llorca et al. (2019) for the Munich metropolitan area in Germany. This
is a promising design, however, not operational yet.
An integration of discrete choice models with an agent-based modeling frame-
work appears currently to be the state-of-the-art approach amongst transport mod-
eling researchers. In the US, implementations were done e.g. in Sacramento, where
an activity-based disaggregate econometric model (DaySim) was developed to sim-
ulate residents’ activity and travel schedules (Bradley et al., 2010). In Europe, an ex-
ample is known from Copenhagen, where the same software platform was used to
develop the activity-based discrete choice model system called COMPAS (Vuk et al.,
2016). Both system built on the software platform DaySim and the work of Bowman
(1998). Also Hörl et al. (2019) extended the MATSim microsimulation framework
with a tour-based discrete mode choice model.
Also, tourism researchers highlight the potential of agent-based models in their
field, but also stress their complexity and challenging communication of simulation
results. A summary of latest applications of agent-based models within the tourism
field has been done by Nicholls et al. (2016). They argue that ABM are better capable
of accounting for the erratic character, instability and unstructured dynamics of
tourism than the existing simplistic linear- and equilibrium-oriented modeling
2.5 DESTINATION CHOICE 35
techniques. Applications to the alpine areas include for example models of winter
tourism demand in the changing climate (snow) conditions (Balbi et al., 2013).
DESTINA T I O N C HO IC E
2.5.1 Destination choice in daily travel
Destination choice belongs to the set of three fundamental choice dimensions: des-
tination, mode and route choice, which are probably one of the most researched
topics in travel behavior science.
Daily travel is largely shaped by trips to primary activities, that is, work and
education facilities, which are stable locations and do not change at short notice.
They are analyzed over long periods, together with other long-term accompanying
decisions like residential location choice, which has attracted considerable attention
in the recent years (Pagliara et al., 2010b; Waddell et al., 2007). This is often com-
bined with spatial aspects of mobility and interrelationships with land-use (Pozsgay
and Bhat, 2001), finally resulting in complex land-use-transport interaction (LUTI)
models (Acheampong and Silva, 2013; Katoshevski et al., 2013).
Destination choice by itself finds more direct application in modeling decisions
concerning secondary activities i.e. shopping (Kristoffersson et al., 2018; Miller and
O'Kelly, 1983) or recreational activities widely studied in environmental economics
and nonmarket valuation (Champ et al., 2017; Mäler and Vincent, 2005; Train, 1998).
Typically, these choices are driven by travel time and cost of travel to the destination
and a set of attributes reflecting the attractiveness of the destination measured by
e.g. retail area, entrance fee or number of opportunities.
As far as the non-work trips with many alternatives are concerned, the choice
set formation process gains in significance. Usually in discrete choice models, it is
assumed that the choice set is given and deterministically predictable (Ben-Akiva
and Boccara, 1995). This assumption is true as long as the number of alternatives
within the choice set is relatively small. However, unlike in mode choice, where the
choice set is finite, small (usually not more than a few transport modes are available)
and easy to determine (available modes for each individual are usually known), the
set of available destinations is usually large and too complex to implement in an
operational analytical model. Therefore, plausible choice sets of reasonable size are
created for destination choice models by sampling the elemental destinations based
on spatial similarities between them and aggregating them to traffic analysis zones
(Kim and Lee, 2017), possibly accounting also for the dominance and perception at-
tributes (Cascetta and Papola, 2009; Pagliara et al., 2010a). In other words, destina-
tion choice models can work at the level of regions, cities, traffic analysis zones or
categories of destinations (e.g. restaurant, beach, school) but not precise locations.
36 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
The further work follows the two-stage modeling approach by Manski (1977) – hav-
ing formed the choice set, a choice conditional on this generated choice set is made
(Zheng and Guo, 2008).
An alternative approach is proposed by Horowitz and Louviere (1995) or Swait
(2001), who argue that the choice set is rather another expression of preferences
than a separate pre-choice step.
2.5.2 Destination choice on vacation
The choice of the vacation destination has always been of interest to researchers
from tourism marketing and tourism management (Decrop, 2006; Sirakaya-Turk
and Woodside, 2005). Learning and dissecting these decisions is crucial for tourism-
dependent destinations to promote their assets, attract more guests, and as a result,
generate more revenue.
Yet, the focus of these studies is long-distance travel and tourism demand, not
necessarily tourist local mobility. According to Bieger and Laesser (2013), who ana-
lyzed the Swiss leisure market, the leisure mobility consists of three major compo-
nents:
− Inter-destination mobility – travelling from home to a destination
− Intra-destination mobility – meaning trips made in order to perform activ-
ities within the destination area
− Leisure mobility at home – induced by sport or cultural activities at home
While the inter-destination travel patterns have been widely investigated in the-
oretical works (Rugg, 1973; Sirakaya et al., 1996; Woodside and Lysonski, 1989) and
numerous case studies (Armstrong and Mok, 1995; Eymann and Ronning, 1997; La-
Mondia et al., 2010; van Nostrand et al., 2013), the research on the intra-destination
movements, i.e. travel within the destination, is relatively limited. As McKercher
and Zoltan (2014) argue, the reasons for that are threefold and pertain to the low
accuracy of the geolocation data, insufficient resolution of travel-activity data col-
lected from tourists, and lack of a theoretical framework. Only recently, there has
been more attention paid to local travel behavior thanks to the use of GPS (Global
Positioning System) traces from mobile devices (Shoval et al., 2014; Thimm and
Seepold, 2016) and GIS (Geographic Information Systems) techniques (Lau and
McKercher, 2006).
However, many of the existing studies are descriptive and focus on visualizing
geographical and temporal dimensions of tourist movements and drawing conclu-
sions on itinerary types and frequency of visits (McKercher et al., 2019; Wu and Car-
son, 2008). Lew and McKercher (2006), in the probably first theoretical work on
2.5 DESTINATION CHOICE 37
tourist intra-destination travel, provide an extensive breakdown of factors
2
impact-
ing intra-destination movements of tourists, ranging from tourist time budget to
personal characteristics to place knowledge.
Works utilizing mathematical models are much less prevalent so far. However,
the topic is slowly acquiring attention of researchers who start applying discrete
choice models to quantify travel behavior of tourists and embed them into models.
A relatively large study on tourist local movement (over 2000 face-to-face interviews
in 29 tourism destinations) was conducted in three regions in Japan by Wu et al.
(2011). Applying a latent class modeling framework, they revealed that, except travel
time and distance, attractiveness of a destination (measured by number of attrac-
tions and number of visitors) is the main factor influencing destination choice,
whereas sociodemographic variables (gender, age, marital status) are decisive for
the travel party choice. Researchers have also started exploiting GPS data for model
building. For instance, Hardy and Aryal (2020) employed neural networks to analyze
GPS tracks of tourist movements in a national park in Australia. Based on survey
data and GPS tracks, Li et al. (2019) built models of destination choice of tourist
visitors to Gulangyu region in China. They observed that tourists who purchased a
joint ticket that includes several attractions tend to travel to zones where these at-
tractions are located. Tourists also avoid areas where they have already been to and
areas with poor signage. As far as the intra-destination mobility within the Alpine
regions is concerned, Zoltan and McKercher (2014) analyzed visitors’ behavior in the
Swiss canton of Ticino based on destination card consumption. Their findings reveal
that tourist movement patterns are defined largely by the spatial dimension rather
than through activity-based segmentation. Nevertheless, none of the papers men-
tioned differentiates between movements that are part of tourist activities (e.g. mak-
ing a hiking trip) and movements to activities (e.g. driving to a zoo), which are of
greater importance for transport planning since they generate road traffic and
crowdedness in public transportation vehicles.
Compared to daily travel, the choice set of available destinations during a vaca-
tion stay can be a more complex issue. Unlike local residents, visitors do not have
equal knowledge about the area and may or may not be aware of some alternatives
(cf. the choice set formation process by Decrop (2010)) depending on whether they
have already been to the area or not or whether they have informed themselves in
advance about available options. Moreover, they usually have no fixed points regu-
lating their mobility patterns (except the accommodation), while residents are con-
strained to the location where they work, or school where they drop their kids etc.,
which imposes limitations on their choice set. Due to the short nature of the stay,
the visitors’ choice set can be dynamic and change quickly over time (Crompton,
1992), making it even more difficult to recognize it in the models. It can be also be
2
Many of these factors are used in the design of the survey instrument in section 3.3.
38 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
driven by habits, attachments or routine (Björk and Jansson, 2008), which contra-
dicts the assumption of tourist’s absolute rationality and optimization character of
the decision process. For instance, the returning tourists might not consider new
alternatives on-site (e.g., a restaurant) since they are used to the ones that they have
been visiting for the last few years.
Besides tourism, there have also been applications of choice models to leisure
trips of domestic populations. For instance, Simma et al. (2002) analyze the desti-
nation choice for leisure activities of Swiss residents within Switzerland. Bhat et al.
(2016) apply the Multiple Discrete-Continuous Probit Model (MDCP) to study the
leisure destination choice of domestic tourists in New Zealand. To the author’s
knowledge, by far the most comprehensive study dealing with leisure and tourism
destination choice specifically in alpine regions was conducted for Switzerland by
Tschopp et al. (2010). In their analyses of various tourism destinations, they utilize
the Multinomial Logit and Nested Logit models. Although the objectives and spatial
area of their work are similar to the ones defined in this thesis, they concentrate
merely on the arrival/departure trips to/from the final destinations for both leisure
and tourism purposes. Moreover, their destination choice model for holiday trips is
limited only to the winter season (skiing activities) and to the trips of Swiss citizens.
An example of a more locally and less state-wide focused study is the one by La-
Mondia and Bhat (2013), who applied the Multivariate Binary Probit Model to study
the visitors’ leisure travel behavior in Northwest Canada. Scarpa et al. (2008) ana-
lyzed the destination choice of members of the Italian Alpine Club (CAI) for one-day
outdoor trips in the Alps and discovered that, except travel cost, also difficulty of
hiking trails and number of mountain huts influence the decisions, while Scarpa and
Thiene (2004) concentrated only on climbers and mountaineers and found travel
cost, severity of the environment and number of alpine shelters to be influential
factors.
MODE C HO I C E
2.6.1 Mode choice in daily travel
Mode choice is the second of the three elementary choices in transportation. Liter-
ature on mode choice is vast and addresses the topic from a number of perspectives
such as modeling methods or applications in cost-benefit analyses.
Thanks to a small number of alternatives in the choice set and conceptual sim-
plicity, transport mode choice is a convenient field to develop and test new model
types ranging from the simple Multinomial Logit to Nested, Cross-Nested, Mixed
models with random coefficients and many others (cf. section 2.4). An interesting
recent development are Discrete-Continuous models (Bhat, 2005) making it possi-
ble not only to model what alternative is chosen, but also how much the alternative
is used given a certain money or time budget (cf. section 6.5). Finally, most current
2.6 MODE CHOICE 39
studies employ Latent Class and Hybrid models. These models often reveal that the
mode choice is strongly affected by personal attitudes (Paulssen et al., 2014). It is
therefore advisable to measure the psychological and sociological constructs in the
survey (e.g., using the Likert-scale questions) and include them in the model
through segmentations, latent classes and latent variables (Leong and Hensher,
2012).
In principle, the two basic factors always present in mode choice models are
travel time and travel cost. They are usually very effective in explaining people’s de-
cisions even if not accompanied by other variables (Frank et al., 2007; Limtanakool
et al., 2006). Other attributes are more mode-specific and pertain to level of service
of a given mode like waiting time, delay and frequency for transit, but may also in-
clude attributes representing perceived comfort or safety (Daziano and Rizzi, 2015).
The literature dedicated to mode choice is split into two branches – one oper-
ating with revealed preference (RP) data and second one using stated preference
(SP) data (Wardman, 1988). RP data provide information on what consumers actu-
ally do, which in transportation means that researchers observe factual choices of
transport system users and collect data on their real market behavior. These data
are considered very reliable in depicting current market equilibrium and personal
constraints of decision-makers but are limited only to the existing alternatives and
are often expensive to collect (Louviere et al., 2000). SP data on the other hand pro-
vide information on what consumers say they will do in hypothetical choice situa-
tions. Unlike RP data, SP data can inform about consumer preferences for new ser-
vices or products with new features, however, at the cost of reliability and validity
of responses. In recent years, also joint models using both RP and SP data have
emerged, which attempt to combine the advantages of both data types (Cherchi and
Ortúzar, 2002; Frejinger et al., 2006; Rashedi et al., 2017).
Substantial part of the research is policy-driven and delivers information on
choice elasticities or Value of Travel Time Savings (VTTS) for various transport
modes, which facilitates project appraisal and evaluation of policy and infrastructure
measures (Graham and Glaister, 2004). This is where the mode choice models uti-
lizing RP data are most useful since they reflect people’s real choices in contrast to
imaginable choices in the SP data, which still need calibration with RP data if are to
be used for forecasting (Hensher and Li, 2010).
However, stated preference data prove more applicable to the experimental re-
search purposes. The SP-based studies explore future mobility forms (Haboucha et
al., 2017; Krueger et al., 2016; Peeta et al., 2008) or estimate the demand for yet non-
existent or emerging modes (e.g. car-sharing, car-pooling, mobility-as-a-service)
(Antoniou et al., 2019; Becker and Axhausen, 2017; Ciari and Axhausen, 2012; Ho et
al., 2018; Wicki et al., 2019; Zhou and Kockelman, 2011). This is possible thanks to
40 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
advance in methodologies, allowing efficient design of choice experiments, survey-
ing larger populations and simulations for scenario predictions (Rose and Bliemer,
2009).
Mode choice models are also widely used for evaluation of transit pricing strat-
egies (Sharaby and Shiftan, 2012) or introducing tolls and congestion pricing (Basso
and Jara-Díaz, 2012; Washbrook et al., 2006).
2.6.2 Mode choice on vacation
Although the transport mode choice is relatively well represented in tourism litera-
ture, studies using discrete choice methods are very scarce. Much research with dis-
crete choice has been done in the fields of tourism demand (Morley, 2012), signifi-
cantly less in long-haul tourism destination choice (LaMondia et al., 2010) and very
little in mode choice modeling (Thrane, 2015).
The transport mode choice is strongly dependent on the destination choice and
hence they should be considered and modeled jointly, which has been repeatedly
demanded in the literature (LaMondia et al., 2010; Masiero and Zoltan, 2013). What
is more, the decision about the transport mode for vacation is not only driven by
factors related to journey to the destination itself, but also factors concerning the
on-site mobility. Visitors decide to travel to alpine regions by car for fear of insuffi-
cient mobility services at their destination and inflexibility of public transportation
(Bursa and Mailer, 2018). In such a case, private car provides a high degree of inde-
pendence and usually ensures the most effective utilization of time. Additionally,
not every single tourist spot in rural regions is accessible by public transport, which
discourages the exploration-focused tourists from relying only on public transport
services on-site (Le-Klähn and Hall, 2013). Luggage transport is another factor de-
terring tourists from choosing a transport mode alternative to car (Böhler et al.,
2006).
So, the decision about the transport mode choice for local trips within the va-
cation region depends strongly on the initial decision about the transport mode for
long-distance trip to the region. However, there are also external conditions, e.g.,
the influence of weather (Becken and Wilson, 2013; Järv et al., 2007).
An extensive review of literature examining factors determining the mode
choice in general is provided by De Witte et al. (2013), while van Middelkoop et al.
(2016) and Thrane (2015) focus on mode choice for long-haul tourist trips. As far as
tourists are concerned, a broad description of factors affecting their mode choice at
the destinations is included in Le-Klähn and Hall (2013). They found that lack of
information and personal preferences are the most common explanations for not
using transit services in rural tourism sites. In urban areas on the other hand, tour-
ists value the ease of use, efficiency and personal safety when choosing public
transport and parking facilities when driving private car, as Thompson and Schofield
2.7 ROUTE CHOICE 41
(2007) point out. Dickinson and Robbins (2008) also narrowed their research to ru-
ral destinations. Apart from identifying general convenience and need to carry
equipment as main reasons for choosing private car, they also highlight a strong car
attachment of some visitors who do not even consider alternatives no matter their
availability, price or other attributes. Gutiérrez and Miravet (2016) analyzed the de-
terminants of public transport use among tourists in a coastal region of Spain. How-
ever, their research is based only on dichotomous statements of visitors whether
they used public transportation during their stay and no data on individual trips
were collected. Moreover, their models ignore the attributes of the alternatives
available at the destination. Gross and Grimm, in their review paper (2018), synthe-
sized outcomes of many existing studies and found that above all the sociodemo-
graphic factors, transport mode chosen for trip to the destination, travel duration
and expenses as well as type of vacation (organized or individual travel) play a role
in transport mode choice at the destination.
Within the alpine setting, specific factors affecting the transport mode choice
for travel to the destination and the mobility at the destination have been investi-
gated by Seltenhammer et al. (2018) and Bieger and Laesser (2013), who revealed that
the family/group size and transport of sport luggage (e.g., skiing equipment, moun-
tain bike) is dominant in the decision process, particularly in the winter season.
Masiero and Zoltan (2013) applied a Probit model for the mode choice of tourists in
the Swiss canton of Ticino and observed, among other things, that domestic tourists
and returning visitors (i.e. tourists who have been to the region before) are more
likely to use public transportation, whereas older tourists and male tourists are more
inclined to use private cars. The work by Pettebone et al. (2011) provides insights
into mode choice at the destination from an American perspective. They found that
visitors to the Rocky Mountain National Park are willing to switch from private car
to shuttle bus if it enhances their chances of being in the park with fewer other
people.
A potential effect of length of stay and (associated with it) satisfaction and well-
being on mode choice is discussed in section 4.3.4 and footnote 10.
Nevertheless, none of the existing studies analyzed the importance of travel
time and travel cost for the transport mode choice of tourists traveling within the
destination in a quantitative way, which is a distinct gap in the research, making it
impossible to apply a monetary measure to improvements or deteriorations in at-
tributes of the available modes (e.g. a higher transit frequency or a longer travel
time).
ROUTE CH O I C E
Route choice is the third component of a minimum set of decisions that have to be
made when planning a trip. It is built of two major elements: the generation of a
choice set of alternative routes and the choice of a route from this choice set. Unlike
42 STATE OF RESEARCH ON TRAVEL BEHAVIOR OF TOURISTS
the mode choice, where the set of alternatives is small and easily identifiable, and
unlike the destination choice, where the set of alternatives is finite (though often
large) and possible to enumerate, the set of alternatives in route choice can be very
large and difficult to identify since the alternative routes share common links and
overlap to some extent (Bovy, 2009). Information about the network as well as the
manner of acquiring this information decides about the size of the choice set of al-
ternatives that the decision-maker is aware of. Out of those, the decision maker
might take into account only the selected ones, depending on specific preferences
and trip constraints, which constitutes the consideration set. Correct replication of
this process is a difficult task. Therefore much emphasis has been put in the last
decades on developing realistic choice set generation methods (Prato, 2009).
Nevertheless, almost all studies on the route choice behavior concentrate on
dense urban networks. This is understandable because these are the most challeng-
ing environments – urban networks are large, multimodal and the route choice plays
a significant role in traffic management and the resulting level of service of network
elements. The research on route choice in non-urban areas is very scarce. Tourism
researchers address the topic from the perspective of destination management and
roadside tourism facilities (Denstadli and Jacobsen, 2011), which is unusable for
transport modeling purposes. However, they provide some interesting observations
about how tourists differ in their route choice behavior from local residents, which
should be considered when developing models of tourist route choice.
Lew and McKercher (2006) have raised the issue of tourists not possessing full
knowledge about the transport system in the region they visit. They also highlight
the different character of transport networks in mountainous regions from the ones
in flat or urban areas, which makes the whole decision process about routing unlike
to what is common in urban areas (shortest route, fastest route):
“A destination’s topography will also influence the siting of facilities and the form
of the transport network, which in turn, will affect tourist flows. Movements in
mountainous destinations intersected by challenging passes will be different than
in flat destinations. Linear, point-to-point touring on clearly defined routes is
more likely to occur in mountainous or island areas, while the potential exists for
more dispersed and alternative routing patterns in destinations located in flat-
lands.” (Lew and McKercher, 2006)
They also mention the factor of picturesqueness of routes that often prevails over
travel time or distance when choosing a route to the destination or moving around
within the destination. This is confirmed by Jacobsen (1996) who discovered that
the views and landscape experience are cherished by motor tourists surveyed in Nor-
way. The component of visual attractiveness of a route plays a particularly important
role on optional (i.e. non-work) trips, which was already confirmed by Ben-Akiva et
al. (1984), who found that the disutility of travel time on non-scenic roads is about
five times the disutility of travel time on scenic roads. Problematic is however how
2.8 IMPACT OF WEATHER 43
to define picturesqueness and how to quantify the scenic attributes of a route. Ali-
vand et al. (2015) developed a very promising approach capable of computing scen-
ery-related attributes ranging from road curviness to the viewshed from the road
elements using data from different sources and providers e.g. volunteered geo-
graphic information (VGI), digital terrain model (DTM), TomTom, Panoramio geo-
tagged photos, Google Earth, census data etc. They found that an increased presence
of water bodies, mountains, forests and parks along a route positively contributes to
the probability of choosing it as a scenic route, whilst urban areas along the route
decrease this probability.
The common use of the built-in and external GPS navigation devices among
tourists should not be neglected. In the context of car use, it is supposed to result in
tourists sometimes having even better knowledge about traffic conditions than local
residents, who rely rather on their habits, common sense and heuristics. This, how-
ever, does not (yet) apply to the knowledge about parking facilities at the tourist
attractions.
Yet, routing decisions are preliminary not going to be considered in the thesis
as the alpine network systems provide limited routing alternatives and this topic is
currently of secondary importance. It is however scheduled as a future research
idea, especially the valuation of the visual component of the route choice is inter-
esting (see Section 6.4).
IMPACT O F W E A THE R
Typical activities performed by tourists in mountain regions, e.g. hiking, climbing,
cycling or skiing, are obviously weather-dependent. One can argue that if the par-
ticipation in activities is weather-dependent then the choice of transport mode used
to reach the locations where these activities are practiced may be affected by
weather too. It is therefore interesting to examine how and to what extent tourists
at the destination locations adapt their travel-activity patterns to unfavorable
weather conditions and whether they react in the same way as they do when they
are at home. This section provides an overview of what is already known in terms of
weather and climate effects on various facets of transportation and tourism.
In response to an increase in unexpected and severe weather events in the re-
cent decades, scientists started exploring their influence on transport more inten-
sively. Except a great deal of research on extreme weather, landslides, floods, unex-
pected snow and heat, all posing a danger to transport networks and causing dis-
ruptions in transit systems, there has also been some interest in weather influence
at the level of individuals and their behavioral reaction to, not necessarily extreme
weather, but above all to normal weather variability on a daily basis (a broad review
of weather effects on all facets of transport can be found in Liu et al. (2017) or Böcker
et al. (2013)).