The Effects of Low-Cost-Carriers on Regional Dispersal of Domestic Visitors in Australia

Examinations of the effects on visitors’ dispersal sourced from intra- modal and inter-modal differentials

Tay T.R. Koo

A Doctoral Thesis Submitted in Fulfilment of the Requirements for the Award of Doctor of Philosophy of The University of New South Wales

Revised July 2009

Supervisor: Dr. Richard C.L. Wu Co-supervisor: Professor Larry Dwyer Department of Aviation

ABSTRACT

This thesis was conceived in the context of post-2000 proliferation of Australian low-cost carriers and regional dispersal policy of the Australian government. The broad aim of this thesis is to examine the effects of low-cost carriers on regional dispersal of domestic visitors. Based on existing theoretical frameworks of tourists’ spatial behaviour and multi-destination travel itinerary, two theoretical constructs - intra-modal and inter-modal effects – were developed to conceptualise the regional dispersal effects of low-cost carriers. The former refers to differences between low-cost carriers and other models of airline business, and the latter refers to differences between low-cost carriers and other modes of transport. Logit models and national-level revealed preference data were used to examine the intra-modal effects, while stated choice method was used to examine the inter-modal effects on two representative regional tourism destinations - Ballina-Byron and Cairns - in Australia.

This thesis provides evidence that suggest low-cost carrier air arrivals tend to disperse for reasons that are different from network carrier air arrivals, supporting the significance of intra-modal effects on regional dispersal. It is claimed that the intra-modal effect is one reason why some destinations observe high growth in airport activity as a result of low-cost carrier entry, but the levels of tourism activities do not match that extrapolated from the level of growth in the incoming air traffic. Two case studies have shown that (1) ground transport policy can completely offset the negative effects on tourists’ dispersal propensity stemming from pre-determined trip characteristics, although the effectiveness of such policy variables varies significantly across destinations; and (2) significant discounts in airfares are sufficient to trigger a modal switch, even in situations when a car is the most suitable mode for the trip, suggesting a real possibility of a bypass of ground-mode-reliant regions. The findings should be of interest to regional destination managers with low-cost carrier services as much as for managers in peripheral destinations without low-cost carrier services.

i LIST OF PUBLICATIONS FROM THE THESIS

Peer-reviewed journal

Koo, T.R, Wu, C. L., and Dwyer, L. M., (2009) Transport and Regional Dispersal of Tourists: Is modal substitution a source of conflict between low-fare air services and regional dispersal? Journal of Travel Research (in press, accepted 20th January 2009).

Koo, T.R, Wu, R. and Dwyer, L. (2009) “Ground Travel Mode Choices Of Air Arrivals At Regional Destinations: The Significance Of Tourism Attributes And Destination Contexts” Research in Transportation Economics: a special issue on tourism (in press, accepted 1st September 2009)

Full conference papers:

Koo, T.R., Wu, C.L., Dwyer, L.M. (2007) Low Cost Carriers, Mode Choice and Regional Tourism Destinations in Australia, Air Transport Research Society (ATRS) conference, Berkeley, California

ii Working papers published in conference proceedings:

Koo, T.R., Wu, C.L., Dwyer, L. (2009) “The effects of affordable air transport on regional dispersal propensity of tourists: a logit analysis of the National Visitor Survey data” CAUTHE February 2009, Fremantle, working paper presentation

Koo, T.R. (2008) "Affordable air travel and regional dispersal in Australia" conference proceedings CAUTHE February 2008, Gold Coast, working paper presentation

Koo, T.R. (2007) “The impact of Low-cost-airlines on regional tourism destinations: issues and challenges” conference proceedings CAUTHE February 2007, Sydney, working paper presentation

iii ACKNOWLEDGEMENT

I am overwhelmed with gratitude for the help and support I received from my principal supervisor, Richard Wu. This dissertation would not have been possible without Richard’s guidance. I am sincerely thankful to my co-supervisor, Larry Dwyer, for his consistent advice and support, as well as his sense of humour and perspective in the midst of chaos. I would like to thank my father, Matt C.D., who has been my informal third supervisor, and my mother, Vivian G.S., and my sister, Su-jie. I would like to acknowledge the Cooperative Research Centre for Sustainable Tourism, established by the Commonwealth Government of Australia, and the Department of Aviation in the University of New South Wales, for financial support and professional development opportunities. I would like to extend my thanks to the staff of Australian Regional Tourism Research Centre, Ballina airport, Cairns airport and the Research and Strategy team in Tourism Australia, for their help with survey design and data collection.

TABLE OF CONTENTS

1. INTRODUCTION...... 1-1 1.1. LOW COST AIR TRANSPORT AND DISPERSAL ...... 1-1 1.1.1. Research significance ...... 1-1 1.1.2. Low Cost carriers' effect on dispersal: an issue of spatial scale...... 1-3 1.1.2. The link between Low Cost carriers and spatial behaviour of tourists …………...……………………………………………………………….. 1-4 1.2. RESEARCH AIMS...... 1-5 1.2.1. Statement of the general aim...... 1-5 1.2.2. Statement of the specific aims ...... 1-6 1.3. NOTES ON METHODS ...... 1-13 1.3.1. Discrete choice models...... 1-13 1.3.2. Stated choice data...... 1-19 1.4. CONTRIBUTION TO KNOWLEDGE ...... 1-21 1.4.1. Contributions and limitations...... 1-22 1.4.2. Key stakeholders...... 1-23 1.5. STRUCTURE OF THE THESIS...... 1-25

2. DISPERSAL AND LOW COST CARRIERS ...... 2-1 2.1 INTRODUCTION ...... 2-1 2.2 DISPERSAL...... 2-2 2.2.1 Definition of ‘regions’, domestic dispersal and regional dispersal.....2-2 2.3 THE CHARACTERISTICS OF LCCS ...... 2-6 2.3.1 The LCC model...... 2-6 2.3.2 Point-to-point network (P2P)...... 2-7 2.3.3 Use of secondary and regional airports...... 2-9 2.3.4 Short-haul and Low-cost customer service...... 2-11 2.3.5 Ticket distribution, fare structure and passenger-handling...... 2-12 2.4 BACKGROUND: PRECURSOR TO LCC GROWTHS IN AUSTRALIA...... 2-14 2.4.1 Deregulation of the airline industry in Australia ...... 2-16 2.4.2 Privatisation of the domestic airports...... 2-17 2.4.3 Foreign ownership cap ...... 2-18 2.5 AUSTRALIAN LCCS AND THEIR IMPACT ON DOMESTIC DISPERSAL...... 2-19 2.5.1 First wave of LCCs in Australia (1990 – 1993)...... 2-19 2.5.2 Duopoly period (1994 – 1999)...... 2-19 2.5.3 Second wave of LCCs (2000 – 2006) ...... 2-21 2.5.4 The ‘third’ wave (post-2006)...... 2-25 2.6 SUMMARY ...... 2-27

3. REGIONAL DISPERSAL PROPENSITY AND LOW-COST CARRIERS ...... 3-1

3.1 INTRODUCTION ...... 3-1 3.2 THE SPATIAL PATTERNS OF TOURISTS’ REGIONAL DISPERSAL...... 3-2 3.3 THE EFFECTS OF LCCS ON REGIONAL DISPERSAL...... 3-8 3.3.1 Spatial configuration of the destinations...... 3-11 3.3.2 Length of stay...... 3-12 3.3.3 Variety and multiple-benefit seeking behaviour...... 3-13 3.3.4 Risk and uncertainty reduction: distance travelled ...... 3-15 3.3.5 Heterogeneity in preferences (Travel party)...... 3-16 3.3.6 Trip arrangement (package tourism)...... 3-18 3.3.7 First timers, repeaters, and destination familiarity...... 3-19 3.3.8 Travel mode choice to and within the destination ...... 3-21 3.3.9 Socio-economic variables...... 3-23 3.3.10 Other variables and issues...... 3-24 3.4 SUMMARY ...... 3-25

4. THE ‘CHARACTERISTICS’ MODEL ...... 4-1

4.1 INTRODUCTION ...... 4-1 4.2 METHOD...... 4-3 4.2.1 Data ...... 4-3 4.2.2 The Model...... 4-4 4.2.3 Dependent and independent variables...... 4-6 4.3 RESULTS AND DISCUSSION ...... 4-9 4.3.1 Number of stopovers ...... 4-13 4.3.2 Length of stay...... 4-14 4.3.3 Distance ...... 4-15 4.3.4 Spatial configuration of the destinations...... 4-16 4.3.5 Accommodation Type ...... 4-17 4.3.6 Accompanying travel party type...... 4-18 4.3.7 Other variables...... 4-18 4.4 LIMITATIONS ...... 4-19 4.5 CONCLUSION ...... 4-20

5. THE CAIRNS EXPERIMENT...... 5-1 5.1 INTRODUCTION ...... 5-1 5.2 REGIONAL DISPERSAL AND TRANSPORT ...... 5-2 5.3 THE MODEL...... 5-5 5.4 ALTERNATIVES AND ATTRIBUTES ...... 5-6 5.4.1 Alternatives...... 5-6 5.4.2 Attributes and attribute level labels...... 5-12 5.5 EXPERIMENTAL DESIGN...... 5-15 5.5.1 Orthogonal main effects design ...... 5-15 5.5.2 Coding and design orthogonality ...... 5-16 5.5.3 The survey...... 5-17 5.6 RESULTS...... 5-18 5.6.1 Descriptive statistics...... 5-18 5.6.2 Model results...... 5-21 5.7 DISPERSAL AND RENTAL CARS...... 5-25 5.7.1 Transport attributes...... 5-25 5.7.2 Trip characteristics...... 5-26 5.8 DISPERSAL AND PUBLIC TRANSPORT ...... 5-28 5.8.1 Transport attributes...... 5-28 5.8.2 Trip characteristics...... 5-29 5.9 LIMITATIONS AND FUTURE RESEARCH...... 5-30 5.10 CONCLUSION ...... 5-32 APPENDIX 5.1 ...... 5-34

6. THE BALLINA-BYRON EXPERIMENT ...... 6-1 6.1 INTRODUCTION ...... 6-1 6.2 TOURISTS’ DISPERSAL ...... 6-2 6.3 THE MODEL ...... 6-5 6.4 DATA ...... 6-6 6.4.1 Case study region...... 6-6 6.4.2 Stated choice data...... 6-9 6.4.3 Choice alternatives...... 6-10 6.5 ATTRIBUTES OF MODAL ALTERNATIVES ...... 6-11 6.6 EXPERIMENTAL DESIGN AND SURVEY ...... 6-17 6.7 RESULTS...... 6-19 6.8 DISCUSSION AND IMPLICATIONS ...... 6-23 6.9 LIMITATIONS AND FURTHER RESEARCH...... 6-27 6.10 CONCLUSION ...... 6-29 APPENDIX 6.1 ...... 6-31 APPENDIX 6.2 ...... 6-32

vii

7. CONCLUSION, LIMITATIONS & FUTURE RESEARCH ...... 7-1 7.1 REVIEW ...... 7-1 7.2 KEY FINDINGS...... 7-2 7.3 CONTRIBUTION TO KNOWLEDGE AND IMPLICATIONS FOR STAKEHOLDERS ..7-5 7.3.1 Contribution to theory...... 7-5 7.3.2 Implications for policy...... 7-6 7.3.3 Implications for destinations...... 7-7 7.4 LIMITATIONS AND FUTURE RESEARCH...... 7-8 7.4.1 Applicability of the results...... 7-8 7.4.2 Limitations of the MNL: utility compensation perspective and taste heterogeneity ...... 7-9 7.4.3 Operationalising ‘dispersal’ ...... 7-11 7.4.4 Integrating destination and mode choice ...... 7-11 7.4.5 The time attribute in leisure and tourism...... 7-12 7.5 TOWARDS AN INTEGRATED MODEL OF INDIVIDUAL TOURISTS' SPATIAL CHOICE AND TOURISM YIELD ...... 7-14

REFERENCES…………………………………………………… .R-1

viii LIST OF TABLES

Table 1.1 Effects of LCCs on the regional dispersal propensity of visitors: intra- modal propositions ...... 1-8 Table 2.1 Summary of definitions...... 2-5 Table 2.2 Product features of LCCs and FSCs (NCs)...... 2-7 Table 2.3 Top 30 Australian domestic airports in terms of incoming passenger flows .....……………………………………………………………………….2-23 Table 3.1 Summary of the relationships discussed in section 3.3...... 3-10 Table 4.1 Summary of the relationships between LCC and dispersal...... 4-2 Table 4.2 Origin-destination sample...... 4-4 Table 4.3 Independent variables...... 4-8 Table 4.4 Model summary ...... 4-12 Table 4.5 Model results...... 4-13 Table 5.1 Three choice dimensions...... 5-7 Table 5.2 List of attribute level labels ...... 5-13 Table 5.3 Model summary ...... 5-21 Table 5.4 Model output: North...... 5-22 Table 5.5 Model output: South...... 5-23 Table 5.6 Inclusive value (IV) parameters...... 5-25 Table 6.1 Attributes...... 6-15 Table 6.2 IV parameter results...... 6-20 Table 6.3 Summary Statistics...... 6-20 Table 6.4 MNL estimation results...... 6-21

ix LIST OF FIGURES

Figure 1.1 Spatial representation of tourists' travel patterns...... 1-10 Figure 1.2 Schematic diagram of the thesis...... 1-26 Figure 2.1 Tourism Regions: An example of New South Wales...... ….……...2-4 Figure 2.2 Revenue Passenger Demand...... ….……...2-20 Figure 2.3 Domestic airfare indices...... ….……...2-21 Figure 2.4 Domestic revenue passenger growth from 1992/1993...... 2-24 Figure 2.5 Overnight trips made by air by purpose...... 2-25 Figure 3.1 Spatial representation of tourists' travel patterns...... 3-5 Figure 3.2 Travel party characteristics of air travellers...... 3-18 Figure 4.1 Regional dispersal: ground transport vs. air transport...... 4-10 Figure 4.2 Regional dispersal by airline...... 4-11 Figure 4.3 Marginal effects of stopovers on dispersal propensity ...... 4-14 Figure 5.1 Map of the Cairns region…………………………….………….….5-11 Figure 5.2 Sample choice shares across alternatives...... 5-20 Figure 6.1 Patterns of multi-destination travel...... 6-3 Figure 6.2 Map of Northern New South Wales...... 6-8

1. INTRODUCTION

This thesis aims to study the relationship between low-cost carriers and dispersal with the aid of Australian data and case studies. Discrete choice analysis is the approach adopted to examine the relationships. The aims, methodology, and anticipated outcomes of this thesis are introduced in this Chapter. In addition, this introductory chapter aims to provide sufficient information so that the reader can obtain a good sense of the links among the five ensuing chapters.

1.1. Low-cost air transport and dispersal

1.1.1. Research significance

When the U.S. domestic market was officially deregulated in 1978, average airfares came down, capacity increased, more airlines commenced services and the aviation network proliferated over the U.S. (e.g. Meyer and Oster 1987, Doganis 2002). One significant development subsequent to deregulation was the proliferation of a new kind of jet carriers in the 1980s. Meyer and Oster (1987) observed,

1-1

regulatory policies, and in keeping with the identity of these under-attended market niches, most of the new entrant jet carriers attempted to do something that predecessor established carriers did not. In most instances they either entered a market that was not previously served or entered a previously well served market while offering substantially lower fares” (p.49-50)

It was evident from the observations made by Meyer and Oster that the new entrant “jets” were loosely equivalent to what is today widely known as the Low Cost Carriers (LCCs) or Value Based Airlines (VBA). There are numerous variations to the LCC business model but research has shown that its low-margin, high-volume and low-fare foci are distinguishing features of the LCCs from network carriers (NC) (Lawton 2002). One significant source of variation within the LCC model is in the way they reduce costs. The cost reduction strategies manifest as characteristics of LCCs, which Gillen and Lall (2004) observed to be uniform fleet, greater use of airports excess-in-capacity, and the specific focus on maintaining a low-cost base in order to maintain the low-fare. Australia too has been subject to the entry of LCCs since the deregulation of the aviation sector in 1990, and as discussed in Chapter 2, Australian LCCs are also broadly congruent to the characteristics mentioned above.

Meanwhile, in recognition of the importance of tourism to regional economic prosperity (as well as alleviating urban congestion by diverting tourist flows), the Australian federal government prioritised the ‘greater regional dispersion of domestic and international tourists’ as a key policy goal in the medium term (former Department of Industry, Tourism and Resources (DITR 2003)). Although there has been a newly elected government in 2008, the emphasis on tourism policy to promote greater dispersal will remain an important policy agenda due to the continuing reliance of the rural regions on tourism for income and employment. For instance, the Jackson report (2009) prepared by the National Tourism Steering Committee to inform the development of a new National Long- Term Tourism Strategy outlined the significance of regional tourism, stating, “tourism provides opportunities for regional and remote communities to grow

1-2 jobs, diversify their economic base, and generate higher standards of living. Nearly half of total tourism expenditure (47 per cent) occurs in the regions” (p.10). Greater dispersal of visitors is maintained in the charter of federal tourism agencies.

Given Australia’s large and highly urbanised geographic characteristics, air transport is a vital form of transport for many tourism destinations located beyond the key metropolises. In some regional destinations, air services are the only real option for the accessing tourists. New air services have the potential to introduce tourism destinations to new markets. Such is the importance of air transport for dispersal, as part of the initiatives outlined in the Tourism White Paper (2003/04), the Australian government commits to ensure that “the airline services to regional destinations are considered as part of a broad Government policy to assist regional tourism” (DITR 2003). As for the definition of ‘regional dispersal’, in policy and practice, the regional dispersal of domestic tourists are defined as ‘a trip that involves at least one night stay outside the state capitals and the Gold Coast’1. Chapter 2 provides a more detailed description of the origins of the definitions adopted.

1.1.2. Low Cost Carriers’ effect on dispersal: an issue of spatial scale

The Australian LCCs exhibit the characteristics of low airfares, excess capacity airports and uniform fleet mentioned previously. The combined effects of these characteristics (the first two in particular) are positive for regional dispersal. Thus, LCC proliferation in the recent years can be viewed as an important agent that assists the government’s tourism policy of greater regional dispersal. More specifically, it can be viewed that the ‘dispersal of tourists beyond capital cities’ helps alleviate urban congestions and contribute towards moderating the imbalances in regional economic development across regions. Hence, in a way, LCC can be viewed as a distributive agent.

1 Dispersal of International tourists involves a stay in Sydney, Melbourne, Brisbane and Perth only, and excludes other state capitals such as Canberra, Adelaide, Darwin and Hobart.

1-3

However, the definition of ‘regions’ encompasses a very large geographic area. In fact, according to the standard definition adopted in the tourism industry, regional dispersal is always achieved as long as an overnight trip is undertaken beyond a few nodes (capital cities). While such measure is sufficient at the level of the federal governance, it is insufficient to bring more localised dispersal issues into light at the State and Territory level. The measure of regional dispersal at the state and local levels are more relevant for state and local governments, especially if they have some sort of regional economic development mandates in their charter (which may be the case for most states and territories). As illustrated later in this chapter, a distinction is made between domestic dispersal and regional dispersal in this thesis to reflect the differences in spatial scales.

1.1.3. The link between Low Cost Carriers and spatial behaviour of tourists

Bieger and Wittmer (2006) asserted that the 21st century proliferation of LCCs in Europe is the third revolution in aviation from the viewpoint of tourism, preceded by the charter sector in the 1970s and the aviation deregulation in the1990s. For tourism, one major consequence of the LCCs has been the generation of new tourist flows to existing and new destinations. This is not surprising given the fact that transport cost is a significant determinant of tourism demand (see for example, Crouch 1995, Sinclair 1998). The low airfares, however, were found to be associated with not only greater demand for tourism, but also demand for a certain type of tourism, as well as different travel characteristics.

An early evidence of the impact of affordable air transport (scheduled air transport) on tourist behaviour can be found in Mings and McHugh (1992). They studied the spatial configuration of the travel patterns of tourists travelling to Yellowstone National Park. They distinguished four types of spatial patterns: direct, partial orbit, full orbit and fly-drive. The pattern most profound at the time was the fly-and-drive pattern, which was associated with a number of key trip and traveller characteristics. Specifically, they observed that the fly-drive pattern was positively associated with increasing trip distance, number of visits to other

1-4 national parks, and length of the trip. Also there were more incidences of first time visits for this pattern compared to others. In regards to the socio-economic characteristics of tourists, they found that the fly-drive travel pattern was associated with tourists belonging to greater income and education levels. Mings and McHugh concluded, “perhaps this reflects increasing affluence, constraints on leisure time, and growing appreciation of … the American West” (p.46).

Mings and McHugh, however, did not explicate the link between the development in the aviation sector and the pattern of tourism they observed. Based on the inverse relationship between tourism demand and transport cost, we can deduce that the emergence of the fly-drive travel in the U.S. and the advent of new entrant jets, are causally related. Thus, we can propose a relationship between the new entrant jets, or the LCCs, and the spatial configuration of tourists’ travel patterns. Dispersal can be viewed as a special case of tourists’ spatial behaviour. Exploring the relationships between these two concepts, i.e. the LCCs and spatial behaviour, is the general premise of this thesis.

1.2. Research aims

1.2.1. Statement of the general aim

G1. Examine the effects of LCCs on the regional dispersal of domestic visitors in Australia.

The over-arching aim of this thesis is to examine the effects of LCCs on the regional dispersal of tourists in Australia. This necessarily involves explicating the theoretical link between LCCs and dispersal, as well as to empirically testing these relationships. It was previously hinted that the LCCs can be viewed as

1-5 agents of change in two facets: (1) in changing the volume of tourist flows to regional destinations and (2) in affecting the trip and tourist characteristics of these flows. To proceed, five specific aims are devised. Although each specific aims address different research questions, these aims were devised in a way that, collectively, contribute towards providing a thesis to the general aim (G1).

1.2.2. Statement of the specific aims

The first specific aim, A1 (see below), is addressed in Chapter 2. Related to this aim are two specific purposes. The first purpose is to provide the necessary background information on the Australian aviation environment to better understand the issues that this thesis aims to address. This is done by outlining the precursor to LCC growth in Australia, followed by a survey of international literature on the LCC models and characteristics. The second purpose is to discuss the Australian LCCs with a focus on their impact on dispersal. In order to do this, there is a need to distinguish between domestic dispersal and regional dispersal. As Chapter 2 will show in detail, domestic dispersal is appropriate for use at the federal level, while regional dispersal is more relevant for State and Territory governments. Furthermore, it is shown in Chapter 2 that while there is evidence of LCCs’ contribution to domestic dispersal in Australia, issues remain as to what the effects of LCCs are on the regional dispersal of tourists. Thus, A1 is to

A1. Provide an interpretative survey of the aviation and tourism research literature, and the secondary data sources relevant in understanding the link between LCCs and domestic dispersal (Chapter 2).

A1 addresses the issue of LCC and the volume of tourist flows. A2 (see below) addresses the characteristics of these flows from the regional dispersal viewpoint. In addressing A2, Chapter 3 aims to interweave the literatures of LCCs and regional dispersal to ascribe a cause-effect structure. Research on multi- destination travel and tourists’ spatial behaviour was found to be the most relevant literature in providing a conceptual framework for the LCC and dispersal

1-6 problem. In theory, at least two underlying sources can be responsible for the differences between the regional dispersal of tourists who used LCCs (called the ‘LCC tourists’ here after) and the regional dispersal of tourists who did not. One is due to the differences between the LCCs and the NCs. This is called the intra- modal source of difference. The second difference arises from the fact that LCC is a type of air transport, thus, it is also subject to the same constraints as all air transport services. This is called the inter-modal source of difference. These two sources form the basis of the cause-effect structure imposed in Chapter 3. A2 is to,

A2. Identify and explicate the relationships between regional dispersal and LCCs based on aviation, tourism and spatial behaviour research (Chapter 3)

The causal relationships between LCC and regional dispersal are summarised below (Table 1.1). Each hypothesis is explained in greater detail in Chapter 3.

1-7

Table 1-1. Effects of LCCs on the regional dispersal propensity of visitors: intra- modal propositions Factors Effects on regional dispersal LCC demand characteristics from propensity dispersal viewpoint 1. Spatial Different tourism regions will be Different tourism regions will be configuration of associated with different levels of associated with different levels of destinations dispersal dispersal

2. Length of stay Length of stay is positively related to LCC demand will be less sensitive to length dispersal of stay than NC demand

3. Variety and Greater variety in the reasons for Variety in the travel purpose, and the multiple-benefit travel, and larger share of VFR large share of VFR travels, are important seeking behaviour related travels, are positively sources of dispersal for the LCC arrivals related to dispersal

4. Risk and Greater risk and uncertainty about LCC demand may be more sensitive to uncertainty the trip may affect dispersal risk and uncertainty, hence the effect of positively or negatively distance on dispersal may be magnified

5. Heterogeneity in Greater heterogeneity in a travel LCCs serve proportionately more couples preferences group may affect dispersal and group travels, but there is no clear positively or negatively proposition on the differential effect of heterogeneity on dispersal between LCC and NC

6. First time or repeat First visitation can have a positive or LCC stimulates first-time visitors to the visitation negative effect on dispersal; repeat destination, which may increase or visitation has a positive effect on decrease dispersal. Second-home dispersal travellers are expected to be an important source of dispersal of the LCC arrivals

7. Package tourism Package tourism is negatively Disproportionately large share of LCC related to dispersal arrivals are FIT tourists, therefore, they are less constrained spatially.

8. Transport 'to' and Addressed in Chapter 5 and Addressed in Chapter 5 and Chapter 6 'within' the Chapter 6 destination

1-8

A2 explored and identified causal relationships between the LCCs and the dispersal propensity arising from intra-modal differences. The natural extension of A2 is to empirically test the proposed relationships. Thus, A3 is to

A3. Build and test a causal model of regional dispersal and the intra-modal differences (Chapter 4);

As previously mentioned, in addressing the issue of differential characteristics of tourists, it is useful to consider intra-modal and inter-modal effects. A2 identified the causal relationships between LCCs and regional dispersal arising from the intra-modal differences, while A3 empirically tests these relationships. The remaining problem is to examine the differences in the dispersal propensity sourced from inter-modal differences. As mentioned previously, the conceptual framework for this problem is based on tourists’ spatial behaviour and multi- destination travel research literature.

Previous research in the field identified a number of trip itinerary patterns, which were empirically found to be robust across spatial scales (e.g. inter-continental scale to local scale) and countries. The previously mentioned study by Mings and McHugh (1992) discovered that the majority of the variation in U.S. domestic trips to Yellowstone Park can be categorised into one of four trip structures: direct route; partial orbit; full orbit, and fly-drive. Lue et al. (1993) introduced structure to these itineraries in developing their conceptual framework for multi-destination travels. The trip itineraries were structured into five basic patterns of single and multi-destination trips, extending the Mings and McHugh’s four spatial patterns.

In this thesis, the trip patterns are grouped into three main types. This is shown in Figure 1.1, which integrates the five patterns proposed by Lue et al. into three patterns: single-destination (SDT), multi-destination type 1 (MD1) and multi- destination type 2 (MD2). SDT refers to a ‘direct-route’ travel that involves overnight stay in a single-destination. MD1 and MD2 represent two ways that LCCs can induce a change in the patterns of regional dispersal in Australia.

1-9

MD1

Regional tour/partial orbit b a c D SDT

Single destination/Base camp MD2

b D a En route b Trip chaining D a c d c

HOME

Figure 1.1 Spatial representation of tourists' travel patterns (modified from Lue et al. 1993)

Moscardo et al. (2004) have shown that the transport modes chosen by travellers to and within the destination constrain travellers’ travel patterns. Furthermore, they show that the ‘access points’ for transport, e.g. the location of the airport in relation to the wider destination region, affect the spatial pattern of the trip. Consequently, a shift in the destination access mode from air towards a car will be accompanied by a change in the trip type to the region. Since 2001, many regions in Australia were subject to LCC entry, increasing the importance of air transport for tourism in the regions. This also amplifies the importance of destination transportation for regional dispersal because the air leisure arrivals, unlike self- drive tourists, rely mostly on travel modes available in the destination.

A challenge for the government at the regional level may be to reconcile the potential conflicts arising from policy objectives that do not necessarily promote

1-10 the same travel mode best for regional dispersal. The examination of the destination travel mode choice of the air arrivals forms the next specific aim. This case study will be referred to as the ‘Cairns experiment’ hereafter, reflecting the name of the case study region.

A4. Examine the trade-offs between destination transport factors and tourists’ travel characteristics in the choice of the air arrivals’ regional dispersal (Chapter 5, ‘The Cairns experiment’)

Specifically, the following research question has been devised:

‘Can (and how) destination transportation policy stimulate the dispersal of the air arrivals, even in situations where the air arrivals exhibit trip characteristics that are dispersal-averse?’

Thus, the following hypothesis will be tested in the ‘Cairns experiment’:

‘Ground travel mode attributes and destination attributes can completely offset the negative effects on tourists’ dispersal propensity stemming from pre-determined trip characteristics’

The first inter-modal issue addressed by A4 pertained to MD1, while the second issue arise from MD2. MD2 includes the trip-chaining and en route patterns. One way that these patterns distinguish themselves from MD1 is through the main mode of travel used by the tourists. Air travel does not offer the spontaneity and flexibility of that offered by cars (e.g. Stewart and Vogt 1997:458). Thus in the regional tourism context, MD2 is difficult to achieve with air travel, but most easily with cars.

In MD2, the peripheral destinations (e.g. ‘a’ and ‘b’ in Figure 1.1) impacted by the LCCs are commonly en route. These destinations usually do not command large enough demand to sustain their own LCC services from ‘Home’. This will be a problem if there is substantial modal shift from the ground modes towards air

1-11 travel on the travel corridor. A sizable substitution effect will adversely impact regional dispersal because it will generate a bypass effect. This modal substitution issue is related to the final aim of this thesis. This case study will be referred to as the ‘Ballina-Byron experiment’ hereafter.

A5. Examine inter-regional travel mode substitution as a source of conflict between low fare air services and regional dispersal by applying a stated choice experiment (Chapter 6, ‘The Ballina-Byron experiment’)

Specifically, the following research question has been devised:

‘Can (and how) low airfares induce tourists to switch from car to air, even in situations where a car may be the most suitable mode of dispersal for the trip?’

The following hypothesis will be tested in the ‘Ballina-Byron experiment’:

Low airfares can induce tourists to switch from car to air by offsetting the positive utility gained from choosing a car, even in situations where the car may be the most suitable mode of dispersal for the trip.

The preceding discussion of MD1 and MD2 demonstrates that different regional destinations will be subject to different channels of LCC impact, i.e. via modal substitution (transport ‘to’ the destination) or modal complementarity (transport ‘within’ the destination and its relationship with the regional dispersal of the air arrivals). The trip patterns, individually or in combination of one another, enables a depiction of large variations of trips into a parsimonious set. When applied to a destination, the trip patterns generate specific LCC and dispersal issues. The applicability for a specific regional destination depends on which trip structure (SDT, MD1 or MD2) characterises the destination’s main demand. For instance, MD1 is most applicable to trips originating from Sydney or Melbourne, travelling to destinations along the Queensland’s Eastern Coast, whereas MD2 is most applicable to shorter trips (less than 800km).

1-12

A4 and A5 help fill a gap in the transport-tourism research. A4 concerns intra- regional mode choice, whereas A5 concerns inter-regional mode choice. Lumdson and Page (2004) noted a need for more cross-fertilisation between tourism research and the established field of transport economics, stating,

“modal competition has attracted highly quantitative and theoretical research by modelling travel behaviour. Yet the explicit tourism and leisure dimension remains a virgin area for research to understand the relationship between the potential for modal switching for pleasure travel rather than the prevailing focus of many transport studies on commuting”

While this thesis will not fully address the gap in the knowledge identified by Lumsdon and Page, it aims to make some progress by cross-applying methods established in the transport economics literature to the problems in tourism.

1.3. Notes on methods

1.3.1. Discrete choice models

The empirical work of this thesis applies several varieties of discrete choice models, namely the multinomial logit (MNL) and nested logit, as well as the basic logit model of binary choice. The aim here is to outline the common and the most fundamental aspects of the choice models applied in Chapter 4, 5 and 6. More information is provided in the methodology sections of each Chapter. The following explanations on discrete choice models are sourced mainly from a classic discrete choice analysis text by Ben-Akiva and Lerman (1985).

1-13

Random Utility Theory (RUT) forms the basis of discrete choice models. The microeconomic consumer theory maps consumption bundles of goods on a continuous space. By assumption, the quantity of goods consumed can be in integers. Consequently, calculus can be used to derive and solve demand functions for a utility maximised bundle of goods. The problem occurs with the standard method when it is applied to situations where a consumer chooses only one option from a number of mutually exclusive alternatives. This is because consumer choice necessarily implicates a consumption of only one good and zero consumption of other alternatives in the choice set, which results in “corner solutions” that cannot be solved by calculus (Ben Akiva and Lerman 1985). Random utility framework provides an alternative approach that overcomes this problem.

RUT assumes that the utility function for a given good can be decomposed into a non-random (or systematic) and a random (or stochastic) component. The randomness is assumed to arise from four sources (Ben Akiva and Lerman attribute this to Manski (1977)): the analyst does not observe all explanatory variables of the alternatives (or often called the ‘attributes’ in the literature); there is taste heterogeneity across individuals that analysts cannot observe; the analyst cannot measure and quantify the variables perfectly; and the use of proxy and instrumental variables results in a loss of information. Due to these reasons, at least some aspect of the utility of a good is uncertain. Random utility function can be expressed in the following way:

= + Ui Vi i Eq. (1)

where Ui is the utility level of a good i (or an ‘alternative’ as often referred to in the literature) and Vi is the systematic component of the utility (the part we can measure and observe), whereas i is the error term that represents the random part of the utility. A key feature of discrete choice models is the probability distribution ascribed to the error term. Different assumptions result in different discrete choice models. The most common assumption is the Gumbel extreme

1-14 value type I distribution. Daniel McFadden was the first to derive a multinomial logit (MNL) model. The MNL takes the following form:

eVni = pni Eq.(2) eVnj j

V where Pni denotes the probability of an individual n choosing alternative i, ni represents the systematic components of the utility described by the attributes, socio-economic and trip characteristics of alternative i for an individual n.

V Likewise, nj represents the observed variables for all alternatives in the choice set. In the MNL, it is the relative utility of one alternative to another that matters. When each of the random term (unobserved) is assumed to have Gumbel distribution, the difference in the random component of each utility function is

V logistically distributed. The linearly additive utility functions, ni , are first estimated from the data, and then it is transformed into probability estimates with the logarithmic function. Hence, the term ‘logit’ comes from the phrase, ‘logarithmic transformation’ (Louviere et al. 2000).

In Chapter 5 and Chapter 6 of this thesis, we primarily make use of the MNL. In particular, the results presented in these Chapters (Table 5.3 and Table 6.3) pertain to the coefficients of the utility function of the following form:

= + + + Vni i i X ni iTni iZni Eq. (3)

where Vni is the level of utility for individual n choosing alternative i. Vni is a function of the levels of the attributes X ni where i is a vector of coefficients to be estimated for each attribute of each alternative i. Tni is the trip characteristics where i represents the vector of coefficients for each trip attribute. Zni is the individual’s characteristics with coefficients vector i.

1-15

One reason why discrete choice models are such powerful analytical tools is because the comparison of coefficients provides trade-off information. Because all explanatory variables are expressed in terms of their contribution to the common unit called utility, when a coefficient of a ‘price’ variable is expressed as a ratio of other variables, monetary values can be ascribed. This subsequently has important interpretation as willingness-to-pay measures and the estimation of welfare and consumer surplus. For instance, a ratio of the coefficient of price to travel time provides a monetary value of travel time. Such interpretation is given when appropriate in this thesis, although in most cases, the MNL in this thesis is used to estimate the coefficients of the explanatory variables and to test for the statistical significance of these variables.

It should be noted that the model applied in Chapter 4 is a simplified version of the MNL. It is a logit model with only two available options. The model is of the form:

eVni = = pn (dispersal 1) Eq. (4) 1+ eVni where all terms are as defined previously. There is another methodological difference between Chapter 4 and the other two empirical Chapters. While Chapter 4 used ‘revealed preference’ data, Chapters 5 and 6 used ‘stated choice’ data. This difference is very important and it is explained in 1.4.2.

The IIA axiom and the limitations of MNL One important limitation in the MNL is the independence of the irrelevant alternatives (IIA) axiom. This property stems from the ‘independent and identically distributed error term’ assumption that gives the MNL the analytically convenient closed-form solution (Ben Akiva and Lerman 1985). In the axiom of IIA, “no provision is made for different degrees of substitutability or complementarity among the choices” (Hausman and McFadden, 1984: 1220). The IIA assumption is equivalent to constant cross effects in the MNL model (Ben

1-16

Akiva and Lerman 1985). Thus, the IIA assumption should be tested, and when the assumption is violated, non-IIA models should be considered. In the following section, the thesis presents commonly used IIA tests and other logit models that relax the assumption of IIA.

Hausman and McFadden (1984) show two ways to test the assumption of IIA. The first test shown below does not require an alternative model, whereas the second test does. For the former, they have shown that a violation of IIA will mean that the coefficients from a MNL with a subset of alternatives, i.e., the restricted model, will be statistically different from the coefficients estimated with all the alternatives, i.e., the unrestricted model. The Hausman-McFadden test provides a way of testing the differences in the coefficients. The test is:

= 1 q [br bu] [Vr Vu ] [br bu] Eq. (5)

br and bu indicate, respectively, vector of restricted (a subset of alternatives) and unrestricted (all alternatives are included in the model) model coefficients. V is the variance-covariance matrix of the estimated coefficients. q is the Hausman- McFadden statistic and this has a chi-square distribution with degrees of freedom equal to the number of coefficients in the restricted model.

Alternative discrete choice models relax the assumption of IIA. A natural extension to the MNL model is the nested logit model (Hensher et al. 2005). The nested logit partly relaxes the IIA assumption by partitioning similar or dissimilar alternatives. If alternative 1 and 2 are considered more similar than they are to alternative 3, then the central idea is that “the individual forms a weighted average of the attributes of alternatives 1 and 2, sometimes called the inclusive value, which is closely related to his consumer surplus” (Hausman and McFadden 1984:1227). Thus, a utility function can be specified to include a ‘composite’ utility (inclusive value),

= + + Vn n in X in (i+1),n IVn Eq. (6)

1-17

where n is a nest of similar alternatives (e.g, Jetstar and Virgin Blue may be in one nest and Train and Bus in the other). The inclusive value (IV) is defined as,

= V j IVn ln e Eq. (7) j n

The IV in the nest, n, can be viewed as a weighted average, or an expected maximum utility from a composite of alternatives (alternative j) (Hensher et al. 2005). The nested logit estimation procedure involves an estimation of an IV parameter for each nest. The IV parameter is a function of the scale parameter, which is assumed away in the MNL due to the independent and identically distributed error term assumption (hence, the scale parameter is absent from equation (2)). Scale parameter, , is equal to

2 = Eq. (8) 6 2

It can be shown that the IV parameter is equal to the ratio of a scale parameter from one nest and a scale parameter from a higher nest (Hensher et al. 2005). If the IV parameter is statistically equal to ‘1’, then the nested logit model collapses to a MNL. The IV test, which involves a Wald-test of significance on the IV parameter, is inclusive in the nested logit model estimation process.

The MNL is the ‘work horse’ in many applications due to its analytically convenient closed form solution (Hensher et al. 2005). By the same token, the MNL is limited in its ability to account for the individual variation in preferences, as well as in its ability to correctly predict market share in situations where the IIA axiom is violated. Given the fact that (1) the thesis aims for understanding than prediction (the former leads to the latter but not the other way around – Louviere et al. 2000); (2) the thesis aims to better understand the general relationships between the independent and dependent variables rather than the taste heterogeneity across individuals, the impact of the MNL’s limitations are

1-18 minimized. The empirical chapters of this thesis apply primarily the MNL, and when appropriate, the nested logit.

While there are extensions to the MNL and nested logit, such as random parameter logit (mixed logit), the application of these models are beyond the scope of this thesis. The random parameter logit (RPL) model enables the estimation of a unique coefficient on the X variables for each individual. However, this research does not need the RPL because the thesis focuses on the estimation of the signs and weights of the coefficients of the independent variables, not the individual variation in the coefficient estimates for each independent variable. This thesis in the final Chapter (Chapter 7) dedicates a section to discuss how the mixed logit models can improve and extend this research.

1.3.2. Stated choice data

In many econometric applications, data are collected on the choices already made in the market. That is, we observe the choices made by consumers in the market and the attributes of the chosen goods and services such as price and product characteristics (Louviere et al. 2000). This type of data is collectively called ‘revealed preference’ or RP data. The National Visitor Survey data used in Chapter 4 is a data of this sort where survey respondents are asked to recall the trip they made in the past and the various aspects of that trip. Although such data are common for various reasons, in social sciences, it is particularly common because data from experiments are not readily available, and often ethically and instrumentally infeasible. Nonetheless, RP data are subject to limitations in the context of choice. Louviere et al. (2000) provided a number of reasons.

Perhaps the most interesting reason concerns the limited variability and high collinearity of the values of explanatory variables in the marketplace. This is partly because competitors match prices, and product features remain constant

1-19 over time. For example, the time it takes to travel from Sydney to Ballina-Byron by car or air; or the qualitative features of public transport services such as driver ‘quality’, which may remain rigid over time due to the time it takes to set up new training programs and the lag time involved with publicly funded projects. Consequently, choices observed on RP data tend to be poorly conditioned (Hensher et al. 2005). Interestingly, Louviere et al. (2000) argued,

“as markets mature and more closely satisfy the assumptions of free competition, the attributes of products should become more negatively correlated, becoming perfectly correlated in the limit … technology drives other correlations between product attributes, so as to place physical, economic or other constraints on product design. For example, one cannot design a car that is both fuel efficient and powerful because the laws of physics intervene. Thus, reliance on RP data alone can (and often does) impose very significant constraints on a researcher’s ability to model behaviour reliably and validly (p.22, parenthesis in original text).”

There are two additional reasons why stated choice data were used in the second and the third empirical studies. One is due to the lack of availability of secondary data sources on the alternatives and the alternatives’ attributes (as well as the alternatives’ attributes’ levels) considered by the decision maker in the choice process. This is particularly the case with data on airfares. Louviere et al. (2000) argued that by creating these data based on a rigorous scientific experimental design procedure, we are able to formulate a causal model of choice, with the added advantage of reducing the invalid inferences from ‘chance’ relationships.

The second reason is related to the fact that stated choice data is capable of accounting for new product features or alternatives that currently do not exist. Given the aim of Chapter 5, it was necessary to create a public transport alternative with some hypothetical attributes. Eaton and Holding (1996) concluded that ultimately, public projects need to be able to induce a change in behaviour - in their use of transport mode from private cars to public vehicles - for policy to be effective. Given the fact that such a policy can be expensive and

1-20 riddled with conflicting interests, as mentioned previously, Eaton and Holding advocated an ‘experimental’ approach to first demonstrate the potential. This is similar to the feasibility assessments akin to transport project appraisals for airports and road infrastructure upgrades, which are often irreversible and have very high fixed costs.

1.4. Contribution to knowledge

1.4.1. Contributions and limitations

Tourism research to this day has largely neglected an analytical approach to assessing the trade-offs between travel mode choice and spatial behaviour. This thesis contributes by providing a utility compensation perspective on the tourists’ choice of transport and the resulting spatial behaviour of tourists. The utility compensation perspective highlights the importance of trip characteristics in a way that can be directly compared to the importance of travel mode attributes. Although not without limitations (discussed in Chapter 7), this approach enables a comparison of the utility gained from paying low airfares with the utility associated with trip context such as length of stay, trip structure (single or multi- destination), or the level of destination familiarity.

Although the application of discrete choice models (micro-econometric choice models) is advanced in transport mode choice research, it is mostly applied in a journey-to-work and intra-urban trip context. In long-distance travel applications, the theoretically important tourism variables are often not included in the analysis. This thesis extends the analysis beyond the traditional economic variables of mode choice by including theoretically significant tourism variables in the long- distance leisure trip context. It is shown in this thesis that in long-distance leisure travels, trip characteristics vary widely across individuals and travel parties, and

1-21 these have significant influence on the choice of travel modes, sometimes to an extent that trip characteristics offset the marginal benefit gained from the changes in travel mode attributes. Therefore, while this research extends the boundaries to which discrete choice models can be applied, the real theoretical contribution of this thesis is in highlighting how our understanding of the relationship between long-distance leisure mode choice and spatial distribution of tourists can improve by accommodating tourism variables in the discrete choice framework. Thus, this thesis’ contribution extends beyond the demonstrating of the applicability of discrete choice models to new settings; rather, the application of choice analysis to long-distance leisure trips raises interesting questions about the choice models. This point is revisited in Chapter 7.

The results from this thesis should be relevant in a setting where the geographic region is large and multi-modes of transport are real options for tourists. However, the results are also sensitive to context; the relevance of the results should be assessed with caution in settings where the transport market is regulated by the government. In other words, the assumption made in this thesis is that all travel modes are free to enter/exit the market, as well as to set their fares and capacity without government intervention, i.e., a deregulated market. Nonetheless, the results should be relevant to assessing the spatial impact of tourism in large developing economies undergoing deregulation of the transport markets (particularly the aviation market). Deregulation of the aviation market is likely to increase discount fares, thereby decreasing average fare levels, which will generate new demand for tourism. This thesis will provide insight into the dispersal impact of such changes for secondary and regional tourism destinations.

1-22

1.4.2. Key stakeholders

State and local governments The findings from this thesis should be of relevance to government with mandates emphasising the greater balance in the distribution of economic benefits across the regions. Transport issues are often at the centre of public policy agenda where the government may promote certain modes of travel over others to meet a wider policy objective (e.g. reduce carbon emissions). Conflicts may arise between policy objectives such as dispersal and environmental preservation; for instance, while car is a pertinent mode of travel for regional dispersal, environmental policy may advocate a shift away from car towards public transport. Furthermore, local level tourism and transport planning issues have a strong political dimension because the competition for public funds increases at this level of governance (Gunn 1988). Thus, information on the trade-offs between travel modes and regional dispersal contribute towards providing diagnostic information, which will help in overcoming these conflicts.

Domestic airports and airlines Australian domestic airports serving regional tourism destinations - regardless of the ownership structure - usually have tourism development objectives in their charter in recognition of the mutually beneficial relationship between the growth of airports and growth of tourism. For smaller regional airports, significant investment is necessary to be able to facilitate the entry of LCC services; for example, on runway and terminal space upgrades and purchase of security equipment. The information on modal substitution and the associated changes in the tourists’ travel patterns will help assess the impact of such investments. While airlines are concerned mostly with the demand for air services between two points, increasingly, the ancillary revenue is becoming an important aspect of LCC business (CAPA 2008). Better understanding of passenger travel behaviour in the destination can increase airlines’ ancillary revenues; for instance, such information can help airlines to exploit opportunities for partnerships and financial innovations with tourism businesses.

1-23

Destination marketing organisations Since the use of one travel mode over another will be associated with a particular travel itinerary, i.e. different patterns of dispersal, mode choice studies can help identify ‘linkage patterns’ of different destinations. Consequently, such information contributes to recognising natural partners in regional or locational cooperation (Opperman 1995; Lue et al. 1993). This is particularly relevant for state tourism organizations whose roles are to facilitate liaison and provide cooperative marketing for the diverse range of tourism regions within their jurisdiction.

1-24

1.5. Structure of the thesis

As mentioned earlier, the general aim of this thesis is to examine the effects of LCCs on the regional dispersal of domestic visitors in Australia. This goal is subdivided into five specific aims. There are five Chapters (Chapter 2 to Chapter 6) that sequentially address A1 to A5. The structure of this thesis is summarised in the schematic diagram below (Figure 1.2). The aims of this thesis are reiterated here, acknowledging that the aims are to understand why relationships occur as well as how they occur.

A1. Provide an interpretative survey of the aviation and tourism research literature relevant to understanding the link between LCCs and domestic dispersal (Chapter 2);

A2. Identify and explicate the relationships between regional dispersal and LCCs based on aviation, tourism and spatial behaviour research (Chapter 3);

A3. Build and test a causal model of regional dispersal and the intra-modal differences between LCCs and NCs (Chapter 4);

A4. Examine the trade-offs between destination transport factors and tourists’ travel characteristics in the choice of the air arrivals’ regional dispersal (Chapter 5);

A5. Examine inter-regional travel mode substitution as a source of conflict between low fare air services and regional dispersal (Chapter 6).

1-25

The effects of LCCs on dispersal

Effect on the volume Effect on travel characteristics of tourist inflow and dispersal propensity (A2, (Chapter 2) Chapter 3)

Dispersal propensity Dispersal propensity differential sourced from differential sourced from intra-modal differences inter-modal differences (LCC vs. NC) (Air travel vs. car travel) (A3, Chapter 4) (A2, Chapter 3)

Dispersal propensity Dispersal propensity and intra-regional and inter-regional mode choice (travel mode choice (travel mode choice within mode choice to the the destination) destination) (A4, Chapter 5) (A5, Chapter 6)

Figure 1.2. Schematic diagram of the thesis

1-26

Chapter 2 introduces the necessary background information on LCCs and regional dispersal. The Chapter discusses the defining features of LCCs and their strategies in lowering costs. The Chapter also introduces the Australian aviation environment and provides an outline of the Australian LCC history. In this Chapter, ‘domestic dispersal’ is distinguished from ‘regional dispersal’. It is argued that while LCCs have contributed to the domestic dispersal of tourists in Australia, more research and data are required to examine the effects of LCC on the regional dispersal of tourists.

Chapter 3 interweaves the literatures on LCCs and the literatures on regional dispersal to impose a cause-effect structure between the two concepts. A distinction is made between intra-modal differences and inter-modal differences. Chapter 3 identifies and explains the relationships between LCCs and dispersal sourced from intra-modal differences. Chapter 3 also provides a literature review that forms the basis for the research issues examined in the Cairns case study (Chapter 5) and the Ballina-Byron case study (Chapter 6).

The empirical work in this thesis is framed in three inter-related empirical research issues. Chapter 4 empirically tests the propositions put forward in Chapter 3, which addresses the intra-modal issue. This is named in this thesis as ‘the characteristics model’. The remaining two Chapters address the regional dispersal issues stemming from the fact that LCC is a form of air transport. Chapter 5 concerns the mode choices made by the air arrivals within destination regions, and the travel modes’ links with regional dispersal. Chapter 6 focuses on the travel mode choices in travelling to the regional destinations. Both Chapters provide an introduction to the research problem before proceeding to the details of the methods used, including the details on the experimental design for the stated choice experiments. The studies are named the ‘Cairns experiment’ and the ‘Ballina-Byron experiment’ respectively. Chapter 7 is a concluding chapter. Research limitations are also discussed, along with future research directions.

1-27

2. DISPERSAL AND LOW COST CARRIERS

2.1 Introduction

Low Cost Carriers (LCCs) are airline business models with the primary aim to achieve lower cost structure. The various strategies they employ to achieve the low cost manifest as a common set of characteristics that are in many ways different from the network carriers (NCs). The core characteristics of LCCs are the offering of affordable airfares and point-point services. These are desirable features from the regional tourism destinations’ perspective because they help stimulate tourism demand. This is also a policy priority for governments willing to alleviate congestion in urban centres, and to capitalise on the economic benefits that tourism is capable of generating for the regions.

The primary aim of this Chapter is to introduce the two central concepts of this thesis - dispersal and LCCs. In doing so, we accomplish the first specific aim of this thesis. Related to this aim are two specific purposes. The first purpose is to provide the necessary background information on the Australian aviation environment to better understand the issues this thesis aims to address. This is done by outlining the precursor to LCC growth in Australia, which is followed by a survey of international literature on LCC models and characteristics. The second purpose is to overview Australian LCCs with a focus on their impact on dispersal. In order to do this, this Chapter first distinguishes between domestic dispersal and regional dispersal. It will be shown that while there is evidence of LCCs’ contribution to domestic dispersal in Australia, issues remain as to what the effect of LCCs are on the regional dispersal of tourists. Regional dispersal is the primary focus of the subsequent Chapters.

2.2 Dispersal

2.2.1 Definition of ‘regions’, domestic dispersal and regional dispersal

A geographical unit, Tourism Region, is important for the definition of dispersal in this thesis. Each state and territory tourism organisations in Australia delineate its territory into a number of Tourism Regions. The delineating method differs for each state, resulting in a variety of number and sizes of Tourism Regions. Tourism Regions are also revised almost every year. In 2007, there were 89 Tourism Regions in Australia.

Australian government agency such as the former Bureau of Tourism Research (Tourism Research Australia as of 2004) defines ‘rural’ as tourism regions outside Adelaide, Brisbane, Canberra, Darwin, Hobart, the Gold Coast, Melbourne, Perth and Sydney. To be precise, this is a de facto definition because the official definition only excludes capital cities. Gold Coast is not a capital city but it is now standard practice to exclude this destination from the regions. Although this definition excludes other large populous centres in the rural regions, these cities are only a small fraction of ‘rural’ Australia, and visitors to these cities “may still pursue activities and experiences which are non-urban in character” (O’Halloran et.al. 2000:60). Often in practice, the term ‘regional’ is used as a synonym for ‘rural’ (Kelly 2001:1 as cited by Centre for Regional Tourism Research). Thus,

2-2 when we refer to ‘regional dispersal’, it is a trip that involves at least an overnight stay in a tourism region other than the cities listed above.

While this thesis adopts the same definition to preserve consistency with government agencies, another branch of classification is added to ‘rural’ to further differentiate the main cities in each tourism region from the rest. This is necessary because a rural visitor - a person staying at least one night in the ‘rural’ tourism region - includes those who visited the main city. If a visitor spent all of his or her overnight stay(s) in the city, then this visitor’s trip is not ‘rural’ in its character. To be more geographically specific in the definition of ‘regional dispersal’, distinction should be made between a visitor who stayed in the main city and a visitor who stayed in the rural regions.

In this thesis, rural regions are divided into ‘cities’ and ‘all other’ regions. In the tourism literature, these cities are often referred to as ‘gateway cities’. Gateway cities provide most of the functional facilities for tourists, as a transport hub for instance, acting as a main point of entry and exit for tourists visiting the wider region (Lew and McKercher 2002). These points in the tourism regions are referred to ‘gateways’ and the remainder ‘periphery’. In Figure 2.1, gateways are Coffs Harbour and Ballina-Byron within the tourism regions of North Coast and Northern Rivers respectively. All other areas of the tourism regions outside these gateways are the periphery. As it will be shown next, a trip to the regions will be referred to ‘domestic dispersal’ and a trip that involves at least one night stay in the periphery, ‘regional dispersal’. This distinction is important; while dispersal as currently defined by the tourism industry remains relevant for tourism and transport policy at the federal level of governance, at the local level, dispersal is important to an extent that travellers diffuse from gateways into peripheral destinations. Definitions are summarised in Table 2.1.

2-3

Figure 2.1 Tourism Regions: An example of New South Wales (based on Australian

Bureau of Statistics Tourism Regions Map release 2007)

Trips of interest in this thesis are those originating from capital cities and Gold Coast. Over 60% of the Australian population resides in these cities (ABS 2007). Many rural trips, however, do originate from non-capital cities (40% of the Australian population live outside these regions). Nonetheless, the focus is on the cities mentioned previously because these are the main origins for domestic air travel demand. In fact, at the time of writing, all domestic routes by Qantas (excluding regional subsidiaries and Qantaslink), Jetstar, Virgin Blue and Tiger airways were either from/to the ‘capital cities or Gold Coast’.

Therefore, domestic dispersal represents trips originating from capital cities (and Gold Coast) destined for the regions. Regional dispersal involves trips that entail at least one night stay in the regions beyond the gateway cities, i.e. peripheral destinations. Regional dispersal includes multi-destination trips as long as the trip 2-4 involves at least one night stay in the periphery. As per domestic dispersal, a trip must originate from capital cities or Gold Coast for it to constitute a regional dispersal. Thus, regional dispersal is embedded in domestic dispersal.

Table 2.1 Summary of definitions

Regions refer to all geographic areas outside the capital cities and Gold Coast tourism regions. Capital cities are Adelaide, Brisbane, Canberra, Darwin, Hobart, Melbourne, Perth and Sydney.

Tourism regions are regional boundaries classified by state tourism organizations. Each state has a different number of tourism regions varying widely in geographic size and the extent to which tourism contributes to the regional economy.

Gateways are the main points of entry and exit for tourists in a given tourism region. Usually, these are the largest cities in respective tourism regions, and each city has an airport (often of the same name as the city) with regular ‘domestic’ air services.

Periphery or peripheral destinations are destinations within tourism regions located beyond the geo-political bounds of the gateway city. These destinations vary in its reliance on tourism ranging from towns to small rural communities.

Domestic dispersal occurs when a trip (1) originates from capital cities (and Gold Coast); and (2) involves at least an overnight stay in tourism regions other than the capital cities (and Gold Coast), i.e. overnight stays in gateways or peripheries.

Regional dispersal occurs only when a trip (1) originates from capital cities (and Gold Coast); and (2) involves at least one overnight stay in a peripheral destination during the trip.

Source: created by the author

2-5

2.3 The characteristics of LCCs

2.3.1 The LCC model

Low-cost carrier (LCC) is a type of airline business model pioneered by Southwest airlines in the U.S. (O’Connell and Williams 2005, Lumsdon and Page 2004, Gillen and Lall 2004, Lawton 2002). It is difficult to provide a definition applying to all low-cost carriers (LCCs) due to the numerous variants of LCCs (Francis et al. 2006). There has been an explosion in the air transport research literature that addressed the definition and characteristics, especially following the successes of European adaptation of the LCC model - Ryanair and Easyjet - since the late 1990s (e.g. Dobruszkes 2007; O’Connell and Williams 2005; Page 2005; Francis et al. 2004; Gillen and Lall 2004; Burghouwt et al. 2003; Lawton 2002; Williams 2002 and 2001).

In general, LCC is an airline business model aiming to have a low cost base to offer lower airfares (Lawton 2002). Lawton (2002) shows that the LCC model is a low-margin and high-volume airline business that relies on the virtuous circle of demand stimulation and economies of density (reduction in unit costs as a result of greater demand density). O’Connell and Williams (2005) summarised the key features of LCC models worldwide. Table 2.2 is based on O’Connell and Williams (2005) detailing the differences in the product features of LCCs and NCs. In many situations, an airline regarded as a LCC will have a combination of these features. It is widely observed that LCC services are predominantly point- to-point, short-haul, and to a less extent, have a uniform fleet, although this is as far as the similarities between LCCs go (Gillen and Lall 2004). Some of the key features are discussed in greater detail below.

2-6

Table 2.2. Product features of LCCs and FSCs (NCs) Product features Low cost carrier (LCC) Full service carrier (FSC or NCs) Brand One brand: low fare Brand extensions: fare+service Fares Simplified: fare structure Complex fare: structure+yield mgt Distribution Online and direct booking Online, direct, travel agent Check-in Ticketless Ticketless, IATA ticket contract Airports Secondary (mostly) Primary Connections Point-to-point Interlining, code share, global alliances Class segmentation One class (high density) Two class (dilution of seating capacity) Inflight Pay for amenities Complementary extras Aircraft utilisation Very high Medium to high: union contracts Turnaround time 25 min turnarounds Low turnaround: congestion/labour Product One product: low fare Multiple integrated products Ancillary revenue Advertising, on-board sales Focus on the primary product Aircraft Single type: commonality Multiple types: scheduling complexities Seating Small pitch, no assignment Generous pitch, offers seat assignment Customer service Generally under performs Full service, offers reliability Operational Focus on core (flying) Extensions: e.g., maintenance, cargo activities (adopted from: O’Connell and Williams 2005)

2.3.2 Point-to-point network (P2P)

A dominant pattern of airline network emerged following the deregulation in the U.S. was the hub and spoke system (HSS) (Meyer and Oster 1987, Doganis 2002, Franke 2004). Franke (2004) has shown that the HSS allows the maximisation of coverage over origin-destination pairs and different customer segments by concentrating the inbound flights into a single hub, while maximising the connectivity for the outbound flights from that hub. This process, however, is inherently complex and entails inefficiencies, which are paid by the passengers through higher fares and inconveniences (stopovers). Franke summarised the negative consequences, stating,

2-7

“the negative aspects of this strategy are a loss of convenience for the passengers (who would prefer direct flights), and a considerable cost penalty for the airline on the operational side. Waved traffic means massive peaks in hub operation, resulting temporary congestion (reduced airside productivity), time-critical connections (special processes required), and strongly fluctuating utilisation of ground handling facilities/workers (reduced landside productivity). Furthermore, congestion plus a multitude of time- critical connections typically lead to poor punctuality performance” (Franke 2004: 16, parentheses in original)

The point-to-point (P2P) strategy represents a case on the other side of the extreme where each destination-origin pair is served directly. In a pure P2P, passengers will use each airport as entry and exit points than as a connecting point. Gillen and Lall (2004) argued that Southwest airlines primarily derives its low cost from this strategy. As Gillen and Lall argued, P2P is the most important feature that enables the fast turn-around of aircraft at airports. This enables the airline to avoid the costly delays associated with connections in hubs. This quick turnaround maximises the aircraft utilisation rate per day. Given that an aircraft is one of the most expensive investments an airline makes, maximization of its use is the most important source of cost saving; for example, a 25 minute turnaround compared to an one hour turnaround will yield an extra two return services on a given day, which results in greater fleet utilisation and staff productivity (Barrett 2004).

Reynolds-Feighan (2001) studied the traffic concentration patterns of LCCs and NCs. He found that the domestic aviation network in the U.S. decentralised over the period between 1969 and 1999. Reynolds-Feighan (2001) argued that this is due to the LCCs’ P2P network. While LCCs, on average, have lower levels of traffic concentration, Reynolds-Feighan also found considerable variations within the LCCs: there were LCCs operating a single hub and spoke system (e.g. America West, TransAir); as well as pure P2P (e.g. Southwest). Swan (2007) in fact argued that the conjecture that LCCs are P2P is a misleading oversimplification because even the ‘purists’ such as Ryanair provides significant

2-8 level of connecting traffic – 15% (Southwest with 30%) - in comparison to traditional carriers, which is often 50%. In fact, Franke (2004) argued that the route network provided by the P2P strategy by some LCCs are so comprehensive that there are opportunities for ‘random connections’ by the passenger themselves.

In Europe, Dobruszkes (2007) concluded that post-deregulation network patterns were largely induced by the LCCs. He observed that the European airline networks changed from a ‘radial’ pattern to a ‘star-shaped’ pattern following the proliferation of Ryanair and Easyjet. While not as spatially comprehensive as that of U.S. or Europe, the Australian LCC network broadly resembles the P2P network structure of U.S. and Europe. Sinha (2001) has shown that the Australian network is mostly P2P because it has a high level of demand concentration on a few large nodes; namely, the demand is concentrated between the state and territory capitals.

2.3.3 Use of secondary and regional airports

Secondary airport refers to an airport providing a ‘secondary’ access to a major population centre. One well-known example of a secondary airport is London’s Stanstead airport used by Ryanair, which is located relatively peripheral to Heathrow and Gatwick. Regional airports, on the other hand, provide access to those travelling to or/and from smaller regions and cities, rather than acting as a substitute to a major gateway airport in large metropolitan centres.

Often, regional and secondary airports are excess-in-capacity; therefore, much less conducive to congestion and delays (Gillen and Lall 2004). Warnock-Smith and Potter (2005), based on a survey of managers across eight LCCs in UK and Europe, found that ‘quick turnaround facilities’ and ‘convenient slot times’ were the top considerations in the choice of airport for entry decisions. In many routes where one end is characterised by a busy and congested airport, and the other, by a secondary or regional airport, convenient slot time introduces greater flexibility

2-9 in scheduling for airlines. Excess-in-capacity airports are much more likely to provide this flexibility than the congested ones.

The Warnock-Smith and Potter (2005) study found that ‘discounts on aeronautical charges’ ranked fourth in importance. Discounts on airport costs can be an important source of cost saving for LCCs. This is because the share of aeronautical charges of total costs will be higher for LCCs than it is for legacy carriers (Lawton 2002). Moreover, during negotiation, LCCs will have greater leverage with the smaller airports because fewer airlines serve them. A widely documented case is Ryanair, which threatens to fly elsewhere if their terms are not met (terms with respect to landing fees, bridge fees, passenger fees, etc). The fact that 93% of Ryanair’s routes are exclusive to the airline (as at the end of 2005; see Dobruszkes 2007) provides some indication as to how important the secondary or regional airports are to Ryanair’s cost reduction strategy.

In summary, the study by Warnock-Smith and Potter found the following factors important in airport choices (from most important to least important):

o high existing demand for LCC services; o quick turnaround facilities; o convenient slot times; o good aviation fee discounts; o positive economic forecasts for the region; o efficient airport management; o high level of airline competition; o good experience of LCCs; o good non-aviation revenues and ownership;

While slot times and turnaround facilities are important determinants of airport choice, adequate demand is the most important factor. Consequently, LCCs tend to favour entry on routes with dense demand and excess-in-capacity airports. As shown later in this Chapter, these characteristics have the combined effect of

2-10 stimulating leisure demand for regional tourism destinations in Australia, i.e. contributes to greater domestic dispersal of tourists.

2.3.4 Short-haul and Low-cost customer service

Most LCC services are short-haul services. Only recently the long-haul adaptation of the LCC model has emerged; for example, with AirAsia X and Jetstar International. ‘Short-haul’ typically refers to up to 3 hours in flight duration. Short-haul flights characterise many flights within large domestic markets such as the U.S., Brazil and Australia, and highly liberalized international markets such as intra-European routes and Trans-Tasman routes between New Zealand and Australia. Thus, the LCC model is highly compatible with the patterns of air transport demand within Australia, as well as between Australia and New Zealand.

The short-haul focus provides a number of important cost advantages. First, it enables the airline to have a uniform fleet, typically that of B737s and A320s, which were proven popular among LCCs due to cost efficiencies for the short- haul stage lengths. For instance, Southwest airlines commands more than 500 B737s and no other type of aircraft, and similarly, Ryanair has a fleet of B737s as does Virgin Blue in Australia (up until the end of 2007). Having a uniform fleet generates significant level of economies of scale in maintenance costs (Lawton 2002, Gillen and Lall 2004, Franke 2004, etc). For instance, Hansson et al. (2003) estimated that 13% of the cost differences between European LCCs and NCs came from lower maintenance costs (as cited by Franke 2004). Long-range aircraft is needed if an airline is to provide long-haul services, and this will require that the airline diversify not only in fleet composition but also in the maintenance and infrastructure costs at the airport. Furthermore, since long-haul services are more likely to need to draw traffic from a wider market catchment, there will be a need for greater coordination with the feeder and spoke services at the origin and destination. This adds to the overall complexity of the airline operation, which is inconsistent with the LCC model.

2-11

The short-haul focus is also closely linked to the customer service costs. LCC is a low-cost customer service model with minimum level of perks (Lawton 2002). The low-cost in-flight service has several advantages for the LCCs. A widely observed feature of the LCC service is the absence of free meals or snacks, in- flight entertainment, and lower staff per passenger ratio. This enables the airline to reduce its cost base, as well as providing the airline with a source of ancillary revenue by charging extra for these services. Second, Gillen and Lall (2004) noted that the avoidance of catering for hot-meals, for example, is possible due to the short-haul nature, and contributes toward reducing the turnaround time. Finally, the short-haul flights enable the LCC to configure its aircraft into single class service. Multi-class configurations add extra costs and detract the focus from the low-margin and high-volume model.

The short-haul focus, therefore, enables the LCC to derive its lower cost structure from simplicity. This strategy is linked to the low-cost customer service, and the greater dependence on ancillary revenue. It is also the case that short-haul services are closely linked to the core strategy of point-point services, which enables faster turnaround time and avoids the complexity in operating a multi-aircraft fleet.

2.3.5 Ticket distribution, fare structure and passenger handling

Internet technology enabled LCCs to reduce distribution costs by bypassing intermediaries. Although travel agents and call centres were replaced by internet for many airlines (not only LCCs), LCCs have generally embraced the internet technology to control costs (Duval 2008). In fact, Hansson et al. (2003) estimated that 15% of the cost difference between European LCCs and NCs comes from ‘innovative direct sales’ and ‘lower Global Distribution System (GDS) charges’ (as cited by Franke 2004).

Simplified fare structure of the LCC not only reflects the differences in the cabin class, such as business vs. economy, it also reflects the differences in the revenue

2-12 management practices and inter-lining employed by the airline. For instance, ‘full economy’ fares with flexible ticket-change conditions by the network carriers are designed for business travellers, with high fares covering the costs of insurance against over-booking, and to compensate for the potential schedule changes by the traveller (Mason 2006). This contrasts with a pure LCC, which offers a single-leg and non-refundable ticket that can be changed to a different flight with a fixed administrative fee and the price differential (Mason 2006). The NCs tend to implement many fare options. In contrast, LCCs often offer a more simplified ticket structure to undercut the NCs on price (Marcus and Anderson 2008).

LCCs’ simplified fare structure has made fares more transparent and amenable for interpretation by consumers. In a P2P network (cf. section 2.3.2), each ‘leg’ of the trip is purchased as an independent trip. If the traveller is travelling two ‘legs’, then the airline treats this as two separate trips. In many instances, even if a traveller is travelling on the same airline on two legs, the traveller needs to check- in and collect her baggage twice. For NCs with hub and spoke networks, the two leg journeys are sold as a single product to the traveller. However, the convenience of the single check-in and baggage transfer is often done at the expense of increasing level of complexity of interlining and coordination with other airlines. In such cases, the airlines usually have agreements on the number of seats per aircraft that can be sold as a two-leg bundled product, or independently sold as a separate journey. This complexity causes the fare itself to be complex because a situation where a two-leg trip is much cheaper than a single-leg (on the same route) can arise. As Clippinger and Strong (1987) noted following the deregulation in the U.S., travel agents were the “official interpreters of the mysteries of air travel” (p.125). LCCs were able to eliminate this problem because their fare structure and network strategy were simple; eliminating one of the important roles of travel agents.

With respect to passenger handling in airports, LCCs often lead the implementation of cost reducing innovations (Swan 2007). Electronic check-ins and electronic tickets reduce labour costs, and effectively transfer the onus of

2-13 some traditional check-in tasks to the passengers. The benefit of unallocated seating is that it gives incentives for passengers to arrive early to secure a good seat, which contributes to reducing delay risks (CAPA 2007). Some LCCs (e.g. Jetstar and Ryanair) offer reduced fares to hand-carry only passengers. Less baggage not only contributes to lower fuel costs (because of lower weight), but also speeds up the check-in process, which reduces the risk of delay and congestions at airport check-in counters. Most of the characteristics and practices discussed above fit quite well with the Australian LCCs. This is discussed in more detail in the following section.

2.4 Background: Precursor to LCC growth in Australia

Aviation research has identified several factors underpinning the growth of LCCs. Francis et al. (2005) noted demand related features such as increasing income and population. Factors more specific to airlines were the entrepreneurial flair of the business leaders (Tony Ryan and Richard Branson) and the brand of affordable air travel (Francis et al. 2005). Furthermore, strong financial backup, which enables the airline to sustain prolong period of losses, was identified as a key factor especially in situations where the incumbents react with strong competitive pressures by matching prices and increasing capacity (Forsyth 2003). Timing of entry also played an important part for some airlines. For instance, the success of Virgin Blue in the Australian domestic market was partly influenced by the collapse of Ansett. Also, technological advances, such as the internet, helped LCCs to reduce costs associated with distribution (Mason and Alamdari, 2007).

2-14

Franke (2004) argued that the existing inefficiencies of the legacy carriers had given opportunities for LCCs to thrive in the short-haul markets. Specifically, Franke (2004) noted,

“[the] carriers had built their complex operational model around the needs of their least valuable clients (low-yield connecting passengers), whom they forced to connect at hubs in order to maximise the airlines’ overall destination portfolio: a situation paid for by their own premium clients. A crisis soon developed during the second half of 2000 when, faced with an economic downturn, these high-value passengers, showed a growing reluctance to pay premium prices” (p.16)

The LCC model is also more resilient in times of weakening demand than the legacy counterparts (Gillen and Lall 2004). Specifically, the legacy carriers adapt to business cycles by shifting the cost base to the premium markets during the high seasons, while shifting to the lower cost module during the troughs (Gillen and Lall 2004). As put forward by Gillen and Lall, the LCCs were permanently on the ‘lower cost model’. Consequently, LCCs were able to continue growing, even in times when the legacy carriers were suffering from heavy losses. Notable examples are Ryanair and Southwest, which continued to make profits in periods of high volatility and levels of bankruptcies. However, the current global financial crisis will affect all air travel as a result of reductions in discretionary income.

Experiences in the U.S. and Europe show that deregulation (of the aviation industry) was a necessary precursor to the entry, adaptation and evolution of the LCC model (Dobruszkes 2007, Franke 2004, Meyer and Oster 1987). Following the deregulation in the U.S., Meyer and Oster (1987) observed,

”the emergence of the new entrant jets was almost surely the least anticipated major event of deregulation prior to the fact ... The niches served by these carriers were largely markets left vacant because of previous regulatory policies, and in keeping with the identity of these under-attended market niches, most of the new entrant jet carriers attempted to do

2-15

something that predecessor established carriers did not. In most instances they either entered a market that was not previously served or well served or entered a previously well served market while offering substantially lower fares” (p.49-50)

Broadly, deregulation refers to a relaxation of set of rules and intervention governing an industry. This is in order to “make markets more effective conduits between consumers and producers”, which includes various measures to make firms to be more productive and efficient in meeting consumers’ needs and wants (Forsyth 1992:5). To achieve this, microeconomic reform involves a “thorough dismantling of the comprehensive system of government regulation and control” (Kahn unknown year). In addition, there are needs to improve the functioning of closely associated markets such as airlines and airports to improve matters overall (Forsyth 1992).

2.4.1 Deregulation of the airline industry in Australia

In 1946, with the view that the Australian airline industry was a natural monopoly, the Australian government established a wholly stated owned airline, Trans Australia Airlines (TAA) (BTCE 1991). Subsequently, this was transformed into a ‘two-airline policy’ to allow for limited competition, which implicitly had an effect of striking some sort of a balance between the benefits of competition and cost savings from scale economies (Hooper and Findlay, 1998). Australian National Airways (ANA) was the other domestic airline, which was subsequently bought by Ansett Airlines (Hooper and Findlay, 1998). In this period, Qantas was the only designated carrier for international operations; Qantas was not allowed to provide domestic services. Thus, there were TAA and Ansett for domestic services, while international services were exclusive to Qantas. Soon after deregulation, TAA was re-branded as Australian Airlines, which was eventually purchased by Qantas in 1991.

The two-airline policy became under increasing criticism because the public could not see effective competitions in the market; for example, BTCE noted that "both 2-16 airlines were operating the same equipment on the same routes with the same schedules for the same fares" (BTCE, 1991:3). An Independent Review of Economic Regulation of Domestic Aviation (the May Review) in the 1980s reached the conclusion that Australian aviation was characterised by low labour productivity, yet, high and stable profits, with its focus almost exclusively on the business market. The consequences were the under-developed leisure air travel market and the absence of charter alternatives (Dwyer and Forsyth 1992).

In October 1990, the two-airline policy was terminated, and this removed constraints for domestic airlines in the following areas (BTCE 1991):

o Control over aircraft imports; o Capacity allowed and supplied on trunk routes by each airline; o Abolishment of the Independent Air Fares Committee in setting fare levels; o Entry/exit barriers to domestic trunk routes.

The effect of deregulation was immediate with the entry of Australia’s first LCC - Compass airline. Before we introduce the topic on LCC entry in Australia, we briefly introduce two other regulatory reforms that accompanied the deregulation.

2.4.2 Privatisation of the domestic airports

An additional barrier to entry for new entrants was removed when the airport sector was privatised. All major airports were privatised in 1997 and 1998. These airports included all capital city airports (Sydney Kingsford Smith in 2002) and a selection of other airportsi (Kain and Webb 2003). Price caps were removed on aeronautical charges in all capital city airports in 2002 except for Hobart (Kain and Webb 2003). Local council owned airports such as Coffs Harbour and Ballina airports were corporatised, and these airports were expected to generate returns through aeronautical charges, as well as non-aeronautical charges (e.g. parking). Pricing reform also took place in the air traffic control and airspace management services provided by Airservice Australia, which involved moves toward user-

2-17 based and cost reflective pricing strategies (Kain and Webb 2003). Consequently, the airports in the regions now had greater bargaining flexibility with the airlines, for instance, on landing fees and passenger charges, which generated commercial opportunities for LCCs to service regional airports at reduced costs.

2.4.3 Foreign ownership cap

Another policy change that led to the entry of LCCs in Australia was the abolishment of the foreign ownership cap on domestic airlines. Full deregulation of the domestic sector is not yet complete because the seventh freedom is permitted only on a case-by-case basis, while cabotage (the eighth freedom) is still forbidden in Australia. Rather, the government preferred lifting ownership controls to promote competition in the domestic aviation sector. In 2000, the Australian government in order to promote competition amended domestic airline guidelines to allow full foreign ownership of Australian domestic airlines, under the proviso it is not in conflict with national interests (Bureau of Infrastructure 2008). Lifting of the foreign ownership cap, for example, enabled Virgin Blue in 2000 and Tiger Airways in 2007 to obtain rights for Australian domestic services despite the fact that their majority stakes were held by foreign investors.

2-18

2.5 Australian LCCs and their impact on domestic dispersal

2.5.1 First wave of LCCs in Australia (1990 – 1993)

The Australian aviation sector was deregulated on the 30th of October 1990. By December 1990, Compass commenced operation with one-class configured A300s (BTCE 1991). Compass, at one point, had 10% of the total aviation market and up to 21% share of the routes it serviced (BTCE 1991). But the airline experienced problems in gaining access to airport slots and suffered from delays in aircraft delivery (Grimm and Miloy 1993). In addition, Compass’ entry was met with strong capacity increases and fare discounting by incumbent carriers; contributing to Compass’ amounting debt. Compass was subsequently grounded within a year of commencing operation. Former regional carrier, Southern Cross Airlines, adopted the Compass brand and launched Compass Mark II in 1992. Compass II, however, lasted less than a year. In the 'first wave' of LCC entry, Ansett and Qantas with their ‘deep pockets’ were able to sustain losses for a longer period of time than the new entrants (Sinha 2001, Forsyth 2003). LCCs failed to sustain their presence in the market; however, their effect on competition perpetuated as competition intensified between the two incumbents in the period following the first wave (BTCE 1991).

2.5.2 Duopoly period (1994 – 1999)

A duopoly comprising Qantas and Ansett emerged in the domestic aviation sector during this period. Although there was no new LCC entry, two distinct post- deregulatory effects were observed in the period between 1993 and 1996 in Australia. First, revenue passenger numbers continued to increase as shown in Figure 2.2. (although it flattened between 1996 and 1999). The second effect was related to airfares. As expected, the average fare levels have decreased following liberalisation (see Figure 2.3). This is a widely documented fact in the aviation 2-19 research literature worldwide (for instance, Williams 2002, Button and Stough 2000, Borenstein and Rose 1994, Meyer and Oster 1987, Winston and Morrison 1986). A notable effect of deregulation on price is the widening of price dispersion, which is also consistent with the post-deregulation effects observed in the U.S. and Europe (Williams 2002ii, Borenstein and Rose 1994iii). As shown in Figure 2.3, disparity is evident in domestic prices throughout 1992 – 2008. However, the period between 1996 and 1999 was characterized by a flat demand, although there were strong fluctuations in the levels of fare discounting. This began to change from the late 1990s when two LCCs made their way into the domestic market.

Figure 2.2. Revenue Passenger Demand (source: Bureau of Transport and Regional Economics, Aviation Statistics 2007)

2-20

Figure 2.3 Domestic Airfare Indices (source: Bureau of Transport and Regional Economics, Aviation Statistics 2008)

2.5.3 Second wave of LCCs (2000 – 2006)

Impulse and Virgin Blue’s entry marked the 'second wave' of LCC entry. Impulse was originally a regional carrier, which expanded its operation to domestic trunk routes, entering into direct competition with Ansett and Qantas. Impulse, similar to the predecessors, did not succeed against the incumbents, and was absorbed by the Qantas group in 2001.

Two important features in the second wave contributed towards Virgin Blue’s success. First, Ansett collapsed in September 2001 leaving a very large capacity gap in Australia. Second, Virgin Blue was in a much better financial position than its predecessors (Compass I and II and Impulse) as part of the international conglomerate, the Virgin Group (Forsyth 2003). Virgin Blue grew to gain over 35% of the domestic market share by 2007 (CAPA, 2007).

2-21

Figure 2.2 shows that the demand for air travel continued to increase beyond the pre-Ansett collapse level following the launch of Jetstar in 2004. By 2007, Jetstar, which was a fully owned subsidiary of Qantas, had 12% of the domestic market (CAPA 2007). Both Virgin Blue and Jetstar commenced services as LCCs, resembling the Southwest airlines, with uniform fleet and direct shuttle flights. Both airlines also adopted the features of Ryanair (at the time of the start-up) by not offering ancillary features such as frequent flyer programs. But their models began to evolve - this is discussed in Section 2.5.4.

From a regional tourism point of view, Virgin Blue and Jetstar sought to link excess capacity airports in the regions with the major cities. Table 2.3 lists the airports with LCC services showing that many regional airports have gained air traffic, in some cases, by multiple-fold in a period of only a few years. Ballina- Byron and Launceston are some of the examples. The effect of LCCs on trip generation is illustrated in Figure 2.4. The generative effect is shown by the ‘wedge’, beginning in 2001/2002, between ‘total domestic passenger demand’ and ‘total domestic passenger demand to capital cities (incl. Gold Coast)’. Figure 2.4. shows a strong growth in domestic air travel demand to regional destinations following the entry of Virgin Blue and that the upward trend continued following the entry of Jetstar in 2004.

2-22

Table 2.3 Top 40 Australian domestic airports in terms of incoming passenger flows

Airport Pax. 2000/01 Pax. 2005/06 Pax. Growth % change Sydney 7,609,862 8,795,031 1,185,169 16 Capital city Melbourne 6,146,495 8,077,308 1,930,813 31 Capital city Brisbane 4,524,200 5,833,024 1,308,824 29 Capital city Adelaide 1,900,557 2,488,121 587,564 31 Capital city Perth 1,629,751 2,327,417 697,666 43 Capital city Gold Coast 898,896 1,653,793 754,897 84 LCC service Cairns 962,124 1,281,078 318,954 33 LCC service Canberra 640,915 1,008,934 368,019 57 Capital city Hobart 276,937 799,558 522,621 Capital city Darwin 418,401 506,208 87,807 21 Capital city Townsville 283,065 471,483 188,418 67 LCC service Launceston 1,280 451,927 450,647 >300 LCC service Maroochydore 69,466 391,690 322,224 >300 LCC service Williamtown 25,666 341,602 315,936 >300 LCC service Alice Springs 350,293 294,439 -55,854 -16 Mackay 94,235 286,314 192,079 204 LCC service Hamilton Island 140,608 199,591 58,983 42 LCC service Uluru 218,415 189,648 -28,767 -13 Rockhampton 78,736 177,552 98,816 126 LCC service Karratha 83,838 120,394 36,556 44 Broome 108,530 118,613 10,083 9 LCC service Prosperpine 21,110 110,573 89,463 >300 LCC service Ballina 76 103,566 103,490 >300 LCC service Coffs Harbour 84 92,845 92,761 >300 LCC service Kalgoorlie 98,068 86,433 -11,635 -12 Hervey Bay 0 55,493 55,493 LCC service Port Hedland 40,827 54,970 14,143 35 Newman 18,868 50,510 31,642 168 Mount Isa 33,664 45,649 11,985 36 LCC service Gove 82,285 45,127 -37,158 -45

(Source: compiled from Bureau of Transport and Regional Economics, Aviation Statistics 2007. Note: [*] shows airports with LCC services as at March 2007)

2-23

Figure 2.4. Domestic Revenue Passenger Growths from 1992/1993 (source: compiled from Bureau of Transport and Regional Economics, Aviation Statistics 2007. note: 1992/93 - 2006/2007 (1992/93 = 0): Capital cities (incl. Gold Coast) vs. All other)

The second wave resulted in the greater domestic dispersal of national visitors. The National Visitor Survey data provides insights into the travel characteristics of domestic dispersal. As shown by Figure 2.5, ‘holiday and leisure’ and ‘visiting friends and relatives (VFR)’ have increased in shares at the expense of ‘business’. The ‘business’ share decreased nine percentage points between 1999 and 2008.

2-24

Figure 2.5 Overnight trips made by air by purpose (source: National Visitor Surey, Tourism Research Australia 2008)

2.5.4 The ‘third’ wave (post-2006)

In late 2007, a Singapore airline backed subsidiary, Tiger airways, made entry in the domestic market. The airline established its base in Melbourne, with low-cost services to Perth and Darwin. The airline’s direct impact on total domestic capacity was marginal because they had only five A320s to deploy in Australia. However, the airline had an impact on the incumbent low-cost carrier, Jetstar, by forcing it to pursue the same routes as Tiger, as well as commencing services to/from Melbourne airport (which Jetstar previously avoided, preferring Melbourne’s secondary airport, Avalon).

Another important change in the Australian airline sector that signified a third wave was the evolution of two incumbent LCCs. Virgin Blue increasingly focussed on becoming a network carrier similar to Qantas. When examined

2-25 against the characteristics of LCCs discussed previously, Virgin Blue’s move towards the NC model is evident because of its expansion into (1) smaller regional markets with lower demand density (with medium size Embraer aircraft); (2) increasing use of hub-spoke strategy (e.g. Cairns – Sydney – Ballina as oppose to Cairns – Ballina direct); (3) introduction of business lounges and premium seating class; (4) code-sharing or/and interlining arrangement with domestic and international airlines (e.g. Delta Airlines, Regional Express, Malaysia Airlines), and (5) multi-fleet, including long-haul aircraft (e.g. B777-ERs for Sydney – U.S., Embraer 117 and 119s for Sydney - Tamworth). This period also marks a new phase in that the focus of the incumbent LCCs became increasingly towards long- haul international markets. Jetstar International was established in 2007 and commenced services to Thailand, Hawaii and Japan as a long-haul low-cost service provider. However, domestic services remain as the most important source of business for all incumbent carriers because they represent the majority of total passenger demand for Australian carriers.

At this point it is worthwhile to note that whether or not Virgin Blue is a LCC is of little consequence to this thesis. This is because most key changes in Virgin Blue’s strategy did not take effect until the final quarter of 2007 – and we will be using secondary data sources from the years 2006 and 2007. Furthermore, given that it is the LCCs’ effect on regional destinations that is of primary interest, the exact location in which an airline is placed along an airline business spectrum is of less importance. Rather, it could be said that the wider focus of this thesis is affordable air travel, which has been instigated by the LCCs and the competition that the LCC entry has introduced to the domestic aviation market since 2000.

2-26

2.6 Summary

This Chapter introduced two central concepts of this thesis - dispersal and LCCs. Background information was provided on LCCs’ main characteristics, as well as information on the precursor of LCC growth in Australia. Against this background, the Chapter introduced the three Australian LCCs – Virgin Blue, Jetstar and Tiger – followed by an outline of the changing nature of these LCCs. The Chapter concluded that the LCCs’ characteristics have had the combined effect of generating air leisure travel demand for regional tourism destinations. Thus, the first specific aim of this thesis has been addressed, which was to ‘provide an interpretative survey of relevant literature and secondary data sources to understand the link between LCCs and domestic dispersal of tourists’. What remains is the question on the effects of LCCs on regional dispersal, which will be the primary focus of the remainder of this thesis. The following Chapter explicates the relationships between LCCs and regional dispersal.

2-27

i The full list of airports leased in this period were: Melbourne, Brisbane, Perth, Canberra, Adelaide, Darwin, Alice Springs, Coolangatta, Hobart, Launceston, Townsville, Mount Isa, Tennant Creek, Archerfield, Jandakot, Moorabbin and Parafield. ii Transportation Research Board (1999) shows that in the U.S., the highest 5% of fare payers’ contribution to airline revenue has increased from 8% to 18%, while the lowest 25% of fare payers’ contribution decreased from 14% to 10% (as cited in Williams 2002). iii Borenstein and Rose (1994) used a Gini Index to analyse the price dispersion and obtained an ‘expected absolute fare difference’ of 36% for a given air service. Importantly, they concluded that although the absolute values vary extensively across routes, the differences in fares paid were prominent across passengers than across carriers.

2-28

3. REGIONAL DISPERSAL PROPENSITY AND LOW-COST CARRIERS

3.1 Introduction

The aim of this chapter is to introduce the relevant literature on the effect of LCCs on regional dispersal. In doing so, this chapter accomplishes the second specific aim of this thesis (A2), which is to ‘identify and explicate the relationships between regional dispersal and LCCs’. This chapter has two sub-aims. The first is to provide an overview of the determinants of dispersal. The second is to identify and explain the relationships between these determinants and LCCs. This research draws from the literatures on tourists’ spatial behaviour, particularly multi- destination travels, because ‘dispersal’ is a special case of tourists’ spatial behaviour. The scope of the literature review extends as far as spatial behaviour research is relevant for analysing the relationships between LCCs and regional dispersal. Section 3.2 outlines the framework applied, while Section 3.3 explains the key determinants in the context of LCCs and regional dispersal.

3-1 3.2 Spatial patterns of tourists’ regional dispersal

This section explains the general patterns of tourists’ spatial behaviour identified in the tourism literature. Due to the spatial nature of the topic, spatial behaviour received much attention from geographers. Pearce (1979), on the subject of tourism geography wrote,

“tourism has been variously defined but may be thought of as the relationships and phenomena arising out of the journeys and temporary stays of people travelling primarily for leisure or recreational purpose … the geography of tourism is concerned essentially, though not exclusively, with the spatial expression of these relationships and phenomena” (p.248)

Dispersal is one such spatial manifestation arising from leisure travels. Inherent in dispersal, therefore, is the spatial expression of tourists’ behaviour. A distinction can be made between ‘behavour in space’ and ‘spatial behaviour’ (Walmsley and Lewis 1994). On this distinction, Walmsley (2004) has shown that the analysis of the former,

“involves description of the context in which the behaviour in question occurs and the relating of behaviour to that specific context … the study of ‘spatial behaviour’ focuses on trying to find the general in the particular in the sense of distilling the rules, principles, and laws that describe behaviour independently of the context in which it occurs. In other words, with “spatial behaviour”, the search is for general principles of people-environment interaction and for understanding of how humans as a whole behave in certain types of settings (e.g. shopping centres, theme parks) rather than with particular contexts (e.g. Oxford Street, London, Disneyland)” (p.50)

This thesis aims to find the general relationships between LCCs (or equivalent air transport services) and regional dispersal. Specifically, dispersal is achieved when many destinations are visited within the same trip, or when a unique trip is

3-2 undertaken on many parts of the destination (Wu and Carson 2008). Although the previous chapter introduced a more simplified definition of regional dispersal, as ‘an overnight trip in the periphery’, it is apparent from Wu and Carson (2008) that dispersal can be achieved by a multi-destination or a single-destination trip to the periphery.

In general, dispersal reflects tourists’ motivation to visit the periphery. Cooper (1981), who was one of the first to have studied the linkages between ‘spatial and temporal patterns of tourists’ and tourist characteristics, noted that a general spatial pattern involves a movement outward from a touring centre, and towards locations with declining tourism facilities. He concluded that the “wave-like pulse of visits outward from a touring centre and down the hierarchy” (p.369) is probably a general phenomenon that can be observed in a variety of places and locations.

Fennell (1996) argued that recognising what is ‘core’ and what is ‘periphery’ depends on tourists’ “inherent activity-based motivations” (p.816). Thus, the ‘core’ for a traveller will depend on the subjective interests and activities sought by the traveller. In Fennell’s study, the ‘special interest’ groups were more specific in their activities; consequently, the special interest groups had a diffused pattern of travel and stayed in the outskirts of the UK’s Shetland region. The majority of the ‘general interest’ group had a high representation in Lerwick, which is the main township of the region. Thus, dispersal reflects tourists’ motivation that tends be more ‘special interests’ than ‘general interests’.

Tourism researchers have identified a number of specific trip itinerary patterns. These trip structures were found to be robust across spatial scales (e.g. inter- continental scale to local scale) and countries. The Mings and McHugh (1992) study was one of the early studies that has identified such patterns. They discovered that the majority of the variation in U.S. domestic trips to Yellowstone Park can be categorised into one of four trip structures: direct route; partial orbit; full orbit, and fly-drive. Lue, Crompton and Fesenmaier (1993) imposed a

3-3 structure to these itineraries. The trip itineraries were structured into five basic patterns of single and multi-destination trips, which extended the four identified by Mings and McHugh (1992). Opperman (1995) further extended this into two single-destination and five multi-destination trips to account for trip patterns common in international travels.

The framework developed by Lue et al. (1993) has been applied to a variety of situations and contexts, and has formed the basis for further studies on multi- destination travel itinerary; for instance, on domestic travel in the U.S. (Stewart and Vogt 1997); domestic travel by international tourists in Queensland, Australia (Tideswell and Faulkner 1999); trip patterns in New South Wales, Australia (Parolin 2001); and South Australia (Wu and Carson 2008); international tourists to Malaysia (Opperman 1995), and the role of Hong Kong in tourists’ international travel itinerary (Lew and McKercher 2002). In Figure 3.1, the five patterns proposed by Lue et al. (1993) were integrated into three patterns of single-destination (SDT), multi-destination type 1 (MD1) and multi-destination type 2 (MD2). Each pattern is discussed below.

3-4

MD1 Regional tour/partial orbit b

a c D

SDT

Single destination/Base camp MD2

b D a En route b Trip chaining D a c

d c

HOME

Figure 3.1 Spatial representation of tourists' travel patterns (modified from Lue et.al.1993)

The single-destination trip (SDT) is equivalent to the direct-route pattern in Mings and McHugh (1992). While Lue et al. (1993) defined the second pattern – ‘base camp’ or ‘BC’ - as a multi-destination trip, Opperman (1995) suggested that this is an extension of the single-destination trip because it involves an overnight stay in a single destination with radiant day-trips to the periphery. Here, BC too will be defined as a SDT. Revisiting the definition introduced in Chapter 2, this type of trip can be both a domestic and regional dispersal. A SDT will constitute a domestic dispersal if the destination of stay is only in the ‘gateway’, while regional dispersal if the chosen destination is in the ‘periphery’.

The second group of travel patterns includes partial orbit and fly-and-drive trips, or in Lue et al. (1993) terms, the ‘regional tour’ pattern. This pattern is of particular relevance to ‘LCC and dispersal’ because it is a representative form of regional dispersal that can be achieved with air travel. Many regional dispersal

3-5 trips are this type in Australia; originating from the large metropolises destined for the ‘sun, sand and sea’ destinations along the Eastern Coast. These patterns are collectively denoted ‘MD1’ (multi-destination 1). Note that when an extra trip links up ‘c’ with ‘D’, this becomes a ‘full-orbit’ pattern.

MD1 highlights the fact that regional dispersal of air arrivals depends on the determinants of travel from ‘D’ to {a, b, c}. These determinants are discussed in greater detail in section 3.3. Moscardo et al. (2004) have shown that transport mode chosen by travellers ‘to’ and ‘within’ destinations constrains the travellers’ travel pattern. Furthermore, they have shown that the ‘access points’ for transport, e.g. location of the airport in relation to the wider destination region, affects the spatial pattern of the trip. Consequently, as illustrated by Moscardo et al. (2004) in their Great Barrier Reef (GBR) case study, a shift in the destination access mode from air towards a car is accompanied by a change in the trip and traveller characteristics to the region.

More recently, the GBR region was subject to several LCC entries on a number of locations (e.g. Hamilton Island, Mackay, Townsville, Cairns and Rockhampton – see Table 2.3 for traffic growth in these airports), rendering air transport as an increasingly significant source of leisure arrivals. A related problem is the question over the type of ground transport mode a destination should promote to achieve the greater regional dispersal of the air arrivals. This issue is a significant one for government policy because governments may have other policy objectives that do not necessarily promote the travel mode best for regional dispersal. This issue will be revisited in Chapter 5.

The final type, MD2 (multi-destination 2), includes ‘trip-chaining’ and ‘en route’ patterns. ‘En route’ occurs when a trip stopovers in ‘a’ or/and ‘b’ on the way to ‘D’ (a trip that takes the route stopping-over at ‘c’ or/and ‘d’ is also en route). ‘Trip-chaining’ occurs when a trip includes destinations {a … d} with different access and return route. A way these patterns distinguish themselves from MD1 is through the main mode of travel used by tourists. Air travel does not offer the

3-6 spontaneity and flexibility of that offered by cars (e.g. Stewart and Vogt 1997:458). Hence, in the regional tourism context, MD2 is difficult to achieve with air travel but easily achieved with cars.

In MD2, the peripheral destinations (e.g. ‘a’ and ‘b’) impacted by the LCCs are en route. This is because these peripheral destinations usually do not command large enough demand to sustain their own LCC services from “Home”. Consequently, these peripheral destinations are bypassed by the LCCs. This will be a problem if there is substantial substitution effect from ground travel modes toward air on the travel corridor. A significant substitution effect will adversely impact regional dispersal because the substitution effect will generate a bypass effect. The modal substitution issue is addressed in greater detail in Chapter 6.

Finally, the preceding discussions on MD1 and MD2 demonstrate that different regional destinations will be subject to different channels of LCC impact, i.e. via modal substitution (transport ‘to’ the destination) or modal complementarity (transport ‘within’ the destination). The trip patterns, individually or in combination of one another, enable a depiction of a large number of trip variations into a parsimonious set of trips. The trip patterns are capable of highlighting specific LCC and dispersal issues. Thus, whether or not a particular issue applies to a regional destination depends on which trip structure (SDT, MD1 or MD2) characterises the travellers, and the destination’s position with respect to the travellers’ overall trip structure. For instance, MD1 is most applicable to trips that originate from Sydney or Melbourne destined along the Queensland’s Eastern Coast, whereas MD2 is most applicable to intra-state trips often shorter in distance (less than 800km).

The review of multi-destination trip factors is a useful starting point for building cause-effect structures of regional dispersal. This is because the force that increases the incidences of multi-destination travels also increases the travellers’ propensity to visit the periphery. The remainder of Chapter 3 focuses on (1) identifying the determinants of regional dispersal and (2) generating propositions

3-7 on how these determinants are affected by the LCCs. As outlined in Chapter 1, this chapter discusses the intra-modal source of difference. Chapter 4 empirically examines these propositions.

3.3 The effects of LCCs on regional dispersal

It was argued that the multi-destination travel literature is a useful starting-point for identifying the dispersal determinants. The Lue et al. (1993) study was one of the earliest to provide a framework on the topic of multi-destination travel. They suggest five main factors. First, heterogeneity of preferences in a travel group can be more easily satisfied with visitations to a greater number of destinations. This is related to another reason given, which is the need for variety by tourists, triggering the need to visit more than one destination, especially if the marginal cost of doing so is relatively low.

There are two other factors. One is related to reduction of risk and uncertainty, and the other, travel monetary costs. Diversification reduces the risk associated with relying on a single destination to provide all the expected utility during the trip. As for travel costs, costs such as long-distance transportation are fixed costs incurred regardless of other attributes of the trip (e.g. length of stay); thus, visiting multi-destinations by combining several individual trips into one, is a way to realise cost savings. Finally, Lue et al. (1993) argued that visiting friends and relatives (VFR) travel purpose is likely to increase the number of stopovers and destinations visited.

Tideswell and Faulkner (1999: 365) added another five factors that can act to stimulate or constrain multi-destination travels. Specifically, Tideswell and Faulkner, based on a review of earlier work by Opperman (1994), Debbage (1991)

3-8 and Lue et al. (1993), added the following factors: package tour or free- independent travel; primary mode of transport used; travel time constraint; repeat visit or not, and the ‘spatial configuration of destinations’. These factors are summarised in Table 3.1. The remainder of this chapter addresses each factor in turn. Emphasis is placed on the relationship of these factors with the LCCs, and where appropriate, the relationship with affordable air travel generally.

3-9 Table 3.1 Summary of the relationships discussed in section 3.3

Factors Effects on regional dispersal LCC demand characteristics from propensity dispersal viewpoint 1. Spatial Different tourism regions will be Different tourism regions will be configuration of associated with different levels of associated with different levels of destinations dispersal dispersal

2. Length of stay Length of stay is positively related to LCC demand will be less sensitive to length dispersal of stay than NC demand

6. First time or repeat First visitation can have a positive or LCC stimulates first-time visitors to visitation negative effect on dispersal; repeat destinations, which may increase or visitation has a positive effect on decrease dispersal. Second-home dispersal travellers are expected to be an important source of dispersal of the LCC arrivals

7. Package tourism Package tourism is negatively Disproportionately large share of LCC related to dispersal arrivals are FIT tourists; therefore, they are less constrained spatially.

8. Transport 'to' and Addressed in Chapter 5 and Addressed in Chapter 5 and Chapter 6 'within' the Chapter 6 destination

3-10

3.3.1 Spatial configuration of the destinations

In the context of tourism destinations, “Wall (1997) is credited with emphasizing the importance of spatial configuration as an attraction attribute” (Weaver, 2006: 93). The spatial distribution patterns of destinations will result in similar patterns of tourist activities. For instance, a ‘node’ will draw a concentration of activities, whereas a linear pattern of attractions will yield linear movement of tourists (Weaver 2006). Tideswell and Faulkner (1999) summarised the influence of spatial configuration on multi-destination travel. The proposition was that “the existence of a range of complementary tourist attractions/destinations within “reasonable proximity” of a region increases the number of stopovers made by tourists” (p.369). Tideswell and Faulkner, however, did not empirically examine the influence of this effect. Hwang and Fesenmaier (2003), in their study of the domestic trips in the U.S., found that the spatial patterns of travel differed widely between and within the Midwest states, concluding that geographic characteristics do influence the spatial behaviour of tourists. Generally, the presence of a variety of activities at a destination causes a spatially concentrated pattern of travel; for example, trips concentrate towards urban areas and gateways for this reason. On the other hand, scattered attractions and destinations cause a spatially expansive behaviour.

The influence of LCC on the relationship between spatial configuration and dispersal is illustrated in the research literature. For instance, Papatheodorou (2002) has documented that the proliferation of affordable air services through charter and low-cost carriers had resulted in the ‘anarchic urbanisation and congestions’ in some tourism centres in the Mediterranean region. Contrasting scenarios also prevail. Francis et al. (2004) illustrated a case where the entry of LCC has resulted in a greater use of the regional airport, but tourists passing through the airport did not travel to the destination that the airport was originally purported to serve. Rather, tourists used the airport as a point of entry and exit to travel elsewhere. The differences in the LCC arrivals’ trip patterns in the two examples given above can be partly attributed to the differences in the spatial

3-11 configuration of destinations and attractions, and the proximity of the airport to these destinations.

3.3.2 Length of stay

Length of stay and spatial behaviour are related. A positive relationship may be expected from the fact that leisure trips are time-constrained, and this renders tourists’ activity patterns highly time-sensitive (Landau, Prashker, and Hirsh 1981 as cited by Debbage 1991). Fennell (1996), in his account of tourists’ behaviour over space and time, added that “when time is short, space is conserved” (p.814). Mansfeld (1991), on the other hand, noted how the effect of length of stay on spatial behaviour may not always be the same, because time constraint may induce a tourist to see as much as possible.

Evidence in the research literature has shown a relationship between airline business models and travellers’ length of stay. Early evidence comes from the study by Pearce (1987) on the spatial pattern of package tours in Spanish destinations. Since the 1970s, charter carriers, as part of their inclusive tour packages, required a fixed duration trip typically for a week or two. However, this changed quickly with the emergence of LCCs on the traditional charter routes in Europe (Williams 2002). This is substantiated by the Alegre and Pou (2006) study of the microeconomic determinants of length of stay; the study shows that the length of stay in the Balaeric Islands between 1989 and 2003 declined by 25%. Although Alegre and Pou (2006) did not explicitly address the emergence of low cost carriers and its potential association with the changes in the length of stay, they did note the shift towards greater flexibility in the length of stay from the traditional ‘bimodal’ distribution (either 1 week or 2 weeks stay) to the ‘four to five day stays’. It is probable, as observed by Pearce (1987), that the previous pattern of one or two weeks stay was an outcome of the package/ticket conditions of the charter carriers.

3-12 In recent years, researchers identified LCCs with short and frequent breaks; for instance, Mason (2005) cited a study by Mintel that found UK residents were 70% more likely to take a short break in 2004 compared to 1999, and argued that the LCCs, to an extent, fuelled this trend. Graham (2006), in exploring the various sources of LCC demand, noted that the short but more frequent travelling to the second homes by the affluent population in UK was an emerging trend associated with the LCCs. She argued that the ‘cash-rich’ and ‘time-poor’ society is conducive to short and frequent trips. This trend is not endemic to UK. The trend towards short and more frequent break is also recognised in Australia (TTF 2003). However, the Australian evidence on LCCs and ‘trips to second homes’ is weak with NVS showing less than 1% of VFR travellers staying in their ‘own property’ during the trip. Generally, the literature concurs with the view that the additional choice brought by LCCs in the selection of trip duration and times, “brought out travel behaviour patterns that were suppressed by the inflexible travel packages that were previously available” (Mason 2005:24).

LCC travellers may be constrained in their time-budget for two other reasons. First, air transport is chosen by those with high opportunity cost of time, compared to other modes of travel. Second, LCCs stimulate disproportionately more of the time-poor travellers, as well as the affluent travellers - particularly with second homes - travelling in greater frequency. Njegovan (2006)’s econometric study of UK residents found that low airfares trigger a substitution from domestic leisure/durable goods (in UK) towards short overseas travel, providing some evidence on the relationship between low airfares and short- breaks. While a trade-off between lower airfares and travel frequency is likely, the lower airfare is unlikely to induce an increase in the length of stay if travellers are time-constrained.

3.3.3 Variety and multiple-benefit seeking behaviour

Lue et al. (1993) argued that a tourist might seek a variety of activities from a single place (such as a gateway) or obtain multiple benefits from multiple places.

3-13 In both cases, the effect on dispersal will be similar because seeking greater variety of activities will generally increase the need to explore the destination region, including the periphery. Given the fact that land-uses become more homogenous down the urban-to-rural hierarchy (Johnston et al. 2000), seeking multiple benefits and activities will increase the need to be spatially expansive in the rural-regions.

Tideswell and Faulkner (1999) argued that the ‘number of different travel purpose stated’ is a good proxy to test for the variety effect. They also accounted for the proposition put forward by Lue et al. (1993) that ‘visiting friends and relatives’ (VFR), as a travel purpose, increases the likelihood of a multi-destination trip. The effect of VFR trips may increase the dispersal propensity because residential areas tend to be located in suburbs, which is often located beyond the functional and recreational centres. It is briefly noted here that VFR as a source of dispersal is less desirable from the expenditure viewpoint because it injects less expenditure into the region’s economy. For instance, NVS (2007) shows that holiday travellers contribute $637 per visitor, or $144 per visitor night, whereas the figures for VFR were $283 and $81 respectively. Thus, even if this type of trip constitutes dispersal, the economic contribution is much less than what the visitor dispersal volume may suggests.

As a result of LCC proliferation, in the short-haul travel market, leisure travellers are increasing in share of total passenger mix relative to business (Dresner, 2006). The significance of the LCC clientele depends on the level of the variety of travellers using LCC services. For instance, Figure 2.5 has shown that LCCs were instrumental in the stimulation of both VFR and holiday flows. As for business travel, Mason (2001) found that business travellers, in particular the small and medium businesses, are important patrons of LCC services. Similar findings appeared in the study by Fourie and Lubbe (2006), who found that LCCs and NCs compete for the business markets in South Africa.

3-14 Overall, it may be argued that a greater range of travel purpose (business, VFR and holiday) characterises the LCC demand than other airline business models, although some variation is expected across routes; for instance, the Sydney- Hamilton Island route which was serviced by Qantas was recently replaced by Jetstar entirely. In such a case, the traveller can only fly with LCCs, hence, the proposed differences between LCCs and NCs cannot be applied. In the aggregate, the greater variety in the reasons for travel and the larger share of VFR related travels, have positive effects on dispersal propensity. Hence, it is proposed that variety-seeking behaviour is a significant source of dispersal for LCC passengers.

3.3.4 Risk and uncertainty reduction: distance travelled

Time constraints, and the desire to enjoy the trip, will lead tourists to visit a large site as a way to reduce uncertainty (Cooper 1981). Thus, one can propose a negative relationship between regional dispersal and levels of uncertainty. In contrast, when tourists diversify their ‘destination portfolio’ in order to diversify risk, multi-destination travels can be positively related to the level of uncertainty. Tideswell and Faulkner (1999), assuming that greater distance is associated with greater uncertainty about the destination, found a positive relationship between distance and the number of multi-destination stopovers. Debbage (1991) argued that greater distance implies greater time and monetary costs; thus, travel to farther destinations increases the tourists’ propensity to ‘see more and do more’. Hwang and Fesenmaier (2003) have shown that 80% of single-destination U.S. domestic travellers travelled round-trip distance of 340 miles or less, while the equivalent figure for multi-destination travellers was 760 miles. In summary, while multi-destination trips are positively affected by uncertainty, it is difficult to ascertain the direction of the relationship between uncertainty and regional dispersal.

Graham A. (2006) argued that increases in the flying propensity of the population is an important source of demand for the LCCs. As discussed previously, Mason (2005) further noted how the ’new’ demand for air travel shows signs of

3-15 destination neutrality. A combined effect of these developments can be summarised as having created a demand that may be more sensitive to risk and uncertainty, because they are likely to be first time visitors (although not necessarily first time flyers) to a particular destination. Thus, LCC proliferation may magnify the effect of uncertainty on dispersal because LCC tourists may be more sensitive to uncertainty. It is possible for LCCs to diminish the effect of uncertainty on dispersal because LCCs lower travel costs, reducing the need to ‘see and do more’; decrease in the travel costs reduces the dispersal propensity.

Distance must be used with caution when it is used as a proxy to travel cost because travel distance does not affect the travel cost as much as other factors - such as the level of competition, the market structure and the regulatory regime on a given route. All Australian domestic routes are deregulated; thus, airlines are free to charge whatever they wish, as long as it does not violate the general competition rules. While this was not a problem for the Tideswell and Faulkner (1999) study because they examined the behaviour of international tourists, the same cannot be said for domestic visitors.

3.3.5 Heterogeneity in preferences (Travel party)

Heterogeneity in the preferences of a travel party increases in the number of people in a travel group, which increases the multi-destination travel propensity (Tideswell and Faulkner 1999). The extent of the heterogeneity also depends on the nature of the travel group; for instance, travelling with ‘family and relatives with children’ may differ from ‘adult couple without children’ for reasons other than the travel party size. Similar to the discussion on the dispersal impact of ‘uncertainty and risk reduction’, the heterogeneity could increase or decrease the dispersal propensity. Figure 3.2 shows that travel party size on air travel increased between 1999 and 2007, and by assumption, the party heterogeneity also increased. Lower airfares tend to promote travels of greater party size for the following but not exhaustive reasons. First, leisure travellers are more likely to be travelling with ‘others’ than business travellers. Even when business associates

3-16 travel together, their heterogeneity in preferences will have little effect because of their restricted freedom over destination choices and length of stay. Second, lower airfares enable larger travel groups such as family and ‘relatives and friends’ to choose air travel. Thus, the share of ‘couples’, ‘family and relatives’ and ‘friends’ increased since the second wave in Australia.

The assumption that preference heterogeneity can be proxied by travel party size requires further empirical investigation. Travel party heterogeneity can be a positive or negative influence on multi-destination and dispersal propensity. The former is possible if the preference heterogeneity requires the party to diffuse in search of greater variety of activities, while the latter is likely if the party, for reasons such as logistics and organisational limitations, is constrained by the large travel party size. By the same token, visitors may concentrate in main tourism centres where large variety of activities can be found to satisfy the variety in the preferences of the travelling group. However, no clear proposition on the differential effect of travel party size (on dispersal) across airlines can be ascertained.

3-17

Figure 3.2 Travel party characteristics of air travellers (source: compiled from National Visitor Survey 1999 and 2007)

3.3.6 Trip arrangement (package tourism)

Package traveller’s behaviour can be spatially confined because of the predetermined routes and places of visits; or it can stimulate a touring type into regions that otherwise will not be exposed to tourists (Tideswell and Faulkner 1999). An important example of a relationship between package tourism and air transport is the inclusive tour charter (ITC) packages developed in Europe 40 years ago. The extent of the ITC was such that the charter revenue passenger kilometres (RPK) surpassed that of scheduled RPK in the 1970s (McDonnell Douglas 1977 as cited by Pearce 1987). Similar to the LCCs, charter carriers and the ITC packages contributed to the domestic dispersal of tourists. Pearce (1987) concluded that the charter package tourism in Europe was characterised by an insular and ‘spatially selective’, ‘pleasure periphery’ in Southern Europe. While Pearce provided an interpretative survey of the spatial patterns of package and charter tourism, the level of analysis did not extend to that of the spatial behaviour in the destination and the surrounds, i.e. regional dispersal.

3-18

As shown in Chapter 2, the LCC model shares similar features with the charter carriers in its various strategies to reduce unit costs; for instance, LCCs target leisure destinations where low-margins but high-volume markets can be sustained, and by directly accessing the smaller regional airports, LCCs can avoid the delays and congestions prevalent in large airports. A major difference between LCCs and charter carriers (other than that LCC is a scheduled service) is that while charter carriers were often owned and operated by vertically integrated tourism firms (Williams 2002), the LCC model derived its cost reduction from simplicity in fares and ‘unbundling’ of air services (CAPA 2006). Thus, the LCC model is an accomplice to ‘free and independent’ (FIT) travellers. With the direct booking and independent pricing of each leg, LCCs effectively passed on the responsibility of ‘packaging’ a holiday to the tourist.

The proposition put forward by Tideswell and Faulkner (1999) suggests that FIT demand may or may not contribute to dispersal propensity. However, this proposition was for international tourists visiting Australia, for whom the role of package tours in introducing destinations and experiences otherwise difficult to gain, is substantial. The same reasoning does not apply to domestic travellers because domestic tourists do not face the same barriers in regards to language, the level of uncertainty, etc. Thus, it is proposed that package tourism is negatively related to dispersal.

3.3.7 First timers, repeaters, and destination familiarity

Cooper (1981) argued that the centralising force of tourists in large sites is prevalent in the early stages of their visits. From a tourist’ motivation perspective, however, the literature suggests an opposite effect to that proposed by Cooper. Opperman (1997) summarised Gitelson and Crompton’s (1984) study of the differences in the first and repeat visitors, noting that first time visitors are younger, have greater motivations and purpose for variety and new experiences,

3-19 and they are relatively distant from travel motivations such as VFR or ‘seeking relaxations’.

Opperman (1997) provides some evidence that first time visitors contribute more to dispersal than repeat visitors. Based on the analysis of international tourists in New Zealand (NZ), Opperman found that first time visitors were more active and explorative, indicative by the fact that they visited more sites during their stay than repeaters. For instance, first time visitors to NZ listed an average of 6.4 activities or attractions compared with 3.6 destinations by repeat visitors. The results also implied that first time visitors, while representing a greater share in the primary destinations, also visited an average of 5.9 destinations compared to 3.6 by repeat visitors. The results have shown that first time visitors had greater relative share in 95 of the 110 destinations surveyed in NZ, which suggests that first time visitors are also important contributors to dispersal.

Recently, Li et al. (2008) provided an overview of the research on first and repeat visitors, concluding that first time visitors may be driven by novelty more than by familiarity (Li et al. 2008: 278). They noted that relaxation and familiarity are the most important reasons for repeat visitors, while gaining new experiences is the motivation for first time visitors. They found that first time visitors were more travel and tourism oriented in comparison to repeat visitors who were more interested in the pursuit of specific activities. Building upon Fennell’s (1996) argument introduced earlier, repeat visitors’ tendency to pursue specific activities implies their greater dispersal propensity. Although Li et al. (2008) did not allude to spatial behaviour directly, they noted that first time visitors are found to be more extensive in their destination exploration, while repeat visitors were more intensive in their use of time across smaller range of destinations.

Going back to the risk and uncertainty reduction perspective, Hwang et al. (2006) wrote, “the more familiar the tourist is with the location, the more knowledge one has of different kinds of local activities and attractions to fill an entire trip schedule” (p.1060). Thus, tourists who are familiar with the destination are able to

3-20 engage in time-consuming activities with less need to diversify their risks across several destinations. In summary, a consensus is yet to be achieved in the literature on the effect of first-time visitation on the dispersal propensity. Repeat- visitors, with their specific activity focus, are an important source of regional dispersal, but it is just as likely that first-time visitors will exhibit high dispersal propensity due to their ‘exploration and new experiences driven’ nature.

As discussed above, a widely observed pattern of travel behaviour commonly associated with the proliferation of LCCs is the emergence of short and more frequent breaks. This increases the number of destination alternatives in a tourist’s holiday choice set. From the viewpoint of a single destination, greater destination alternatives for the tourists will increase the level of first time visitors. However, as previously discussed, it is uncertain whether or not the greater incidences of first time visits will increase dispersal.

3.3.8 Travel mode choice to and within the destination

As discussed in 3.2, there are two regional dispersal issues closely related to ground travel modes. The first issue is related to the transport mode choice within the destination. Thus, the dispersal of the air arrivals is influenced by the ground travel attributes and ground travel mode availability at the destination. The second issue is related to the regional dispersal impacted by the substitution of modes from the car to air in getting to a destination. We briefly discuss these relationships in this chapter. These issues will be revisited in much greater detail as case studies in Chapter 5 and Chapter 6 respectively.

Tourists arriving on air transport need to rely on the access and availability of local travel modes to realise their desired spatial behaviour. Visitors not restricted in travel mobility are “more spatially adventurous” (Debbage 1991: 368). Travel modes at the destination are important determinants of dispersal because the different modes are related to the different levels of ‘mobility’ (Lew and McKercher 2006).

3-21

There is little evidence available on the type of travel modes used by the air arrivals in the regions. Nor is there information on the preferences of LCC tourists toward a certain ground mode, and whether or not there is a significant difference between those who were enticed by a lower fare to travel as oppose to those who were not. However, we can assert that a greater proportion of LCC passengers are likely to be enticed by low fares than NCs. One European example suggested that LCC tourists have the greater propensity to hire rental cars (Barrett 2004), which supports the industry observations on the emergence of fly-drive travel in Australia during the second-wave (Tourism Australia, 2005). Data suggests that the rental car industry has been one of the winners from LCC proliferation; for instance, the Australian tourism satellite account (TSA) shows that the rental car industry has grown proportionately more (34% - although from a lower base) between 2000/01 and 2006/07 than the total industry gross value-added, which has been 17% (ABS 2009). A counter argument is that LCC tourists, as air travellers, face an additional burden of organising transport at the destination, which negatively affects their dispersal propensity.

The second issue is related to the bypass of destinations as a result of modal substitution. These two issues are linked in that a travel mode choice to get to a destination influences the travel mode used in the destination; for instance, if a tourist drove from the origin to the destination then that tourist will presumably also use the same vehicle in the destination. Thus, it is reasonable to think that tourists factor-in their need and desire for mobility at the destination in their choice of the main travel mode. Limtanakool et al. (2006) argued that the choice of private car in long-distance journey partly arises from the fact that car offers the flexibility to visit the attractions that have poor accessibility, e.g. residential neighbourhoods and out-of-town recreational areas. They argued that leisure trips are more likely to use private vehicles because leisure trips often involve travel with other people, which makes private car cheaper and more convenient, etc. Chapter 6 aims to address the question, ‘do cheap fares induce substitution

3-22 towards air travel even in situations where the private vehicle is more convenient and appropriate?’

3.3.9 Socio-economic variables

Most of the variables identified previously are readily measured for empirical analysis. However, other variables such as travel motivations and traveller personality are much more difficult to measure. Mansfeld (1990) noted how it is the variety in travel motivations and decision processes that underpins the variety in spatial behaviour. Motivations and attitudes are often measured by psychographic approaches (Mansfeld 1990); e.g., the study by Moscardo and Pearce (2004) on the interaction between travel mode choice and travel motivations. However, one limitation of such an approach is that data is often not available through secondary sources. This is perhaps in reflection of the fact that data obtained on psychographic studies tend to be highly destination, place or product specific (Mansfeld 1990). An alternative approach is to use proxies to capture the differences in tourists’ motivations from one individual to another.

Mansfeld (1991) argued that the use of socio-economic variables is a feasible approach in discriminating between tourists’ motivations (and the implied spatial behaviour) because tourists’ motivations are formed in the context of the socio- economic ‘environment’. Thus, there is a good reason to believe that tourists who exhibit similar background will show similar spatial behaviour. Socio-economic variables are not “direct travel determinant(s), but as a personal situation that might result in or impinge upon certain subjective travel motivations” (Mansfeld 1991:383). It is assumed that these factors can “effectively discriminate between different patterns of tourist spatial behaviour” (Debbage 1991, p.254).

Similar rationale underpins the use of socio-economic variables in the spatial applications of micro-econometric choice models. In disaggregate behavioural analyses, it is common to include socio-economic variables as ‘conditioning variables’ for variation in tastes and preferences of individuals. Income and age

3-23 are frequently used variables to approximate the effect of taste heterogeneity (Hensher et.al. 2005). Income, for example, is included in the model based on the assumption that individuals with high income display considerably different tastes compared to lower income individuals; income is not included as a measure of the purchasing power (Jara-Diaz, 1991). In the context of multi-destination travels, Tideswell and Faulkner (1999) used income as a proxy for the level of ‘economic rationalism’, arguing that higher income individuals are also more economically rational, which increases their tendency to visit multiple destinations. Thus, in the studies of spatial behaviour, if data permit, socio-economic variables should be included in the analyses.

3.3.10 Other variables and issues

Type of Accommodation Information on the type of accommodation chosen is relevant to dispersal for two reasons. First, accommodation type such as resorts provides facilities within the complex that may reduce the need to venture out. Second, the chosen accommodation is a useful indicator of the intention and motivation of the traveller. For instance, choosing to use ‘camping grounds and caravan parks’ indicate the travellers’ trip motivation and special interests, which is associated with a greater tendency to disperse to the periphery. Another example is the choice of ‘friends and relatives’ property’. This is related to dispersal because such choice of accommodation contains some information about the location of travellers’ main overnight stay. Given the fact that 80% of VFR trips stay in ‘friends and relatives property’ (NVS 2007), it is plausible to suggest that VFR trips will have the greater tendency to visit the residential areas of the destination region.

Trip expenditure In the microeconomic theory of choice, there is an important difference between income and expenditure. The former, as previously discussed, is a determinant of taste heterogeneity (Jara-Diaz 1991). Mansfeld (1991) observed that higher

3-24 expenditure is associated with a more expansive spatial behaviour (greater incidences of ‘moving from places to places’). While greater trip budget may allow for greater spatial behaviour, it is also the case that the desire or the need to realise an expansive spatial coverage usually requires larger trip spend. In addition, greater expenditure may arise from heavy shopping activity and city- centre activities that are not necessarily spatially expansive. This issue of ‘association not causality’ applies to other variables outlined in this Chapter. For instance, with respect to length of stay, a traveller may choose a short duration trip because that is all that’s needed for the trip, not because the traveller has a strict time constraint. Having said this, in a time-poor and cash-rich society, it is likely that the traveller is time constrained.

3.4 Summary

This Chapter focussed on the relationships between the LCCs and regional dispersal. The outcomes from this Chapter are twofold. First, the spatial patterns identified (SDT, MD1 and MD2) provided a framework in which specific LCC and regional dispersal related issues could be clarified and addressed. Two such issues emerged; (a) the extent to which destination policy control variables, in particular ground transportation, can influence the regional dispersal of the air arrivals; and (b) the extent to which LCCs can trigger a ‘bypass’ of the regional destinations without domestic air services, by promoting a modal substitution of tourists from ground travel modes towards air travel. These issues are further examined in Chapter 5 and Chapter 6 respectively. Second, this chapter generated propositions on how the LCC demand may be different from the NC demand with respect to their effect on dispersal. Propositions were explicated and summarised in Table 3.1. The following chapter empirically tests these propositions.

3-25

4. THE ‘CHARACTERISTICS’ MODEL

4.1 Introduction

The proliferation of Low Cost Carriers (LCC) marked a new era in air travel, generating much interest in the industry and academic literature on the LCC model and its impact on various aspects of aviation and tourism. This research concerns the LCCs’ impact on regional dispersal of tourists. If we have an empirical model that specifies the relationships between dispersal propensity and trip characteristics, we can then ask the question, ‘how are the empirical models of dispersal and trip characteristics differ between LCC and NC travellers?’ Put differently, ‘are there sufficient differences between the LCC and NC travellers to imply a divergent behaviour at the destination?’ The propositions were introduced in Chapter 3. The primary aim of this Chapter is to test these propositions. Many studies have outlined the various factors influencing spatial behaviour. But the relevance of these studies in understanding the impact of LCC, and more generally, the impact of the increases in the air arrivals, is left unexplored. The contribution of this Chapter is in partly filling this void in the research literature.

This Chapter briefly re-introduces the relationships between LCCs and dispersal, then we outline the steps involved in building the dispersal model with the National Visitor Survey data (section 4.3). Logit model results are presented and discussed (section 4.4). As for definitions, the LCCs in Australia are Virgin Blue and Jetstar, and a trip is classified as regional dispersal if it had at least one night stay in the periphery of a given Tourism Region. Note that Tiger airways has been excluded from the analysis due to very low sample size (NVS 2007 has one

4-1 month of Tiger airways data because the airline entered the market in December 2007).

In this empirical study, six of the propositions derived from Chapter 3 are tested. These are summarised in Table 4.1 along with the test results. Details of the test results will be discussed in later sections.

Table 4.1 Summary of the relationships between LCC and dispersal

Factors Effects on regional dispersal Propositions on the characteristics of LCC Test results propensity demand from a dispersal viewpoint

Number of stopovers* Number of stopovers is positively Number of stopovers is positively related to supported related to dispersal propensity dispersal propensity

Preference Greater heterogeneity can have a There is no clear proposition, but LCCs serve - heterogeneity* positive or negative effect on proportionately more couples and family dispersal travels than NCs.

Risk and uncertainty* Greater risk and uncertainty can LCC demand may be more sensitive to risk supported have a positive or negative effect and uncertainty, hence the effect of distance on on dispersal dispersal may be magnified

Length of stay* Length of stay is positively related LCC demand will be less sensitive to length of supported to dispersal stay than NC demand

Spatial configuration Different tourism regions will be Different tourism regions will be associated supported of destinations* associated with different levels of with different levels of dispersal dispersal

Variety and multiple- Greater variety in the reasons of LCC demand has higher dispersal propensity not tested benefit seeking travel increases dispersal because of the greater variety in the reasons of (data behaviour propensity. travel limitation)

VFR travel purpose* VFR related travels are positively VFR is an important source of dispersal of the supported related to dispersal LCC arrivals

Package tourism Package tourism is negatively Disproportionately large share of LCC arrivals not tested related to dispersal are FIT tourists, therefore they are less (data constrained spatially. limitation)

Note: asterisk (*) indicates variables empirically examined. Other variables were omitted from the analysis due to data limitations.

4-2

4.2 Method

4.2.1 Data

National Visitor Survey 2006 and 2007 National Visitor Survey (NVS) is the largest (in terms of sample sizes) travel survey available in Australia with approximately 40,000 trip samples each year. The survey collects a large amount of information on a range of variables. All variables are collected at an individual trip level. It is the most comprehensive disaggregate data source on Australian domestic travel, which began collecting information on the domestic airline used from 2006. This survey is managed quarterly by the federal tourism research agency, Tourism Research Australia (TRA).

Study sample This study examines ‘all leisure trips (Holiday or VFR) made by Australians originating from state capitals and destined (with at least one night stay) to Tourism Regions directly serviced by LCCs’. Tourism Regions are administrative boundaries set-up by state and territory tourism organisations (as outlined in Chapter 2, Figure 2.2). The origin-destination pairs are shown in Table 4.2. Six pairs1 were of distance too short (less than three hours drive) for air travel to be of any significance. Consequently they were removed from the sample. The study sample also excluded trips with greater than four stopovers (overnight). To maximise sample size, 2006 and 2007 samples were combined to obtain a total sample of 3,761 trips, of which 3,042 trips were leisure.

1 Brisbane – Maroochydore; Brisbane – Ballina; Gold Coast – Ballina; Gold Coast; Maroochydore; Sydney – Newcastle; Hobart – Launceston.

4-3

Table 4.2 Origin-Destination sample

Origin Destination State (destination)

Cairns Queensland Sydney Launceston Tasmania Melbourne Townsville Queensland Brisbane Maroochydore Queensland Adelaide Williamtown New South Wales Perth Mackay Queensland Hobart Rockhampton Queensland Darwin Broome Western Australia Canberra Proserpine (Whitsundays) Queensland Gold Coast Hervey bay Queensland Ballina New South Wales Coffs Harbour New South Wales

The National Visitor Survey (NVS) shows that between the year-ending 1998 and 2007, the proportion of gateway only visitors (non-dispersal) varied between 67% and 71% (or the regional dispersal varied between 29% and 33%). However, even the NVS is limited at this level of disaggregation due to high confidence intervals. Thus, these figures should be used as a guide only.

4.2.2 The Model

We apply a binary logit model to the regional dispersal problem. Dispersal is defined as having a discrete binary outcome, i.e. to disperse or not. The model applied in this study is advantageous over alterative methods considered, such as analysis of variance, in that the model provides a ceteris paribus effect of the independent variable on the discrete dependent variable (disperse or not).

In general, the following utility function is estimated for each option in the conditional logit model (introduced in Chapter 1),

= + + + Vni i i X ni iTni iZni Eq. (1)

4-4

There is one important difference between the conditional logit above and the multinomial logit model applied in this study. In the conditional logit, the effects of both attributes and individual characteristics can be estimated. The distinction is that the former varies across choice alternatives (e.g. airfares on Jetstar vs. airfares on Qantas), while the latter varies across individuals (e.g. airline used, length of stay of a trip, or income level of an individual). The multinomial logit model (which often refers to conditional logit today – and referred to as such in the subsequent Chapters) is technically a ‘characteristics model’ (Maddala 1986). We apply this model because the alternatives, to disperse or not, are functions of the individual trip characteristics, which varies across individuals not across choice alternatives.

The characteristics model is algebraically equivalent to the conditional logit model (Maddala 1986). Simply put, for a binary outcome, the probability of dispersal reduces to the following form:

eVni = = pn (dispersal 1) Eq. (2) 1+ eVni where

= + + Vni i iTni iZni Eq. (3)

4-5 All terms are defined the same way as Eq.(1). Trip characteristics, Tni , refers to trip factors outlined in Table 4.1.

4.2.3 Dependent and independent variables

The dependent variable: operationalising ‘dispersal’ Regional dispersal is defined as a trip with at least one night stay in a region outside the gateway city. Thus, an issue arise as to how the boundary of the gateway is defined. In this study, we chose to use a politically salient boundary, the Local Government Area (LGA), to distinguish the gateway from the periphery. This is the smallest political spatial unit in Australia. The most basic unit is the Statistical Local Area (SLA) established by the Australian Bureau of Statistics (ABS) for the Census. In most cases, several SLAs form one LGA. Most gateways in the regions are made-up of either one or two LGAs. Several, or sometimes numerous LGAs form a Tourism Region (there are approximately 1,000 LGAs in Australia but only 80 Tourism regions). Now that we have defined the basic spatial unit of a gateway and the periphery, we can move onto the task of defining whether or not an individual trip constitutes dispersal.

An overnight trip can belong to one of the following mutually exclusive category of trips:

o Trips that involve stay only in the gateway (denote this SDG); o Trips that involve stay only in the periphery (denote this SDP); o Trips that involve multiple stays only in the gateways (denote this MDG); o Trips that involve multiple stays only in the periphery (denote this MDP); o Trips that involve mixture of nights in both the gateway and the periphery (denote this MDX)

Based on the regional dispersal definition adopted, we can reduce the five categories above into a binary outcome:

4-6 (1) No dispersal option: a gateway(s) only trip (SDG or MDG); or (2) Dispersal option: a trip that involves at least one night stay in ‘periphery’ (SDP or MDP or MDX)

Thus in Eq[2], P(Y=1) is P[(SDP or MDP or MDX) = 1].

4-7 The independent variables

Independent variables consist trip characteristics, Tni , and individual characteristics, Zni (summarised in Table 3).

Table 4.3 Independent variables

Factors (variable codes) Coded values

Number of stopovers (up to 4 stopovers only) One overnight stopover* Dummy (0 or 1) Two overnight stopovers (stops2) Dummy (0 or 1) Three overnight stopovers (stops3) Dummy (0 or 1) Four overnight stopovers (stops4) Dummy (0 or 1)

Preference Heterogeneity (travel party) Alone* Dummy (0 or 1) Couples (coup) Dummy (0 or 1) Family (fam) Dummy (0 or 1) Friends and Relatives with Children (vfrch) Dummy (0 or 1) Friends and Relatives without Children (vfrnoch) Dummy (0 or 1)

Risk and Uncertainty (short - 800km or less) Dummy (0 or 1)

Length of stay (nights) Number of nights

Spatial configuration of destinations Major International* Dummy (0 or 1) Peri-Capital regions (pcap) Dummy (0 or 1) Coastal with significant international visitors (coint) Dummy (0 or 1) Coastal with mostly domestic visitors (codem) Dummy (0 or 1)

VFR travel purpose (proxy: type of accommodation) Friends or relatives' property (frp) Dummy (0 or 1) Repeat visitation (proxy: type of accommodation) Dummy (0 or 1) Own property (own) Type of accommodation (remaining types) Four star or greater hotels or resorts (lux) Dummy (0 or 1) All other* Dummy (0 or 1)

Note: asterisk (*) denotes the reference level

Three variables require further explanation: number of stopovers; type of accommodation; and spatial configuration of destinations. The previous Chapter has shown that dispersal is positively related to multi-destination travel patterns;

4-8 thus, ‘number of stopovers’ variable is included in the model. Accommodation types are used as proxies. There are two such proxies: friends or relatives’ property (FRP), and ‘own’ property. The former is used as a proxy for VFR. Given the fact that 80% of VFR trips are FRP (NVS 2007), the inclusion of both VFR and FRP would result in collinearity problems. The FRP variable was chosen over the VFR variable because the former provided a better model fit. ‘Own property’ was the best available data to account for the effect of ‘repeat’ visitation, which is not available in the NVS.

As for the spatial configuration of destinations, there is no clear guideline as to how this variable should be operationalised in the literature. What we know is that a desirable feature of the variable which operationalises the spatial configuration, should capture the ‘principal components’ of destinations from a tourism viewpoint. Tourism and Transport Forum (TTF 2002) groups the 80 Tourism Regions in Australia into eleven geographical categories based on their essential tourism and physical geographic characteristics. This was used to differentiate between tourism regions in the sample.

4.3 Results and Discussion

The model was estimated with maximum likelihood. The modelling process involved the evaluation of likelihood ratio tests, asymptotic t-tests and comparisons of Akaike information criterion (AIC). The asymptotic t-test results are shown with the coefficient estimates. In the estimation, a weighting variable was used for each individual trip. Tourism Research Australia (who manages the NVS) calculates the weights for each sample to account for the fact that respondents are asked for the last two trips and the fact that single-person households are over-represented in the sample. The sample is also adjusted to the

4-9 known age-sex distribution of the population. The weighting variable was obtained with the unit record data from Tourism Research Australia.

Figure 4.1 shows that ground transport modes (car, train and coach) are much more widely used for regional dispersal than air transport. In fact, 83% of leisure tourists who used ground travel modes as their ‘main’ mode of travel have undertaken ‘dispersal’ trips, whereas the equivalent figure for air transport is only 42%. This is not surprising as ground modes, especially the car, offers the most spatial flexibility – a feature of this travel mode widely recognised in research (e.g. Page 1994). Interestingly, LCCs are associated with lower dispersal than NCs. Figure 4.2 shows that dispersals for Virgin Blue and Jetstar were 38% and 33% respectively, while Qantas’ dispersal was 51%. The confidence interval is large at this level of detail; therefore, the percentage figures should be viewed only as a subject of interest and not as evidence of conclusive findings.

Figure 4.1 Regional Dispersal: ground transport vs. air transport (source: NVS 2006 and 2007)

4-10

Figure 4.2 Regional Dispersal by Airline (source: NVS 2006 and 2007 (1,190 observations))

Table 4.4 shows the summary results of the NC model and the LCC model. Both models were identically specified. The ratio of the two log-likelihood values is an indicator of ‘goodness of fit’, i.e. the pseudo R^2. There is no standard to which pseudo R^2 can be compared against, except that higher pseudo R^2 indicates a better fit (e.g. Borooah 1996). Hensher et al. (2005), based on simulations, have shown that a value of 0.3 is a good benchmark. The model falls short of this mark. However, this was expected given that information such as airfares, which is an important determinant of airline choice, was missing from the model. For the purpose of this Chapter, the current models are sufficient for our primary aim to hypothesis-test the factors in Table 4.1. An extension of this research will be to build on this model to incorporate travel modal attributes using stated choice data. The stated choice experiment approach is adopted in Chapters 5 and 6.

4-11 Table 4.4 Model summary

FSC LCC

LL of no coefficient model -260.23 -551.75 LL of Constant only model -258.55 -501.88 LL of the full model -213.78 -420.47 Observations 383 796 DF 19 18 P(Y=1) 0.51 0.34 Pseudo R^2 (LL constant only) 0.17 0.16 Pseudo R^2 (LL no coefficient) 0.18 0.23

In Table 4.5, NC and LCC columns show the coefficients and the statistical significance of the X variables. The final column, ‘NC-LCC’, shows the results from a statistical test to see whether or not the difference between the coefficients of NC and LCC on a common factor is statistically significant. We used the following asymptotic t-test suggested by Ben Akiva and Lerman (1985: 202):

NC LCC k k NC + LCC (var( k ) var( k ))

The results reveal that dispersal factors differentially affect NC and LCC.

4-12 Table 4.5 Model results

Factors (variable codes) NC LCC NC - LCC

Constant -0.38 -1.39 *** -

Number of stopovers (up to 4 stopovers only) One overnight stopover* reference level Two overnight stopovers (stops2) 2.43 *** 2.32 *** - Three overnight stopovers (stops3) 3.24 *** 3.34 *** - Four overnight stopovers (stops4) 1.97 * - - -

Travel Party Alone* reference level Couples (coup) -0.03 - 0.41 - - Family (fam) -0.64 * 0.63 ** *** Friends and Relatives with Children (vfrch) -0.67 - 0.78 ** ** Friends and Relatives without Children (vfrnoch) 0.38 - 0.08 - -

Distance (short - 800km or less) 1.68 *** -1.11 *** ***

Length of stay (nights) 0.05 ** -0.04 * ***

Spatial configuration of destinations Major International* reference level Peri-Capital regions (pcap) -0.44 - 0.73 *** ** Coastal with significant international visitors (coint) -0.45 - 0.90 *** *** Coastal with mostly domestic visitors (codem) -0.89 * 0.02 - -

Type of accommodation Own (own) 0.14 - 3.83 *** ** Friends or Relatives' Property (frp) -0.02 - 0.62 ** * Four star or greater hotels or resorts (lux) -0.61 * -0.04 - - All other* reference level

Age -0.02 - 0.01 - -

Note: [*] next to results of NC and LCC indicate 10% level of significance, [**] 5% and [***] 1%. [*] in the variable column indicates the reference group (cf. Table 4.2).

4.3.1 Number of stopovers

As expected, the greater the number of overnight stopovers, the greater the dispersal propensity. This variable is one of the most significant independent variable in terms of coefficient size. Furthermore, the relationship between the

4-13 number of overnight stopovers and dispersal is non-linear: the marginal utility of moving from one to two stopovers is greater than that of moving from two to three stopovers (see Figure 4.3). The final column of Table 4.5 shows that the coefficients of NC and LCC models on the number of stopover are not statistically different between the two models. The 4th stopover variable in the LCC model was omitted due to absence of valid observations.

Results indicate that the marginal effect on the dispersal propensity will be the greatest if single stopover tourists were targeted. This way the limited marketing and investment resources can be used to their most effect. For greater dispersal, research and management should focus on trips up to three stopovers, not only because they take up the majority of visitations in Australia, but also because the marginal improvement in the dispersal propensity is the greatest in this range.

Figure 4.3. Marginal effects of stopovers on dispersal propensity (source: modelled results of NVS 2006 and 2007. Note: based on the ‘NC’ model results)

4.3.2 Length of stay

Length of stay has a positive and statistically significant coefficient in the NC model. However, length of stay is statistically significant but negative in the LCC model. Two points are worth noting. First, the average length of stay of LCC tourists is one night less (6.9 nights) than NC nights (8 nights), and the standard

4-14 deviation is also lower among the LCC tourists. Second, the LCC model shows that even if length of stay increases, it will be of little effect towards dispersal. Both points support the hypothesis devised earlier in Chapter 3, although the level of support is not strong (shown by the weak coefficient). A case could be made that LCC tourists are constrained by time, and this time-constraint provides little room for tourists’ dispersal propensity to be swayed by the length of stay. However, an average of 6.9 nights against an average of 8 nights is not a wide difference. Therefore, the evidence on length of stay as an intra-modal source of difference is weak. It may be the case that for the LCC demand, when length of stay increases, the desire for expansive spatial behaviour is met with other forms of travel such as day-trips, rather than a change in the overnight destination (because of the time constraint). However, the day-trip hypothesis was not testable from the current study.

The result on the length of stay may be affected by the fact that 30% of the leisure travel samples were VFR. VFR travellers’ spatial behaviour will be determined largely by the residential locations of friends and relatives. Thus, at least from a dispersal point of view, length of stay may not be so relevant for VFR trips. This does not mean that VFR travellers do not disperse; rather, the length of stay of VFR travellers, independent from other variables, has no bearing on the propensity to disperse.

4.3.3 Distance

It was proposed that the LCC demand is more responsive to distance, i.e. the coefficient of the LCC model will be larger than the NC model. The results show that the two models differ in their signs (+/-). The NC model has a positive coefficient on ‘short’ (less than 800km), whereas the LCC model has a negative coefficient. In the light of the hypotheses summarised in Table 4.1, this means that the LCC model supports the view that greater risk and uncertainty stimulates dispersal. In contrast, NC tourists focus on gateways as a consequence of greater

4-15 risk and uncertainty. The differences in the coefficients are statistically significant at the 1% level.

This evidence suggests a link between passenger characteristics, airlines and tourists’ destination spatial behaviour. In other words, LCC demand is associated with tourists who respond in a spatially expansive manner in order to reduce the uncertainty and risk, whereas NC demand is characterised by tourists who focus on the gateway in response to the uncertainty. One potential explanation for this difference may be traced to the operational differences between the airline business models. LCCs commonly adopt Point-Point (and direct) routes, whereas NCs commonly use hub-spoke (hence the name ‘network’ carrier). As distance increases, NC services are more likely to be hub-spoke to regional destinations. The network strategy may have a limiting effect on the spatial behaviour of tourists because greater time is spent on connections and in-flight; potentially at some expense of time and energy at the destination.

4.3.4 Spatial configuration of the destinations

The characteristics model supports the proposition that different spatial configuration of destinations (reflecting different physical factor endowments, as well as the patterns of human landscape) will produce different dispersal propensity. In the LCC model, coastal international destinations have the highest influence on dispersal (0.9), followed closely by peri-capital regions (0.73). The model shows statistically insignificant differences between major international destinations and coastal domestic tourism regions.

In the NC model, major international destination tourism region is the only region with a statistically significant effect on dispersal (positive effect). One potential explanation for the differences in the results of the NC and the LCC models is that major international destination such as Cairns (based on TTF classification described earlier) is by far the largest market for Qantas (the NC) in the study sample. It is also the case that Cairns and the surrounding destinations (between

4-16 30% and 44%) have much greater incidences of multi-destination travel by domestic visitors than the national average (11%). The ‘major international destination’ variable consequently stands out from all other variables in the model. Given the fact that spatial configuration is often fixed, or supply inelastic, understanding the response of the new air arrivals to this configuration is important for destination managers and planners.

4.3.5 Accommodation Type

In the LCC model, ‘own property’ has a large positive effect on dispersal. Nonetheless, this variable is a proxy, and not fully reflective of the broad range of repeat visitors. Similarly, ‘FRP’ (friends and relatives’ property) has a positive coefficient. This is partly due to homes located in the residential areas of a destination, which is in the outskirts of the main centres (same reason applies to interpreting the coefficient of ‘friends and relatives property’). As for the NC model, luxury resorts and hotels have a positive effect on dispersal. This is expected because they are often located in the CBD and offer superior facilities, reducing the need for tourists to venture beyond.

It is briefly noted here that ‘FRP and ‘own’ as sources of dispersal are less desirable from an expenditure viewpoint because they inject less expenditure into the region’s economy. For instance, NVS (2007) shows that holiday contributed an average of $637 per visitor, or $144 per visitor night, whereas the figures for VFR (who make up the majority of FRP) were $283 and $81 respectively (Tourism Australia 2008). Thus, even if this type of trip contributes to greater dispersal, its corresponding economic contribution to the region is much less than what the visitor dispersal volume may indicate. Dispersal arising from VFR may add to dispersal visits and nights, but comparatively little to expenditure and financial yield. Further to financial yield, given the traditionally labour-intensive nature of the accommodation industry (Dwyer, Forsyth and Spurr 2003), the marginal effect of a dollar spent by dispersing tourists may contribute little to the economic yield in the regions. Moreover, the level of leakages will be significant

4-17 in peripheral regions because small regional economies tend to have a more homogenous industry base; consequently, significant share of tourists’ expenditure will leak-out as import payments to other regions and abroad.

4.3.6 Accompanying travel party type

In the LCC model, compared to ‘travelling alone’, ‘travelling with family’ and ‘travelling with friends and relatives with children’ have a greater effect on dispersal. In the NC model, ‘travelling with family’ has a positive effect on dispersal. This travel party type, however, was the only party to have a statistically significant effect in the NC model. The result supports the view that greater preference heterogeneity will cause greater dispersal among the LCC tourists. However, in the NC model, the negative coefficient indicates a tendency for NC tourists to spatially ‘agglomerate’ than to disperse, when the heterogeneity increases. The LCC model supports the view that greater heterogeneity is a positive source of dispersal, whereas the NC model supports that greater heterogeneity constrains tourists’ spatial behaviour.

4.3.7 Other variables

In discrete choice theory, income is often used as a taste parameter (e.g. Ben Akiva and Lerman 1985). The behavioural implication of income differentials is that the spatial behaviour of tourists with low income will be different from that of high income (e.g. Mansfeld 1992). Unfortunately, the income variable had large incidences of missing data (approximately 20% of the sample). Although various strategies to overcome the missing data problem were considered, it was judged most appropriate to exclude this variable from the analysis. The ‘package- trip’ and ‘rental vehicle’ variables were also omitted from the analyses for similar reasons. Age variables were included in the model but they were found statistically insignificant.

4-18

4.4 Limitations

Some of the sample route pairs shown in Table 4.2 have small sample sizes, and not served by all three airlines. For instance, Jetstar completely replaced Qantas services to Hamilton Island in 2006. Consequently, on this route, the route specific effect on dispersal is confounded with the airline specific effects. Future studies should be more case specific, although this may be a problem due to low sample sizes. Furthermore, alternative methods of operationalising dispersal should be considered in the future. The use of LGA boundaries, while politically salient, is arbitrary for the tourists (since they have little or no knowledge of these boundaries). One way this can be done is to specify a multinomial model with levels of dispersal as dependent variables.

Two methodological limitations are noted: the use of characteristics model, and the use of revealed preference data. The characteristics model applied in this study specifies an outcome (dispersal or not) as a function of trip characteristics. With such an approach, there is a problem of not knowing the direction of cause and effect, i.e. the length of stay may be negatively related to dispersal because tourists are constrained by the time-budget, or it may be that tourists choose short-trips (a short getaway in the gateway). One solution to this problem is to apply an experimental approach. This way, the researcher controls the environment in which tourists make choices. This method (stated choice method) has been applied to tourism related issues for some time (e.g. Louviere and Hensher 1983). Stated choice method will be applied in the following two case studies.

4-19

4.5 Conclusion

This Chapter aimed to empirically test the relationships between regional dispersal and affordable air services. The study found some clear differences between the LCC and the NC model; preference heterogeneity (larger travelling party size); travelling to second homes; staying at friends and relatives’ property, and the risk and uncertainty factors were major sources of dispersal in the LCC model. The evidence from this study supports the view that intra-mode differences can be a differentiating factor of the behaviour of tourists at the destination. It was shown that some of this information is contained in the tourists’ airline choice.

4-20

5. THE CAIRNS EXPERIMENT

5.1 Introduction

Low Cost Carriers (LCCs) stimulate domestic dispersal of tourists in Australia. Regional destinations experienced increased number of air arrivals as a consequence. The greater pool of visitors creates opportunities to increase regional dispersal. The corollary is the increasing reliance of a destination on air transport. What are the factors that influence the air arrivals to disperse? Is it possible to induce tourists to disperse by implementing appropriate destination transport policy? Specifically, can destination transportation policy stimulate the dispersal of the air arrivals, even in situations where the air arrivals exhibit trip characteristics that are dispersal-adverse? These are the questions that this Chapter aims to answer.

In this Chapter, we address the fourth specific aim of this thesis, which is to ‘examine the effects of destination transport factors and tourists’ travel characteristics on air arrivals’ regional dispersal by applying a stated choice experiment’. A research design that accommodates trip characteristics and destination transport attributes so that their influences on regional dispersal can be compared, will provide us with the capacity to draw conclusions on the likelihood of destination transportation policy to stimulate dispersal (of the air arrivals), even in situations when the air arrivals exhibit dispersal-averse propensity.

Transport issues are often at the centre of public policy agenda where governments promote certain modes of travel over others to meet a wider policy objective (e.g. reduce carbon emissions). Thus, public policy can be subject to

5-1 conflicting interests (Eaton and Holding 1996). For instance, a conflict may arise between the objective to maximise dispersal and the objective aiming to maximise the use of public transport; this is because while a car is a pertinent mode of travel for regional dispersal, environment-led policy may advocate a shift away from a car towards public transport. Research on the connection between local travel modes and regional dispersal can provide diagnostic information to help make more informed choices on the allocation of public funds.

Eaton and Holding (1996) concluded that public projects need to be able to induce a change in behaviour; thus, given the fact that public projects can be expensive and riddled with conflicting interests, an ‘experiment’ may be desirable to first demonstrate the potential of the project. In addition to the reasons discussed in Chapter 1 (p1-19) and Chapter 4 (p4-19), stated choice method was chosen for this case study because the method enables the researcher to control the levels and type of travel attributes whether or not the attributes are actual or hypothetical. As highlighted in 1.4 (Chapter 1), this provides an opportunity to empirically test the effects of travel mode variables under ‘what if’ scenarios. An example may be to examine the effectiveness of public bus attributes designed to facilitate greater regional dispersal of tourists in the regions. This Chapter adopts such a method by applying a stated choice experiment.

5.2 Regional dispersal and transport

A case study approach is adopted in order to progress through the research aims. Cairns and the Tropical North Queensland tourism region (TNQ) is the largest regional destination in Australia in terms of domestic air arrival volume (see Table 2.3). Prideaux (2000) has illustrated the relationship between the growth in domestic and international tourism with the development of air transport services and transport infrastructure in Cairns. He noted that beginning in the early 1980s,

5-2 air transport (and its declining relative costs compared to other modes of transport) became an important part of tourism development in the region. In the mid to late 1990s, a different trend emerged in the region. Moscardo et al. (2004) noted that between 1996 and 2001, the trend from air towards the use of long- distance road was related to the greater use of Whitsundays as the point of access to the Great Barrier Reef (GBR) than Cairns. Moscardo et al. (2004) argued that access modes are significant constraints for the tourists because a given mode fixates the arrival point to a particular access node, in which the subsequent travel patterns are influenced. The modal shift, they argued, was associated with more than just a shift in the access points, rather it was related to certain characteristics and constraints of tourists. They argued that the modal switch was associated with a shift from mass tourism towards smaller and specialised tourism. They noted that the patterns of tourism also became more peripheralised and diffused across the wider regions. The characteristics of tourists, they argued, also changed towards that of more repeat visit oriented, variety and activity-seeking tourists. Overall, Moscardo et al. (2004) stressed the importance of the arrival transport mode as an agent of change in the patterns of regional tourism.

Since the Moscardo et al. (2004) analysis of change in regional tourism, the ‘second-wave’ of Low Cost Carriers (LCCs) proliferated in Australia. This accelerated the tourist flows to Cairns. Cairns airport is today the second largest non-capital airport (after Gold Coast) in both international and domestic arrivals, although LCCs’ impact was mostly on domestic air services (but this does not mean that the impact was mostly on the domestic tourists because most international tourists travelling to Cairns have to use the domestic network as well).

Despite the reversal of trends towards air travels in the regions, it is very unlikely that the patterns of regional tourism will revert to the pre-1996 scenario for at least two reasons. First, LCCs tend to seek excess-in-capacity airports located in destinations with dense demand for leisure travels. During the second-wave, five airports (Cairns, Townsville, Mackay, Hamilton Island and Rockhampton) in the

5-3 GBR region gained direct access from the key metropolises through LCCs. These airports were previously not attended by direct domestic services, or they did, but did not have the exposure to markets that LCCs create. These airports are roughly equidistant from one another by approximately 4 hours drive along the 2,000km stretch of the GBR region (with the exception of Hamilton Island, which is nearby Mackay), which opens up possibilities for a numerous variation in the travel patterns compared to the pre-1996 era. A consequence of the modal shift has been the bypass of ground-mode-reliant smaller and peripheral destinations in the North Queensland region (Whyte and Prideaux 2007). Second, as discussed in Chapter 3, LCC demand characteristics tend towards free and independent travellers (FIT), which lends support towards the continuation of the post-1996 trends observed by Moscardo et al.

The increasing emphasis on air travel for tourism is representative of the experiences of many other Australian regional destinations. An obvious example is the airports outlined above, all of which have experienced large increases in the volume of air traffic inflows since 2001 (please refer to Table 2.3). The peripheral destinations surrounding the gateway cities, however, do not command sufficient demand for a separate LCC service; rather, they must rely on the air arrivals to disperse from the gateway. The issue this research aims to highlight is the effectiveness of ground transportation in increasing the dispersal propensity of tourists from the gateway.

The air leisure arrivals in Cairns have available to them the following trip alternatives (please refer to Figure 3.1): (1) at least one night stay in the periphery (single-destination trip destined for the periphery or a full/partial orbit pattern); (2) gateway only trip; and (3) stay overnight only in the gateway but take day- trips (base-camp). In the definition we adopted, only the first of these patterns constitutes regional dispersal. We build a choice model to test the effectiveness of ground transport factors in tourists’ choice between the three trip alternatives. The following sections introduce the methodology. The methodology sections describe

5-4 the discrete choice model applied, the factors influencing dispersal and transport mode choice, the experimental design, and the data collection procedure.

5.3 The Model

The basic discrete choice model for multi-alternatives is the multinomial logit model (MNL). In this Chapter, the following utility function is estimated for each mode of transport and in each destination context.

= + + + Vni i i X ni iTni iZni Eq. (5.2)

One limitation of the MNL is the independence of irrelevant alternative property (IIA), which results in the constant cross elasticity of the attributes. This occurs as a consequence of the assumption of ‘independent and identically distributed (IID) error term’, which enables the derivation of the simple MNL form. Violation of this assumption generates unrealistic market share prediction of the choice alternatives (see Ben Akiva and Lerman (1985) illustration of the ‘blue-bus and red-bus’ problem). Nested logit is a natural extension of this model by partly relaxing the assumption of constant and equal variances of the error terms (i.e., IID). Both models were estimated in this study. As shown later, MNL was found sufficient for our purpose.

5-5

5.4 Alternatives and attributes

5.4.1 Alternatives

Three pieces of information is required for the model to yield the results necessary to achieve the aim of this research: information on the trip structure chosen by the tourists; information on the travel mode used by the tourists; and information on the context in which these decisions were made. The third is related to the destination context in which the tourists make their choice. Destination contexts are used here to represent a collection of destination attributes. Thus, the contexts are expected to have a significant influence on the choice behaviour of tourists. The three choice dimensions are illustrated in Table 5.1. Each dimension is discussed below.

5-6 Table 5.1 Three choice dimensions Number of Choice dimension and attributes Alternatives attributes

North South Cairns and Trip structure (dimension1) [Overnight trip] [ Day-trip ] GBR only

- Travel mode (dimension 2) RC PB RC PB TOUR

North/ North/ North/ North/ North/ - Destination (dimension 3) South South South South South Abbreviation for the choice alternatives RCD PBD RCB PBB TOUR Gateway

Attributes [Destination expenditure] [Generic for o/n ][ Generic for d/t ] -2 2 Attributes [Price] Yes Yes Yes Yes Yes - 5 5 Attributes [Travel time] Yes Yes Yes Yes Yes - 5 5 Attributes [car type] Yes - Yes - - - 2 2 Attributes [sightseeing stopovers] - Yes - Yes - - 2 2 Attributes [Driver characteristics] - Yes - Yes - - 2 2 Attributes [Frequency] - Yes - Yes - - 2 2

Total 20 20

Note: Rental (RC); Public Bus (PB); Small organised tour (Tour); Take a overnight trip and travel by rental car (RCD); Take a day-trip and travel by public bus (PBB) etc. Destination expenditure attribute is generic for all ‘overnight’ and ‘day-trip’ alternatives.

Trip structure (dimension 1) As previously mentioned, regional dispersal is a trip that involves at least one night stay in the ‘periphery’. This is defined as a region outside the politically salient boundary of Cairns city. An alternative to dispersal is a trip that involves overnight stays only in the gateway, i.e. Cairns city. Finally, a day-trip option from Cairns to the periphery is added to the choice experiment as an alternative to an overnight trip. Thus, the three trip alternatives are: ‘gateway only’ vs. ‘base- camp (day-trip beyond the gateway)’ vs. ‘at least one overnight stay beyond the gateway’.

Travel mode (dimension 2) There are several alternative modes of travel available for the type of trips mentioned above; including, rental vehicles, taxi, tour shuttles or four wheeled- drive operators, rail services, as well as non-motorised travel modes. According to

5-7 information available from the regions’ travel internet sites, the most popular form of travel is a day-trip by a car or through a tour operator. Currently, public transport (e.g. Sunbus) is available within town centres and the suburbia, as well as on selected inter-regional routes. Other services such as Skyrail and boats also provide transport for tourists, although they are more limited to specific locations and tour activities, such as rainforest tours and the tour of certain islands in the Great Barrier Reef.

Three salient travel mode alternatives were identified from the viewpoint of transport and tourism policy on regional dispersal. In addition to rental cars, public transport was identified as a potential alternative. The third alternative is located ‘in-between’ on what may be called the ‘characteristics space’ with public bus on one end of the spectrum and the car on the other. ‘Small-group tours’ often offer a level of flexibility and privacy that may be perceived to be a mixture of the car and the public bus; for instance, while the car is wholly flexible for the trip desired and the public bus more restrictive because of its scheduled and ‘public’ nature of its characteristics, small-group tour belongs to neither of the categories. Rather, it shares some aspects of both. The three alternatives are different from each other to an extent that it helps to preserve the IID assumption, which renders the MNL model more appropriate.

V Thus, there are six alternatives each with the ni in Equation 5.2. The alternatives are products of the ‘trip structure’ dimension and the ‘travel mode’ dimension shown in Table 5.1. These are:

o Overnight trip beyond Cairns using a rental car (denoted by RCD); o Day-trip beyond Cairns using a rental car (denoted by RCB); o Overnight trip beyond Cairns via public bus (denoted by PBD); o Day-trip beyond Cairns via public bus (denoted by PBB); o Day-trip beyond Cairns with a small group tour operator (denoted by Tour); o Stay in Cairns only

5-8

Destination (dimension 3) Finally, two destination contexts are added to the experiment to account for the effect of destination characteristics. Tropical North Queensland region (TNQ) is characterised by its diverse range of attractions. This engenders a major challenge for delineating an appropriate ‘destination’ boundary, as well as a challenge in the identification of a parsimonious set of destination contexts so that the size of the experimental design remains practically feasible. For instance, it is commonly cited in traveller information brochures that TNQ offers experiences ranging from the City (Cairns) and beaches, to rainforests and tablelands, and the GBR. Current travel patterns were used as the basis for reduction in the number of destination contexts. Tourism accommodation establishments, bed-spaces and room number statistics released by the Australian Bureau of Statistics (ABS) were consulted to identify the key destinations of overnight stays. Based on these figures, it was identified that most (over 90%) of accommodation establishments were in the Coastal regions (including Cairns, which has a 67% share).

Within the coastal regions, Local Government Area (LGA) profiles published by Tourism Research Australia (TRA) were used to delineate two contrasting geographic regions: the North and the South (Figure 5.1 shows the map of the region used in the actual survey). Douglas and Johnstone are the representative LGA in the North and the South respectively. The LGA profiles show that the Johnstone LGA (south of Cairns) and the Douglas LGA (north of Cairns) are similar in that:

• a high proportion of travellers to those regions are for leisure (holiday or VFR) purpose (87% and 91% respectively); • ‘beach’ is the main activity that overnight tourists engage in these destinations (53% and 61% respectively); • a high proportion of visitations is likely to be part of a multi-destination travel itinerary, probably involving overnight stay(s) in Cairns (44% and 41% of Johnstone and Douglas overnight visitors also stayed overnight

5-9 elsewhere in their trip, compared to 30% of visitors to Cairns and 11% national average).

However, these are as far as the strong similarities are observed. Key differences are:

• the average spend differs significantly, with per night expenditure of $92 in Johnstone against $223 in Douglas, indicating that Johnstone is a more affordable alternative; • Johnstone has a much higher share of ‘caravan parks’ accommodation than Douglas (according to ABS, Johnstone shares 2% of ‘hotels, motels and apartments’ bed spaces, but 13% of the region’s caravan parks. Equivalent figures for Douglas are 20% and 12% respectively); • the length of stay in Douglas is higher (5.2 nights) than Johnstone (3.8 nights), thus, the South is relatively less popular in both volume of traffic and in number of nights; • only 24% of Johnstone visitors are of interstate origin whereas the equivalent figure for Douglas was 65%. This reflects the fact that many of the air arrivals (from Sydney, Melbourne) are also less familiar with the South, and the visitation to this region is of lower priority than the North for these visitors.

Thus, it was judged that these two regions – the North and the South - were sufficiently different in characteristics to interact differently with the mode choices of the air arrivals.

5-10

Figure 5-1 Map of the Cairns region shown to the survey respondents (drawn by the author)

5-11 5.4.2 Attributes and attribute level labels

Some of the most common mode choice attributes in the journey-to-work trip contexts are price and time. In addition, there is a wide range of qualitative variables (although in practice, some attributes are used more often than others) such as frequency, expected delays, etc (see Hensher and Prioni 2002). These variables are also sometimes collectively referred to as ‘instrumental’ variables, including ‘flexibility’ and ‘convenience’ as well as costs (Anable and Gatersleben 2005).

Eaton and Holding (1996) suggested that the following factors are important in the choice of public transport for recreational travel to National Parks (in order of importance): punctuality; convenient park; lower fares, and the use of ‘novelty vehicles’. More recently, Lumsdon (2006) provided a qualitative study on the issues surrounding the promotion of public transport for tourism in the UK. Based on in-depth interviews of key stakeholders, Lumsdon (2006) found that ‘sightseeing’ is an important market segment for leisure and recreational use of public bus services. This implied that certain public transport attributes were more desirable than others. Two of the main attributes noted by Lumsdon were ‘en route stopover opportunities’ and ‘driver knowledge about the destination and friendliness’. These are also examined in this Chapter. Interestingly, Eaton and Holding (1996) and Lumsdon (2006) did not stress travel time as a significant factor in the patronage of public transport over private car. The discussion section revisits this topic on travel time.

Two groups of travel mode attributes were identified above: economic variables and ‘tourism’ variables. Economic variables are widely used in urban mode choice research, but tourism variables are rarely considered. Destination expenditure attribute was added as a quantitative measure of destination characteristics. As previously mentioned, the use of ‘destination contexts’ design accounted for the destination attributes. The attribute level labels are summarised in Table 5.2 below.

5-12

Table 5.2 List of attribute level labels

Attributes Attribute level labels Rentcal car alternative Daily rate (incl. fuel) $50, $100, $150 One-way 'in-vehicle' travel time 1 hour, 2 hours, 3 hours Car type economy, luxury, 4WD Public bus alternative Price per person Free, $40, $80 One-way 'in-vehicle' travel time 1 hour, 2hours, 3 hours Driver attribute below expectation, average, above expectation Sightseeing Non, 1 or 2 stopovers, more than 2 stopovers Frequency every 1 hour, every 2 hours, every 3 hours Small group all-inclusive tour Price per adult (child) $100($50), $150($75), $200($100) One-way 'in-vehicle' travel time 1 hour, 2hours, 3 hours Destination expenditure (per night per person) Northern destinations $120, $170, $220

Southern destinations $70, $120, $170

(i) Price For the rental car alternative, the attribute level labels were based on daily and five-day rates of the major rental firms. Public bus fares were based on current inter-regional bus fares (e.g. SunExpress). A ‘free’ ride attribute level label was added to the experiment to maximise the level of conditioning for this alternative. The ability to analyse an effect of a hypothetical alternative such as a ‘free public bus’ is one important advantage of the stated choice method. Finally, the labelling of price attribute for the tour alternative was based on day-tour information from brochures and websites. All websites were accessed in the first week of August for prices in the period between 21st and 27th of August, which was the actual survey period in Cairns.

5-13

(ii) In-vehicle travelling Time (one-way) and frequency The attribute level labels for the travel time attribute was based on ‘google map’ information on distance and imputed travel time. The attribute labels for frequency were based on current frequency of regional bus services.

(iii) Rental vehicle type For the rental car alternative, rental vehicle type was also added as an attribute. This was considered important because consumers often associate quality and price in their choice (Hensher et.al. 2005); thus, to not include information on the quality of rental vehicle may induce tourists to choose a high price alternative because they associate this with higher quality. Consequently, this has the danger of measuring the combined effects of price and quality, not only price.

(iv) ‘Tourism variables’ for public transport As mentioned previously, two tourism attributes are added to the public transport alternative: ‘one or two stopover in special places’ and ‘driver knowledge and friendliness’.

(v) ‘Comfort’ factors Anable and Gatersleben (2005) have shown that ‘freedom’ and ‘control’ are the affective qualities of a car that travellers emphasise over public transport. In addition, ‘flexibility’ and ‘convenience’ of a car are also important (e.g. Anable and Gatersleben 2005). While each of these factors could not be included in the experimental design for practical reasons (to contain the size of the experimental design), the survey asked the respondents to rate the ‘comfort’ of travel modes on a Likert scale. Although this is an imperfect measure of the affective factors, it captures some aspects of the qualitative attributes that may not be so easy to explicate in the experimental design. Similar methods have been used, for instance by Koppelman and Sethi (2005), in inter-regional mode choice experiments.

5-14 (vi) Expenditure at the destination It was shown previously that destinations in the South are much more affordable than those in the North. Different levels of expenditure per night per person were specified in reflection of this difference. It was shown that average daily expenditure in Cairns is approximately an average of the expenditures in the North and the South.

5.5 Experimental design

5.5.1 Orthogonal main effects design

A key issue in the experimental design for choice modelling is whether or not a design should allow for testing of the violation of the identical and independently distributed error terms (IID) assumption. The outcome from such a test subsequently provides the basis for extending the analysis with more sophisticated models (Louviere et.al. 2000). Given the choice dimensions of this study, it was appropriate for the experimental design to be non-IID, so that non-IID models could be estimated from the data collected (e.g. nested logit). A sufficient condition for a non-IID design is when all attributes are orthogonal with one another within and between alternatives (Louviere et.al.2000). Thus, for this study, a design that can accommodate at least 20 orthogonal attribute columns was required (the number of attributes shown in Table 5-1).

A fractional factorial of 320 was selected. The fractional factorial allows up to 20 orthogonal columns, each with three levels. In 54 treatment combinations (choice sets), this is an orthogonal main effects only plan. Thus, the effects of two-way and higher order interaction are not protected from confounding with the main effects; for instance, the effect of price of an attribute is independently estimated

5-15 from effects of all other attributes, but there is no guarantee that this effect will be independent from the interaction effect of, say, price and time. This design was replicated to produce choice scenarios in the context of trips to the Northern region and another complete set of scenarios in the context of trips to the Southern region. Thus, there are 104 treatment combinations in total (after removing the treatment combinations without any designed trade-offs), and this was blocked into 26 versions to produce four choice scenarios for each respondent. All alternatives are present in the choice scenarios, and each respondent received two scenarios each from the North and South destination contexts.

5.5.2 Coding and design orthogonality

Effects coding enables the model to estimate the effect of a particular variable as a deviation from the grand mean (the mean of the unobserved utility). This coding scheme is necessary in order to estimate non-linear effects without the non-linear effects confounding with the alternative specific constant (e.g. Hensher et.al. 2005). However, the effects coding scheme generates correlation among the levels of the same variable. As previously discussed, one advantage of using a stated choice experiment is that the values of the explanatory variables are not correlated. But orthogonality can be lost in many ways (see Louviere et.al. 2000 for details). In this study, the effects-coding structure of the variables from ‘high’ to ‘medium’ to ‘low’ is one source of correlation within a given attribute of an alternative. Pairwise correlation matrix is a common strategy to test for design correlations. As expected, the effects coding structure gives rise to a correlation of approximately 0.5 within the levels of a given independent variable.

The extent to which 0.5 is a problem is difficult to know, although a rule of thumb value, for instance, 0.8 can be used as a benchmark (Hensher et.al. 2005). As a consequence of the correlation, the coefficient estimates may become unstable and standard errors may become very large, affecting the asymptotic t-tests of statistical significance (Greene 2002). Thus, we interpret the coefficients with caution in discussing the results. It is noted that correlation is only a problem between the levels of an attribute of an alternative; for example, the correlation is

5-16 introduced between a ‘high’ price and a ‘low’ price of the rental car alternative, and not between the price and time attributes of rental cars. All designed attributes, by the virtue of the design, are orthogonal with respect to all other attributes within and across alternatives.

5.5.3 They survey

The survey was conducted at the Cairns domestic airport terminal in the period between 22nd and 27th of August in 2008. The peak period in Cairns tourism is between April and October, as other months are part of the wet season. There was a continuous flow of visitors throughout the day, to Sydney, Melbourne, Brisbane, Perth and Adelaide. All visitors who regarded themselves as residents of these cities were eligible for an interview; provided the purpose of their trip was ‘visiting friends and relatives’ or/and holiday, and they had taken one of Jetstar, Qantas or Virgin Blue flights.

While the primary component of the survey was the hypothetical choice scenarios, other trip information was gathered. The questionnaires on trip information were designed to mimic that of National Visitor Survey conducted by Tourism Research Australia. While most of the questions on trip details and personal information were not found intrusive, a question on ‘income’ was ignored by more than 20% of the respondents. Consequently, this variable was dropped from the models. Pilot surveys were distributed to the students and staff of University of New South Wales, Research and Strategy division in Tourism Australia, and Cairns airport for feedback. A sample of the survey is provided in the Appendix.

Collection method was ‘simple random’ in that, for instance, interviewers approached travellers taking seating on every second row in the departure lounge area. The turnover of travellers was high. The final two days focussed on obtaining a more representative sample across demographic groups (age and gender), representing a stratified random sampling technique. The data collection exercise aimed for eight replications of the entire design, or 208 respondents.

5-17 After discarding unreliable responses, 196 surveys were judged usable, providing at least seven replications with a total of 784 choice observations. The face-to- face survey helped to assure reliable and informed responses.

5.6 Results

5.6.1 Descriptive statistics

The sample collected was slightly skewed towards male (59%). As a benchmark, the National Visitor Survey statistics on Cairns show that the share of 25-44 and 45-64 should be approximately the same (TRA 2008). Age groups of 18-25, 36- 45 and 46-55 were approximately equally represented with shares between 16 – 20% of the total sample. The age group 26-35 represented 31% of the sample, while the 56-65 age group accounted for 11%, and over 65 with 4%.

Trip characteristics information is presented below. 100% of the sample departed Cairns via air transport. However, there was a small percentage of sampled individuals who arrived on modes other than air travel such as train or rental vehicles. These respondents were subsequently removed from the analysis. The following trip characteristics are highlighted:

o Nearly half of the visitors sampled used ‘rental cars’ as a main mode of ground transport in the destination (43%). This is followed by walking (20%), private vehicle (11%), tour company (7%) and public bus (5%). The cases of private vehicles apply to friends’ and relatives’ vehicles.

o Half of the sample stated ‘hotels, motels and apartments’ as their main accommodation (51%). This type of accommodation, together with

5-18 ‘resorts’, accounted for 80% of the sample, while 10% indicated friends’ and relatives’ property.

o In regards to destination activities, 80% of the sample stated ‘eating out’ as their main form of travel activity, followed by ‘walk or drive around’ (75%) and ‘visiting the rainforest’ (56%). This pattern is consistent with the LGA profiles mentioned previously. Surprisingly, only 47% of the sample stated ‘Great Barrier Reef’ as one of their travel activity, indicating the diverse range of activities tourists seek from Cairns and the TNQ region. Further, the high proportion of ‘walk or drive around’ (75%) and ‘go to the beach’ (58%) against relatively low incidences of ‘day-trips with a tour company’ (31%) indicate that tourists prefer to do things themselves than to rely on the services provided by the local tour operators.

o Couples represented 46% of the sample, travelling alone represented 30% and group of three represented 14%. The goal was to obtain one survey per travel group, however, on several occasions ‘couples’ participated in the survey separately. This most likely contributed to the inflated sampling of couples. Nonetheless, their stated choice data remains valid.

o As for length of stay, the sample median was 4.5 nights, while the average was 5.4 nights. This is consistent with the 4.8 average nights found in the published sources (e.g. TRA 2008). Over 92% of the sample recorded trip durations less than 11 nights. Finally, 59% of the sample was repeat travellers, and 41% was first time visitors to Cairns. Unfortunately, the data on repeat visitation for domestic travellers are not available to compare.

While the author is not claiming the data to be statistically representative of Cairns’ entire visitor population, the data collected are demonstrated to be consistent with the best available secondary data on the region.

5-19 Figure 5.2 shows that the option to ‘hire a rental car and take a overnight trip’ (41.5%) is the most popular choice, followed by ‘staying in Cairns’ (25.6%). There appears to be a small market for public transport with a choice share of 13.3% for both overnight and day-trips. The surprising result was the little choice preferences for ‘organised day-tours’, reflecting the potential cannibalisation of this market when leisure-purpose-built public transport alternatives are introduced. By the same token, the small sample choice shares of the ‘Base-camp PB’ and ‘Base-camp Tour’ alternatives suggest that the interpretation of the results on these alternatives should be undertaken with caution. Consequently, the discussion in this Chapter focuses mostly on the alternatives with more statistically reliable results such as ‘Dispersal RC’ and ‘Gateway/GBR’. In aggregate, the distributional patterns across choice alternatives in the North and South are similar. However, Public Bus and day-trips are more popular for travel to the South.

Figure 5-2 Sample choice shares across alternatives

5-20 5.6.2 Model results

Various model specifications and nested logit structures were applied to the choice alternatives above. However, the evidence from these models provided support for the use of a multinomial logit model with each of the ‘trip-structure’ – ‘travel mode’ combination as an independent alternative. This is discussed later in this section. Table 3 shows the MNL model performance indicators. The model fits slightly better for the ‘trip to the North’ scenarios than the South.

Table 5.3 Model summary North South Adjusted pseudo R^2 0.267 0.239 Log likelihood (model) -505.8844 -521.2965 No coefficient LL -702.3697 -702.3697

The model results are shown below (Table 5.4 and Table 5.5). ‘Organised all- inclusive tour’ (Tour) was estimated without the alternative-specific-constant (the Tour option was the base alternative). There are two models – one for the North and one for the South – shown in Table 5.4 and Table 5.5 respectively. Some variables were excluded during the modelling process because they were not statistically significant across all alternatives. The coefficients represent the marginal effect of a variable on an alternative’s level of utility (see Equation 1). For example, the coefficient value of -0.88 on ‘PBD $80’ variable in Table 5.4 shows that the utility from choosing ‘overnight trip beyond Cairns on public bus’ decreases by 0.88 unit of utility (‘utils’) when public bus fare to travel to the Northern destinations is $80. The actual trip characteristics, which were also collected from the survey, are dummy coded. Thus, the base case is shown in brackets, e.g. ‘Repeat visit’ (first-time). Interpretation of coefficients is similar to the travel mode attributes. Thus, -0.629 on ‘RCD repeat’ variable shows that repeat visitors obtain less utility from choosing ‘overnight trip beyond Cairns on a rental vehicle’ by 0.629 than first-time visitors. The primary interest here is in finding the significant factors, and the extent to which these factors may affect

5-21 relative utility levels. Thus, the discussion on actual probability values and predicted choice shares is omitted.

Table 5.4 Model output: North

Trip to the Northern region Variables Coefficient P-value Variables Coefficient P-value Constants Repeat visit (base: first time) RCD 1.263 ** RCD repeat -0.629 *** PBD -1.795 RCB -1.156 Accommodation type (base: 'all other') PBB -1.329 * RCD resort 0.247 Gateway 2.090 *** RCD HMA -0.078 RCD FRP -1.406 *** Price PBD resort 2.166 ** PBD $80 -0.880 *** PBD CNC 3.841 *** PBD $40 0.154 PBD HMA 2.463 ** Tour $200 -0.939 * Tour $150 0.070 Travel party # (base: travelling alone) PB two adults 1.067 ** Time PB three or four adults RCD 3 hours 0.075 0.348 RCD 2 hours -0.288 * Age group # (base: 18-25) Destination Expenditure ^ PB 26-35 -1.168 ** Overnight trip in the North $220 PB 36-45 -0.275 -0.474 *** PB 46-55 -0.449 Overnight trip in the North $170 PB 56-65 0.101 0.220 PB over 65 -0.409

Comfort RCD comfort 0.201 *** RCB comfort 0.306 ***

Note: Asterisk [*] indicates asymptotic t-test significance at 10%, [**] 5%, [***] 1%. [^] indicates generic parameter within trip structure (e.g. for overnight trip regardless of travel mode). Please refer to Table 5.3 for more details on generic parameters. [#] indicates generic parameter within travel modes (e.g. the coefficient on PB means that the coefficient is equal for both PBD and PBB). Abbreviations used: hotels, motels and serviced apartments (HMA); friends and relatives’ property (FRP); camping and caravan parks (CNC).

5-22 Table 5.5 Model output: South

Trip to the Southern region Trip to the Southern region (cont…) Variables Coefficient Variables Coefficient Constants Repeat visit (base: first time) RCD 1.13 * RCD repeat -0.56 ** PBD -0.32 RCB repeat -0.83 *** RCB 0.50 PBB -1.02 Accommodation type (base: 'all other') Gateway 2.19 *** PBD CNC 2.90 *** PBD HMA 0.79 Price PBD FRP 1.48 PBD $80 -0.50 * PBD resort -0.11 PBD $40 -0.26 PBB $80 -1.39 ** Travel party # (base: travelling alone) PBB $40 -0.14 RCD two adults 0.46 * RCD three or four adults Driver knowledge and friendliness 0.08 Above expectation (PBD) RCD more than four adults 0.48 ** 3.40 *** As expected (PBD) PBB two adults 1.42 ** -0.47 * PBB three or four adults 1.09 ** Sightseeing (number of stopovers) More than two stopovers -0.60 ** Length of stay (base: 1-3 nights) One or two stopovers RCD 4-6 nights 0.32 ** 0.61 ** RCD 7-10 nights 0.07 RCD 11 or over 0.42 ** Comfort PBD 4-6 nights 1.02 *** RCD comfort 0.14 * PBD 7-10 nights 0.40 ** RCB comfort 0.14 PBD 11 or over 0.17 RCB 4-6 nights 0.50 *** RCB 7-10 nights 0.19 Destination Expenditure ^ RCB 11 or over 0.14 Overnight trip in the South $170 -0.26 * Age group # (base: 18-25) Overnight trip in the South $120 PB 26-35 -0.19 -0.22 PB 36-45 -0.32 Day-trip in the South $170 PB 46-55 -0.91 * -0.55 *** PB 56-65 -2.69 ** Day-trip in the South $120 PB over 65 0.71 -0.07

5-23 Two tests described in Chapter 1 were applied to the Cairns data. The Hausman- McFadden test was applied in the following way. For models ‘North’ and ‘South’, two most prominent alternatives (in terms of sample choice shares) were removed from unrestricted models. Thus, four tests were conducted in total: a restricted model without the ‘rental car overnight alternative’ (one for north and one for south); a restricted model without the ‘public bus overnight alternative’ (one for north and one for south). The tests revealed that when the rental car ‘overnight’ alternative was removed from the North and the South model, evidence to reject the IIA assumption was insufficient (Hausmand and McFadden statistics of -18.4 and -12.53 respectively). As for the public bus alternative, the North model violated the IIA assumption at the level of 1% significance, whereas the South model did not. The reliability of the Hausman-McFadden test has been called into question for relatively small sample sizes (Fry and Harris, 1996). Given the small choice shares of public bus alternatives in the sample, it is appropriate that further tests are applied.

The second IIA test applied was the IV test. Table 6 shows the inclusive value parameters (IV parameters) of the model nested in travel mode and that nested in trip structure. The nested logit models were specified as per the models that generated the results in Table 5.4 and Table 5.5. The IV parameter estimation results are either statistically insignificant from ‘0’ or ‘1’, or they exceed the value of ‘1’. The latter case violates the utility maximisation assumption that underpins discrete choice analysis, whereas values of 0 or 1 indicate that the specified nest is not significant statistically (Hensher et.al. 2005). In particular, given the incidences of the statistically equivalent value of ‘1’ in these models (IVRC and IVPB in the travel mode nest, and IV day-trips for both North and South in trip structure nest), there is evidence that the nested model collapses to a simple multinominal logit model. Thus, this simplifies our modelling task to a situation where each of the ‘trip structure’ – ‘travel mode’ combinations is an independent alternative uncorrelated in their stochastic utilities of one another.

5-24

Table 5.6 Inclusive value (IV) parameters

North South Travel mode nest IV RC 1.1 ** 0.17 IV PB 11.6 1.2 *** Gateway 11 Trip structure nest IV overnight 7.5 2.6 ** IV day-trip 1.6 ** 0.997 *** Gateway 11

Note: [*] indicates not statistically different from ‘1’ at 10% level, [**] 5% and [***] 1%.

The following discussion concentrates mostly on the overnight trips of tourists and the significant travel mode attributes associated with the overnight trips. Overnight trips typically inject greater expenditures into peripheral destinations; thus, this type of trip may be of most interest to them. Overnight trips were also the most popular choices (see Figure 5.2) and consequently less subjected to the problems associated with low choice samples.

5.7 Dispersal and rental cars

5.7.1 Transport attributes

Destination expenditures and perceived comfort exert significant influence on the choice of RCD in both destination contexts. The perceived comfort of the rental vehicle is a strong source of utility (the coefficient indicates a change in one unit in the Likert scale). In fact, it can be concluded that perceived comfort is one of the most important reason why a car is a popular choice, supporting the Anable and Gatersleben (2005) study that has shown the importance of affective factors (such as ‘freedom’ and ‘control’) of a car over public transport. Furthermore, the flexibility the rental vehicles offer (much in the same vein as the private vehicle),

5-25 is a significant factor that is difficult to be replaced by other modes. The significance of ‘comfort’ reflects these qualities. This supports Eaton and Holding (1996) who argued that the popularity of the car cannot be replaced by other modes, especially when the travel modes are compared against the same attributes. Rather, they argued, other modes must capitalise on what the private vehicle cannot offer.

Destination expenditures influence tourists’ choice of modes and trip structure. Tourists were unresponsive to price and travel time attributes of rental cars, at least when compared with the utility gained from qualitative (and affective) features such as comfort. One key feature of rental vehicles is that the cost per head decreases up to the vehicle’s capacity limit, diminishing the importance of price as travel party size increases. This helps to explain why the price of travel mode is not significant but the price of destination is significant (destination expenditure variables). Destination expenditures are typically greater than expenditures on transport, and this renders the responsiveness to destination expenditures greater. In addition, the rental vehicle rate was indicated in the survey as ‘per day’ cost, whereas the destination expenditure was ‘per person per day’ cost.

5.7.2 Trip characteristics

Repeat visitors to the region are less likely to choose RCD and RCB. In both destination contexts, the magnitude of the negative effect of repeat visitation (- 0.56 in the South model) is strong enough to offset the utility gained from savings in destination expenditure (+0.48 utility earned by saving $100 in expenditures (going from $170 to $70 per day)). Thus, it is more difficult to entice repeat travellers to choose RCD or RCB option with control variables such as price, than it is for first time visitors. The result suggests that first time visitors are more likely to use rental cars for dispersal, while repeat visitors are less likely to do so, possibly because the repeat visitors’ greater destination familiarity enables them to exploit other alternatives. As discussed in Chapter 3, Li et al. (2008) noted that

5-26 first time visitors are more extensive in their destination exploration, while repeat visitors are more intensive in their use of time across a smaller range of destinations and activities. Further, it has been suggested that “the more familiar the tourist is with the location, the more knowledge one has of different kinds of local activities and attractions to fill an entire trip schedule” (Hwang et.al. 2006: 1060), which renders repeat visitors more specific in the activities pursued, but also less explorative, diminishing the need for a travel mode that provides this capacity for the visitor.

Individual trip characteristics are significant constraints for the choice of RCDS alternative (superscript denoting ‘South’). The utility functions differ for the two destination contexts in two ways. First is that the attribute coefficients, such as ‘destination expenditure’ has less influence on the trip to the South than to the North. Second, trip characteristics such as length of stay and travel party size exert significant influence on the choice of RCDS but not for RCDN. This is an interesting finding given that these two trip characteristics are important determinants of multi-destination travels and dispersal. The utility from choosing RCDS increases as travel party size increases; for instance, compared to solo traveller, couples yield statistically significant 0.46 utils, three or four adults yield 0.08 (but not statistically significant), and more than four adults yield statistically significant 3.4 utils. The utility from choosing PBD increases as length of the trip increases; for instance, compared to a trip between 1-3 nights, a trip between 4-6 nights yields statistically significant 1.02 utils, while a trip with 7-10 nights yields a statistically significant 0.4 utils.

Length of stay is positively related to greater dispersal and multi-destination travel (“when time is short, space is conserved” - Fennell 1996). Greater travel party size indicates heterogeneity in preferences, which results in greater need to visit multiple places (Tideswell and Faulkner 1999), and by implication, greater need to be more spatially expansive and disperse. In other words, a trip to the South becomes more likely only when there is sufficient time and preference heterogeneity in the travelling group, reflecting the fact that the South is less

5-27 popular and known to the tourists. Importantly, both variables are in many instances determined prior to the arrival, thus this result shows the relative ineffectiveness of destination control variables, e.g. price, for dispersal to the South.

5.8 Dispersal and public transport

5.8.1 Transport attributes

There are significant differences in the factors that determine PBDN with PBDS. A key finding is that a choice of PBDN and PBDS is associated with a different responsiveness to different public transport attributes. The PBDS alternative is determined by the qualitative attributes of public bus, as well as price, whereas only price matters for the choice of PBDN. A high level of ‘driver knowledge and friendliness’ and ‘1 or 2 stopover for sightseeing’ are qualitative features of public bus design that may contribute to greater rider-ship, but only for trips to the South. For this alternative, the disutility of price (PBD $80 coefficient of -0.5) can be almost completely offset by offering good driver service (‘above expectation’ yields 0.48 unit of utility) or more than offset by a stopover opportunity en route for sightseeing (‘one or two stopovers’ yields 0.61). The combined offering of two attributes will increase the utility of PBD to go to the South by 1.09 units (0.48 + 0.61) (or even more if interaction effects are present).

The differences in the utility functions of alternatives between destination contexts can be attributed to two factors. First is the relatively unknown status of the South compared with the North. Thus, qualitative attributes of public transport services are important for tourists with little familiarity and knowledge of the destination. The second explanation refers to market segments. Lumsdon (2006) described two market segments for public bus services: ‘sightseeing’, and ‘activity seekers’. The study argued that the latter will be much less concerned

5-28 with the ‘transport as tourism’ aspect of the trip; rather, this group will use the bus purely as a vehicle to travel between origin and destination in pursuit of their sought activities, or in Lumsdon and Page’s (2004) terms, ‘transport for tourism’. The significant utility gained from the qualitative attributes indicates that sightseeing tourists may be the primary source of demand for the South. Southern destinations may generate demand from the sightseeing tourists because it is an unfamiliar destination.

5.8.2 Trip characteristics

Trip characteristics have statistically significant influences on the choice of the PBDS alternative but not on the choice of the PBDN alternative. Greater travel party heterogeneity and length of stay positively influence the choice of PBDS. The coefficients are large relative to transport modal attributes, implying that attractive travel mode attributes themselves may not be sufficient to compensate for pre-determined trip characteristics such as short length of stay. Generally, ‘camping and caravan’ (CNC) is associated with the greater use of public transport service. Furthermore, tourists with ‘friends and relatives property’ (FRP) as their main accommodation, were observed to be public transport averse, presumably because family and friends are able to provide the necessary mobility in the destination.

Travel decisions to the South depend on the length of stay. Greater length of stay tends to promote overnight trips as well as day-trips, which is expected given the positive relationship between length of stay and dispersal. However, this is not the case for the North, where length of stay was found statistically insignificant in all alternatives (subsequently dropped from the model). This reflects the fact that the northern region is a prime attractor of tourists to Cairns and TNQ. The effect of length of stay (less than 4 nights in this study) may not be an important variable for many of the well-known regions because these destinations are often the main reason for the trip to Cairns. However, for a relatively unknown periphery, length of stay is an important determinant. This is not surprising because the southern

5-29 destinations will be ranked lower in the tourists’ priority list, which will be considered for a visit when the utility from visiting the primary destinations has been fulfilled. One implication is that the trend of short-frequent break will not contribute dispersal to the peripheral destinations in the South and alike. This has important ramifications for the dispersal of LCC-induced tourists because the LCCs have been observed to be associated with short-frequent breaks.

5.9 Limitations and future research

The time and frequency variables were statistically insignificant in this study. The determining power of travel time in travel mode choice is significant in the context of journey-to-work (JTW) trips (e.g. Redmond and Mokhtarian 2001) and in long-distance inter-regional trips (e.g. Hensher 1997, Koppelman and Sethi 2005). This insignificant result may be a reflection of the relatively time- insensitive nature of leisure travellers, in particular when the range of travel time examined was between one to three hours. The result implies that peripheral destinations are not significantly disadvantaged by the fact that their destinations are an hour further from another destination. In fact, the evidence supports Page’s (1994) argument that in tourism, transport is not only a cost to be minimised, but also an integral part of tourists’ overall travel experience. An extension of this argument is a possibility of positive utility attached to travel time, in which case we should not observe a significant negative relationship between utility and travel time. The positive utility in travel time is illustrated in the intra-mode JTW trips; for example, Redmond and Mokhtarian (2001) show that commuters prefer a short commuting time than none. Perhaps future studies can apply a similar approach to a finer market segment in order to isolate the positive and the negative effect of time on utility.

5-30 Overall, the finding on the time variable is in-line with the qualitative work of Lumdson (2006) and Eaton and Holding (1996) reviewed earlier. In both studies, in-vehicle travel time is not mentioned as a key determinant for the demand of public transport in the context of recreational trips. Nonetheless, the importance of the frequency attribute is noted in their studies. Surprisingly, this research found no significant effect of frequency on mode choice. This potentially illustrates one important issue with stated choice experiments. The choice scenarios are formulated with pre-determined set of attributes that describe a choice alternative, which cannot be exhaustive for practical reasons. Thus, attribute specification must be parsimonious. While frequency is an important attribute, from respondents’ viewpoint, this may be a proxy for a more salient and ambiguous feature such as ‘convenience’. Specifying a ‘convenience’ attribute that summarises frequency, as well as features such as schedules and reliability, may yield a different outcome. Such specification should be explored in tourism problems in the future.

Finally, while this study provides insights into dispersal and travel mode choice behaviour of the air arrivals, the results and conclusion cannot be extended to the behaviour of some market segments in the Cairns region. For instance, campervans and backpacker segments were not explicitly considered in this study. The backpacker segment is related to the high level of international visitors in the Cairns region, which highlights another limitation of this study - that only domestic visitors’ dispersal and mode choice behaviour were considered. Furthermore, due to resource constraints, the survey could be carried out over a limited period. While the survey period has been carefully selected (e.g., avoiding special events, etc.), the author acknowledges that the short survey period imposes some limitations on the findings. In general, the samples collected are more representative of the behaviour of tourists during peak-holiday season than off- peak. Thus, in situations of excess capacity, the behaviour of tourists is likely to be different from that observed in this Chapter.

5-31

5.10 Conclusion

The aim of this Chapter was to provide insight into the likelihood of destination transportation policy to stimulate dispersal of the air arrivals, even in situations where the air arrivals exhibit trip characteristics that may be dispersal averse. The use of stated choice data and the application of choice modelling provide the ceteris paribus effects of attributes (both actual and hypothetical attributes) and trip characteristics on choice. This allows a direct comparison of transport attributes and trip characteristics from a utility compensation perspective. This study has shown that appropriate ground travel mode attributes can offset some or all of the negative effects of trip characteristics on tourists’ dispersal propensity. However, the extent to which this is feasible depends on the destination contexts. Dispersal to the North is easy to entice because northern destinations are one of the primary reasons why travellers fly to Cairns in the first place. However, this is not the case for the southern destinations.

One significant outcome from this study was the importance of trip characteristics on dispersal to the southern destinations. The relative importance of trip characteristics compared with the coefficients of modal attributes was very strong, indicating that individual trip characteristics are binding constraints to dispersal to the South. The length of stay and travel party size variables were constraints that tended to reduce the propensity of air arrivals in Cairns from dispersing to the southern destinations. Hence, ground transport is of little effect in promoting dispersal of the air arrivals to the South because trip characteristics are in many instances pre-determined.

For those choosing rental cars, perceived ‘comfort’ is the primary source of utility for using this mode for dispersal. Thus, the quantitative attributes such as price and time are relatively ineffective in contrast to the subjective and more

5-32 qualitative elements. As expected, there was a strong relationship between a car and dispersal. This relationship was evident in both destination contexts. However, destination context changed the relationship between dispersal and public transport markedly. The clear difference was that the travel to the northern region was related to the functional elements of the public bus alternative such as price, whereas the South emphasised the qualitative attributes such as adequate ‘stopovers for sightseeing’ and good ‘driver knowledge and friendliness’. This is in part a reflection of the ‘sightseeing’ market characteristics to the South, which is related to the fact that tourists are generally less familiar with the South.

The findings are relevant for destination managers and policy makers. Firstly, destination transport policy aimed at assisting dispersal must be devised upon adequate assessments of the factors that constrain tourists’ travel. Specifically, this study provided some evidence supporting the attractiveness of qualitative attributes of public bus services, and importantly, demonstrated how the effectiveness of such design differs across destinations. Public transport is often an important component in the pursuit of environmental objective by government. This research has generated empirical evidence highlighting the importance of weighing up tourism and regional dispersal implications of public transport policy. Although the data examined in this Chapter were collected in the Tropical Northern part of Australia, this research should be of relevance to many regions interested in understanding the relationship between destination transport and spatial behaviour of the air arrivals, which experienced vast growth in the recent years due to the advent of low-cost carriers.

5-33

Appendix 5.1

5-34

6. THE BALLINA-BYRON EXPERIMENT

6.1 Introduction

The emergence of LCCs has improved air travel access to regions outside the capital cities in Australia by offering discounted tickets and non-stop services from key domestic origin markets. By the same token, it has also increased the competitiveness of air travel against other modes of travel in regions traditionally reliant on ground modes. Recent research by Whyte and Prideaux (2007) in North Queensland (Australia) has shown the relative decline of car and long-distance coach travel between 2001 and 2005, while air travel increased in the same period largely marked by the proliferation of two Australian LCCs (Virgin Blue and Jetstar). As a result, tourism businesses located between tourism generating regions and regional destinations experienced declines in visitation (Whyte and Prideaux 2007).

In Australia, car is the dominant travel mode used for visiting rural regions (TTF 2002). The car allows travellers the flexibility to establish their own travel itinerary (Taplin and McGinley 2000), whilst air travel often does not offer the same flexibility and spontaneity in the choice of travel routes (Stewart and Vogt 1997). Consequently, travel mode is an important means by which the different levels of spatial ‘degrees of freedom’ for tourists are achieved (Lew and McKercher 2006). In fact, recent research has shown that the spatial pattern of travel and travel mode used are related to the travel experience sought. Moscardo and Pearce (2004) studied the moderating role of lifecycle factors in the choice of long-distance mode of travel, and found that self-drive tourists are considerably

6-1 different from non-self drive tourists in the travel experience sought in the North Queensland Region. In particular, the study found that self-drive tourists tend to place more importance on visiting rural communities than other travellers.

There is a potential conflict between the increasing use of air travel and dispersal. This is because dispersal typically requires a high degree of mobility, which can be most easily met by using the car, but is most difficult to meet by air transport. Conversely, according to the law of demand in microeconomic theory, the improved affordability of airfares is a potent force in increasing the demand for air travel. Specifically, the objective of this Chapter is to examine the proposition that LCC proliferation adversely affects regional dispersal. This shall be approached via the analysis of the trade-offs involved in leisure travellers’ travel mode choice decisions. This Chapter accomplishes the final specific aim of this thesis (A5), which is to examine inter-regional travel mode substitution as a source of conflict between low fare air services and regional dispersal by applying a stated choice experiment.

6.2 Tourists’ dispersal

Australia’s national tourism organization, Tourism Australia, uses the definition of ‘regional dispersal’ as trips originating in State and Territory capital cities into destinations other than these cities and the Gold Coast. In this chapter the regions are dichotomised into ‘gateways’ or ‘periphery’. Lew and McKercher (2002) defined gateways as the first destination of overnight stay in the trip, which can be either a point of entry or the main destination itself. In Australia, the gateways are almost always the largest townships of the tourism-regions. For the purpose of this research, a single destination trip is defined as a trip that involves a stay only in one gateway, whereas a multi-destination trip involves at least one overnight

6-2 stay in the gateway and one in the periphery. The cases in which a trip involves stopovers on more than one gateway are not considered in this research.

Dispersal is achieved when many destinations are visited within the same trip, or when a unique trip is undertaken in many parts of the destination (Wu and Carson 2008). From the viewpoint of individual preferences, it is possible for there to be as many variations in spatial behaviour of tourists at the destination and in the region surrounding the destination, as there are individuals travelling. Lue, Crompton and Fesenmaier (1993) conceptualised the variation in the patterns of trip itinerary into five basic patterns of multi-destination trips. Oppermann (1995) developed this further into two single-destination and five multi-destination trips. The multi-destination trip patterns identified have been applied to differing contexts by researchers; on a domestic-regional level (Stewart and Vogt 1997), to travel by international tourists (Tideswell and Faulkner 1999), as well as inter- continental travel (Lew and McKercher 2002). Some of the common trips featured in these studies that are relevant to this research are the patterns of ‘regional tour’ and ‘en route’ travels (Figure 6.1). In this research, a single- destination trip refers to a trip that only involves an overnight stay in the ‘gateway’ (denoted ‘D’), while multi-destination trips involve overnight stops in at least two different destinations, one of which is the gateway.

Regional tour/partial Combined en route and orbit regional tour e e d d En route f f D D D t c b c b a a

HOME HOME

Figure 6.1 Patterns of multi-destination travel (modified from Lue et.al. 1993 and Oppermann 1995) 6-3

The consequences of modal substitution towards air travel can be detrimental to the peripheral destinations. Substitution away from ground modes implies bypassing smaller destinations located between major origin markets and popular domestic leisure destinations. A destination such as Port Macquarie, a seaside town located between Sydney and Byron-Ballina in New South Wales, is an examplei. This relationship can be seen in Figure 1. In the ‘combined en route and regional tour’ diagram, ti represents the transport linkage between home and destination, and the subscript (i) represents the available travel mode on this link, such as car or air. If substitution occurs toward air travel due to low fares, then the smaller destinations ‘a’, ‘b’ and ‘c’ will be bypassed, with the only possibility of visitation conceivable when the traveller travels back from ‘D’.

Modal substitution is not the only channel of influence of affordable air travel on dispersal. If the cheap and direct flights stimulate a greater number of tourists to ‘D’ then this increases the pool of tourists that may travel further to the peripheral destinations of ‘d’, ‘e’ and ‘f’. In some circumstances, even the destinations ‘a’, ‘b’ and ‘c’ may experience an increase in visitations from the travellers flying into ‘D’. This may occur when the return route or mode is different from that used for access, such as when the tourist uses a car to travel back ‘home’, or when the air arrivals take day-trips from ‘D’ to the surrounding periphery using local transport. It is acknowledged that these sources of change in spatial patterns have important implications for the evaluation of the net effect of affordable air travel on dispersal. The two sources outlined above, however, were not considered in this study because it was assumed that the majority of travellers on the corridor use the same mode to travel both ways. Second, day-trips from ‘D’ represent a ‘base- camp’ pattern, which does not constitute the ‘dispersal’ defined in this study. The primary focus of this research is on the effect of modal substitution on regional destinations, e.g. ‘a’, ‘b’ and ‘c’. As explained below, the decision by tourists to disperse to ‘d’, ‘e’ and ‘f’ is viewed as an exogenous factor that this research controls using a stated choice experiment.

6-4

Tourists’ travel mode choice on each leg of the journey does not occur in isolation; rather, it is influenced by the entire trip and the context in which travel decisions are made (Page 2005). Thus, in light of the ‘combined en route and regional tour’ diagram in Figure 1, while leisure tourists’ long-distance travel mode choice applies only to the ti segment of the journey, the decision of whether or not the tourists’ trips involve dispersal to ‘d’, ‘e’, ‘f’ will affect the mode choice on ti. For instance, on distances where ground modes compete with air travel, a possible scenario is that if the tourist’s itinerary includes a visit to ‘d’ then driving the entire trip may become more attractive than when the tourist only requires a trip to ‘D’. Subsequently, a tourist may decide to make this switch in travel mode. This implies a linkage between the destinations ‘d’, ‘e’, ‘f’ and ‘a’, ‘b’, ‘c’, because driving the entire distance inadvertently provides opportunities for en route visitations along ti. In contrast, flying will preclude this possibility, resulting in a complete bypass (corridor effect) unless some form of vehicle is used to travel back down to ‘c’ from the gateway (D). In this chapter, we examine the effect of multi-destination trips on mode choice, i.e. the effect of trips with and without visits to ‘d’, ‘e’, or ‘f’, on mode choices along ti.

6.3 The model

Similar to Chapter 5, the MNL model is applied in this study. Please see Chapter 1 for details on discrete choice models. This Chapter examines the factors affecting mode choice in differing trip contexts, e.g. single-destination vs. multi- destination. The experimental design used in this chapter enables the estimation of the mode choice model for each trip context separately (i.e., two separate equations), as well as in a single equation that includes both contexts. For the

6-5 former, the following utility function is estimated for each mode of transport in each trip context.

= + + Vni i i X ni iZni Eq. (2)

Vni is the level of utility for individual n choosing alternative i. Vni is a function of the levels of the attributes X ni where i is a vector of coefficients to be estimated for each attribute of each alternative i. Zni is the individual’s characteristics with coefficients vector i. As for the single equation approach, Oppewal and Timmermans (1991) have shown that the following utility function can be estimated given an appropriate experimental design:

= + + + + Vni i d i i X ni d i X ni iZni Eq. (3)

The additional term in Eq (3) is d , which is a dummy term that takes the value of ‘0’ when the choice is made under a ‘single destination trip context’ and ‘1’ when the trip is ‘multi-destination’. d interacts with the alternative specific constants ( i) and the alternative specific attributes of travel modes ( i X ni). The latter enables, in a single model, the estimation of separate coefficient for each trip context of the same attribute. Both models were applied in this study.

6.4 Data

6.4.1 Case study region

The data collection regions were Ballina and Byron in the Northern Rivers tourism region of New South Wales, Australia (Figure 2). Byron is a popular

6-6 seaside leisure destination, where 22% of total trips originate from Sydney and 26% from Brisbane (TRA 2008)ii. The Ballina-Byron airport is located in Ballina, which is a 25 minute drive from Byron. The leisure travelers (holiday and ‘visiting friends and relatives’ travel purpose) on the corridor from Sydney to Byron were chosen as study subjects for two main reasons. First, two LCCs, Virgin Blue and Jetstar, commenced services to the Ballina-Byron airport introducing low fares and greater ticket discounting practices. Thus, it was expected that travelers on this route are familiar with the air travel alternatives and the low fares frequently advertised. Second, the corridor is approximately 800km, a distance sufficient for competition to prevail between private car, coach, rail and air travel.

6-7

QUEENSLAND Gold Coast (850km OR 10 hours drive from Sydney)

Byron Ballina (800km OR NORTHERN 8.5 hours drive RIVERS from Sydney) NEW ENGLAND TOURISM TOURISM REGION REGION

Coffs Harbour (550 NEW SOUTH km OR 6 hours drive from Sydney) WALES

NORTH COAST TOURISM REGION Port Macquarie (380km OR 4.5 hours drive from Sydney)

HUNTER TOURISM REGION Main Highway (Train runs roughly parallel)

Sydney

Figure 6.2. Northern New South Wales Coast (Source: drawn by the author based on ‘Tourism Regions classification’ of New South Wales State Tourism Organisation)

6-8 6.4.2 Stated choice data

Econometric models often use data collected on choices already made in the market, commonly referred to as ‘revealed preference’ data. Revealed preference data suffer from a lack of variation in the levels of explanatory variables and difficulties in observing the alternatives actually considered by the decision maker (Hensher et.al. 2005). ‘Stated choice’ data on the other hand, involve presenting to a decision maker a combination of alternatives (e.g. flying or driving) and attributes (e.g. price) as hypothetical scenarios (see Figure 5.2). An example of stated choice application on long-distance travel mode choice is the study by Hensher (1997), which used this method to estimate the demand for a then planned high-speed-rail between Sydney and Canberra. More recently in tourism, Crouch et.al. (2007) applied the stated choice method to examine preferences in the allocation of discretionary expenditure on domestic tourism against alternatives such as reducing household debts and overseas holiday, while Huybers (2002) and Huybers (2003) applied this method to the short-break destination choice of Sydney and Melbourne residents.

The stated choice method was used in this research for several reasons. First, stated choice method is an experiment that manipulates the control variables. For example, airfares are systematically varied across the choice alternatives so that their influence on respondent’s choice of travel mode can be estimated in a controlled environment. This approach overcomes the pitfalls in the revealed preference data such as lack of variation in the levels of variables (Louviere et.al. 2000). Additionally, alternatives considered and the prices paid by tourists are information often not readily available in secondary data sources or in the form of revealed preference data. Finally, this method allows the analyst to vary other aspects of the trip so as to answer a question central to this chapter: “How would you change your current choice had your trip involved a stay at least two hours drive away from the main town centre?” This allowed the researchers to estimate the effect of change in trip context on travel mode choice in a controlled environment.

6-9

By applying the stated choice framework, we are able to estimate the extent to which each factor influences travel mode choice. The controlled factors are travel mode attributes (e.g. airfare) and trip characteristics (or trip context) (multi- destination vs. single-destination trip). The stated choice method is particularly appropriate when the study is interested in the willingness-to-pay and trade-offs among choice alternatives, rather than market share predictions (Hensher et.al. 2005). Since the objective of this chapter is to extract the trade-offs between modal specific attributes (e.g. price) and trip context (single destination vs. multi- destination), stated choice data were chosen. The following sections on research methodology outline the discrete choice model, the data collection region, choice alternatives, attributes considered, and experimental design for the stated choice survey.

6.4.3 Choice alternatives

The feasible set of alternatives for this study included Car, Rental Car, Bus/Coach, Train, Virgin Blue (DJ), Jetstar (JQ), Regional Express (REX), and flights to Gold Coast airport. Technically, transport to Gold Coast airport is not an independent mode; rather, it represents an alternative route. The decision to include flights to the Gold Coast was made in consultation with local industry practitioners and researchers. Gold Coast airport is only one hour driving distance from Byron and there are high levels of air service frequencies to the Gold Coast compared to only daily services on the route between Ballina-Byron and Sydney. Thus, withdrawing this alternative would exclude a prominent form of competing air transport to Ballina-Byron. Whilst trains no longer operate directly to Byron, the inclusion in this study does not pose a problem. In fact, the ability to account for an unavailable mode is an important advantage of stated choice experiments, applied previously in studies examining the viability of currently unavailable alternatives (e.g. Hensher 1997).

6-10

6.5 Attributes of modal alternatives

Modal attributes in the model were based on the literature review of inter-regional mode choice studies. This research aimed to provide a comprehensive specification of modal attributes recognising that under-specified models will increase the likelihood of violating the identical and independent distribution (IID) assumption of the error terms in MNL models (Louviere et.al. 2000; Hensher et.al. 2005). Consequently, attribute specification was based on a literature review of modal attributes not only on inter-regional mode choice, but a wider survey of the literature including those studies that examined the importance of ‘qualitative’ variables such as road conditions, safety, schedules and delay risks for public transport alternatives.

Service qualities are generally more difficult to account for in models because of their subjective nature (Hensher et.al. 2005). Service ‘convenience’ is often associated with service schedule and frequency in the travel mode choice literature. Frequency of the transport service, as with price and time, frequently appears in the attribute specification and is easily quantified. For example, Koppelman and Sethi (2005) used a schedule convenience attribute that included arrival and departure time of the day as dummy variables, as well as a measure of the reliability of the transport service by incorporating an ‘unreasonable delay’ dummy variable.

The nature of the qualitative variables is likely to differ for each mode. On non- urban driving it was found that, in Australia, the top three issues for the regional motorists were: behaviour of other drivers, condition of roads, and safety and

6-11 accidents (ANOP Research Services, 2005). Hence, the model specification for a car alternative should include road quality and safety variables. Previous studies such as Greene and Hensher (2003), in specifying the stated choice experiment attributes for road types in long-distance travel, used attributes such as number of lanes, the existence of median strips and percentage of free flow time etc. On road safety and risk, Rizzi and Ortuzar (2003) investigated the impact of perceived road risk on route choice for inter-urban trips using the yearly fatal accident rate on the given route.

In regard to attribute levels, most of the attribute level labels were based on real market information so that the designed choice scenarios were as realistic as possible. The attributes and attribute level labels are explained in detail below, and summarised in Table 6.1.

Price The prices for air transport mode were obtained from Jetstar, VirginBlue and Regional Express websites on the 17th of November 2006 for the period between 18th of November and the 25th of January; and again on the 27th of December 2006 for the period between 28th of December 2006 and 29th of January 2007. Based on the published fares in the period above, this experiment controlled for three levels of air ticket price: $80; $150, and $220. $80 was one of the lowest fares available in that period, and $220 was the highest. Similarly, the train and coach prices were based on the published fares on company websites (Countrylink, Greyhound and McCafferty). The price attribute level labels for train and coach were $60, $120 and $180. Finally, rental rate per day was specified in the model for the rental car alternative. As per the other alternatives, the labels were based on real market price of several rental car companies in Ballina-Byron. These were $30, $60 and $90 per day.

In Australia, more than three-quarter of the motorists have a good idea of the petrol price at a given point in time (ANOP, 2005). Therefore, it was viewed that the fuel price per litre was an appropriate measure of the motorists’ perception of

6-12 the price of travel on car. Fuel price ranges were obtained from the Australian Automobile Association monthly average fuel prices for Sydney Metropolitan Area between December 1998 and December 2006. The prices fluctuated around $1.10/litre. Based on the 1998-2006 fuel price time series, three fuel price level labels were $0.70, $1.10 and $1.50.

Time The time attribute is in two parts: ‘in-vehicle time’ (IVT) and ‘out-of-vehicle time’ (OVT). The attribute level labels used for all modes are based on published information from airport transfer operators, flight schedules and travel guides. For Jetstar, Virgin Blue, Regional Express and the flight to Gold Coast, the IVT was controlled at the levels of 1 hour, 1.5 hours and 2 hours. For OVT, this varied among 2 hours, 3 hours and 4 hours. For other scheduled transport services such as Coach and Train alternatives, the IVT was varied among 11 hours, 13 hours and 15 hours, whereas the OVT ranged from 1 hour, 3hours and 5 hours. For private and rental car alternatives, combined IVT and OVT were specified. The ‘door-to-door time’ variable had three levels, e.g. 7 hours, 9 hours and 11 hours.

Road risk The Pacific Highway is the major artery that runs for most of the Sydney- Ballina/Byron route and it rates as one of the worst roads in regards to safety and risk (AAA 2005)iii. Road safety and risk was measured by the level of fatal accident rate with the following labels: ‘50% reduction in fatal accidents’, ‘no change’, and ‘50% increase in fatal accidents’. Such approach to road safety and risk in stated choice experiments has been demonstrated in previous studies such as Rizzi and Ortuzar (2003).

Road condition At the time of the survey, 243km of the 618km (40%) of Pacific Highway was in the form of dual divided lanes with a median (RTA 2006)iv. The reminder of the highway was in the form of undivided two or four lanes. However, it was expected that additional sections of the undivided lanes were to be upgraded in the

6-13 following years. The road condition attribute labels were ‘30% upgrade’, ‘60% upgrade’ and ‘90% upgrade’ of the highway.

Reliability Airline on-time performance data are available from the Bureau of Transport and Regional Economics (BTRE) aviation statistics. The figure shows the percentage of airline arriving and/or departing within 15 minutes of scheduled time. REX and Jetstar had a 90% on time performance, while Virgin Blue’s performance was 83% in 2006. In the experiment, the labels were ‘75%’, ‘85%’ and ‘95%’ on-time performance. The reliability attribute was omitted for the coach and train alternatives due to limited data on the actual levels.

Schedules Schedule attributes were labelled according to arrival and departure times. For ‘air’ alternatives, these were departures and arrivals in the ‘morning’, ‘afternoon’, and in the ‘evening’. For train and coach alternatives, the equivalent labels were ‘morning departure and night arrival’, ‘night departure and morning arrival’, and ‘arrival between 12am and 6am’.

Frequency Virgin Blue and Jetstar operate daily services, whereas Regional Express (a regional carrier) alternative and Sydney-Gold Coast alternative operate more frequently. Thus, the attribute level label has been adjusted accordingly. Virgin Blue and Jetstar attribute labels were ‘4 per week’, ‘daily’ and ‘4 per day’, while for Regional Express and Sydney-Gold Coast, the labels were ‘daily’, ‘4 per day’ and ‘10 per day’. At the time of the survey there were between three and four daily coach services on the travel corridor. To reflect this, the experiment controlled coach schedule frequency for ‘daily’, ‘4 per day’ and ‘10 per day’. For the train alternative, the labels were ‘4 per week’, ‘daily’ and ‘4 per day’ to reflect the fact that train services are less frequent than coach services in the current market.

6-14

Table 6.1a Attributes (abbreviation used for model estimation in brackets e.g. (price1))

Ticket price Fuel price In-vehicle time Out-vehicle time Door-Door time Frequency Jetstar 1 hour 2hr 4/week $80 $150 (price1) 1.5 hour (it1) 3hr (ot1) Daily (freq1) - - $220 (price) 2 hour (it) 4hr (ot) 4/day (freq)

Virgin Blue $80 1 hour 2hr 4/week $150 (price1) 1.5 hour (it1) 3hr (ot1) Daily (freq1) - - $220 (price) 2 hour (it) 4hr (ot) 4/day (freq)

Regional Express $80 1 hour 2hr Daily $150 (price1) 1.5 hour (it1) 3hr (ot1) 4/day (freq1) - - $220 (price) 2 hour (it) 4hr (ot) 10/day (freq)

Fly to Gold Coast $80 1 hour 2hr Daily $150 (price1) 1.5 hour (it1) 3hr (ot1) 4/day (freq1) - - $220 (price) 2 hour (it) 4hr (ot) 10/day (freq)

Rental car $0.70/litre 7 hours $1.10/litre (price1) 9 hours (it1) - -- - $1.50/litre (price) 11 hours (it)

Private car $0.70/litre 7 hours $1.10/litre (price1) 9 hours (it1) - -- - $1.50/litre (price) 11 hours (it)

Train $60 11 hours 1 hr 4/week $120 (price1) 13 hours (it1) 3 hrs (ot1) Daily (freq1) - - $180 (price) 15 hours (it) 5 hrs (ot) 4/day (freq)

Coach $60 11 hours 1 hr Daily $120 (price1) 13 hours (it1) 3 hrs (ot1) 4/day (freq1) - - $180 (price) 15 hours (it) 5 hrs (ot) 10/day (freq)

6-15

Table 6.1b Attributes cont. (abbreviation used for model estimation in brackets e.g. (price1))

6-16 6.6 Experimental design and survey

A fractional factorial of the 344 full factorial design was selected for this experiment. This fractional factorial only allowed for the independent estimation of the main effects of each attribute. This orthogonal array provided up to 44 control variables in three levels so that non-linear effects could be estimated. After removing two treatment combinations without designed trade-offs, 106 choice sets were generated with a total of 44 attributes across eight alternatives, and three attribute levels for each attribute (please see Table 1 for each alternative’s attributes). Thus, each attribute of an alternative is orthogonal to all other attributes of that alternative as well as the attributes of all other alternatives. This constituted an orthogonal main effect only design, where the main effects are not protected from potential confoundment with two-way and higher order interaction effects (Louviere et.al. 2000). All alternatives were available in all choice scenarios.

In addition, to test for the effect of trip context on mode choices (single destination trip vs. multi-destination trip), the design was duplicated so the context of a single destination trip and a multi-destination trip could be presented with the exact same design. That is, respondents were asked to make mode choice decisions under scenarios when the trip involves only a single destination and scenarios of multi-destinations. This duplication procedure is in-line with that suggested by Oppewal and Timmermans (1991) for a single equation model with context effects. Thus, a complete design had 212 choice sets (106 multiplied by two) blocked by 53 so that four choice sets (212 divided by 53) were shown to each respondent during the survey.

The survey was undertaken in the main beach area of Byron Bay and at the departure lounge of Ballina-Byron airport, which is the main gateway airport of the region. Simple random sampling strategy was employed. Departing travelers

6-17 were approached in the departure lounge, or while they were in the long queue to the security screening point. For tourists surveyed at the main beach area, each survey distributor approached newly arriving visitors in their allocated area of the beach. The respondents were screened to ensure their trips involved ‘a stay of at least one night in Byron, on a trip purpose other than business or work’. In addition, the visitors had to be permanent residents of Greater Metropolitan Sydney to ensure that all travelers meet the basic choice context, thus excluding those who used Sydney as a transit point. Upon consultation with local tourism research office, the survey was undertaken over the course of eight days between the 20th – 27th of January 2007 with five survey distributors. This period is traditionally the final week of the summer peak in Ballina-Byron. The survey was face-to-face where possible (except in the departure lounge) to assure response quality.

In total, 340 respondents attempted the survey of which 302 were usable for empirical analysis. The survey distributors were asked to keep records of the number of people they approached, and from this we were able to impute that the response rate was approximately 20% for the beach visitors and 10% for the departing visitors at the departure lounge. 80 of the 302 valid samples came from the surveys conducted at the airport. This gave a total sample of 1,202 observations (excluding six missing observations) across 302 individuals. The samples were:

• 49% between the age 18 and 35, which is consistent with the fact that Byron is favored by young travelers as a beach and surfing destination; • age group 36-45 and 46-55 represented 19% and 20% respectively, while only 3.5% was over the age of 65; • gender distribution was slightly skewed towards female (62%); • over 94% of the sample was traveling in a party size of four or less and 29% of the total was traveling alone.

6-18

6.7 Results

The results discussed herein pertain to the utility function only. That is, we

V present the outputs for the utility functions, i.e. the ni in Eq.(2) and Eq.(3), and do not produce probability estimates. Thus, the emphasis in this chapter is on the effects of the attributes and trip context (trip characteristics) on the utility levels relative to the base alternative, train. The results between the two approaches, i.e. single equation and separate equation approach, are nearly equivalent to one another. To preserve flow, the single equation model outputs are shown in the main text, whilst the separate model outputs are shown in Appendix 6.1. All control variables were effects-coded.

Hausman-Mcfadden test was applied to single and multi-destination models. Due to a large number of alternative specific parameters in these models, a procedure outlined in Hensher et al. (2005) was followed. For both models, the effect of an absence of the car, Jetstar and Virgin Blue was tested (these three alternatives were most popular in the choice sample). In all six cases, the Hausman-Mcfadden test was negative; this indicates that there is insufficient evidence to reject the IIA axiom and that the MNL model is adequate. To be sure, Inclusive Value (IV) test was conducted. The IV test can be more powerful in revealing IIA violation (Hausman-Mcfadden 1984). Nested multinomial logit specification between ‘air’ and ‘ground’ alternatives revealed a violation of the IID assumption for the single-destination model. This is shown by the significant IV parameter on the ‘air’ nest at the 5% level (Table 6.2). Nonetheless, the nested logit did not contradict the results of the MNL (see Appendix 6.2). We persevered with the MNL model results because they illustrate our points in a simpler manner. Table 6.3 shows the summary statistics of the MNL models (both single equation and separate models).

6-19

Table 6.2 IV parameter results

Single-destination model (for Multi-destination model (for the 'Air' nest; 'Ground' IV = 1) the 'Air' nest; 'Ground' IV = 1)

IV parameters 0.64577 0.15311

P-value 0.0564 0.4212

No. of observations 604 604

Table 6.3 Summary Statistics

Single equation Single-destination Multi-destination model model model

Log Likelihood (no coefficient) -1831.582 -1249.7444 -1249.7444

Adjusted pseudo R^2 0.262 0.282 0.246

No. of observations 1,208 604 604

The final result of the single equation model is presented in Table 6.4. Each of the coefficients is interpreted as a ceteris paribus effect on the total utility of a given travel mode (relative to the train alternative). Attributes found to be insignificant for all the alternatives in the model were dropped during the model estimation process. The key purpose of Table 6.4 is to show the results we wished to highlight the most in the context of the research question of this chapter. Consequently, some context-interaction variables were omitted from the model. The train mode was the base alternative for all alternative specific constants and variables. In Table 6.4, the variables with a single asterisk (*) are significant at 10%, two (**) and three (***) represent significance at 5% and 1% respectively.

6-20

Table 6.4 MNL estimation results

Variables Coefficients P-value Variables Coefficients P-value Variables Coefficients P-value

Constants MD constants Inertia (drove before) CAR 0.28 CAR -0.27 Car 0.42 ** COACH -1.70 * COACH -1.03 DJ 0.43 DJ -1.03 ** Inertia (flew before) GC -0.85 GC -0.71 DJ 1.08 *** JQ 0.37 JQ -0.85 * GC 0.68 ** REX -0.07 REX -0.91 ** JQ 1.22 *** RC -1.89 ** RC 0.17 REX 0.90 ***

Price ('High' price) MD-on-price ('High' price) Travel party size CAR -0.02 -- CAR 0.40 *** COACH -1.24 * -- COACH 0.14 DJ -0.58 *** DJ 0.16 DJ 0.29 GC -0.29 GC -0.55 GC 0.38 ** JQ -0.72 *** JQ 0.23 JQ 0.34 ** REX -0.71 *** REX 0.09 REX 0.32 * RC 0.11 -- RC 0.40 **

Price ('Medium' price) MD-on-price ('Medium' price) Age CAR -0.02 -- CAR 0.84 *** COACH 0.41 -- COACH 0.54 DJ -0.06 DJ 0.08 DJ 0.92 *** GC -0.57 ** GC 0.37 GC 0.79 *** JQ 0.03 JQ -0.26 JQ 0.82 *** REX -0.22 REX 0.15 REX 0.85 *** RC -0.11 -- RC 0.73 **

Freq ('High' frequency) MD-on-freq ('High' frequency) COACH 0.96 ** -- DJ -0.15 DJ 0.37 * GC -0.19 GC 0.06 JQ 0.47 *** JQ -0.37 * REX 0.29 ** REX -0.08

Freq ('Medium' frequency) MD-on-freq ('Medium' frequency) COACH -1.24 * -- DJ 0.11 DJ -0.18 GC 0.11 GC -0.04 JQ -0.30 ** JQ 0.11 REX -0.15 REX 0.31

Schedule (arrival in the afternoon) COACH 0.00 DJ -0.23 ** GC 0.00 JQ -0.11 REX 0.00

Note: DJ = Virgin Blue; JQ = Jetstar; REX = Regional Express; GC = Gold Coast; RC = Rental car; MD = multi-destination; freq = frequency

6-21 Price (Price ($220) and Price1 ($150)) Price variables were highly significant for all air alternatives. For instance, when the price was ‘high’ ($220 one-way) Virgin Blue yields a loss of 0.58 in utility, but gained 0.58 when the price was very low ($80)v (given that the coefficient on ‘price $150’ is effectively zero). Thus, there is a 1.16 utility difference (0.58 – (- 0.58)) between a Virgin Blue flight when the price is $80 compared with a Virgin Blue flight when the price is $220. The same applies to all other alternatives.

Time variables All time variables were not significant at 5% level. They were subsequently removed from the model and the table. This is surprising because time is often an important explanatory variable in urban mode choice studies, although it is usually the case that leisure tourists are less responsive to time than business travellers. Potential reasons for this result are discussed in the next section.

The surrogates for convenience (schedules and frequency) Virgin Blue’s morning arrival was preferred to an afternoon arrival. For a given frequency, it appears that tourists will derive some additional utility if the arrival time is earlier than the current 12pm arrival service. Frequency is statistically significant for Jetstar and REX. For instance, thrice-daily frequency is a positive source of utility for tourists choosing Jetstar.

Other variables The significant age coefficients for each mode show that, as age increases, the attractiveness of alternatives other than train increases relative to the train alternative. Risk, road condition, fuel price and reliability variables were either insignificant, or statistically significant but too small relative to other statistically significant variables such as price and trip context.

Inertia effect A person’s choice in the experiment is explained, to an extent, by the mode actually chosen in the current trip. If the person drove to Byron, then the utility

6-22 from choosing the car mode increases by 0.42 units of utility relative to choosing the train mode in the choice scenario. Similarly, if the person actually flew to the destination for the trip on which the survey was undertaken, choosing to fly again generally yields much greater utility than choosing the car mode or any other alternatives.

Trip context effect (multi-destination (MD) constants, multi-destination effect on price (MD-on-price) and multi-destination effect on frequency (MD-on-freq)) The ‘context’ effect has a similar level of influence on JQ, DJ and REX. If a visitor, ‘in addition to a stay in Byron, is to stay at least one night in regions at least two hours drive away from Byron’, then the utility derived from air transport diminishes. For example, the utility earned from flying with Virgin, Jetstar and REX decrease by a constant of 1.03, 0.85 and 0.91 respectively. The context can also moderate the influence that modal attributes have on choice. The variables under ‘MD-on-price’ and ‘MD-on-freq’ show the effect of context on price and frequency. With the exception of frequency and price, the context-and-attribute ‘interaction’ effects were mostly insignificant. These variables were subsequently omitted from the model and the table.

6.8 Discussion and implications

The results show that multi-destination context has an effect of shifting the utility functions of air transport alternatives by a negative constant relative to single- destination trips. This is shown by the significant MD-constant variables but mostly insignificant MD-on-price and MD-on-freq variables. Whilst the overall utility function shifts, the ‘slopes’ of the utility functions remain equal. This has an interesting interpretation in random utility theory. The alternative specific

6-23 constants can be viewed as the average impact of the unobserved utility on the alternatives (Hensher et.al. 2005). This suggests that the important determinants of travel mode choice from single vs. multi-destination point of view were not captured by our model’s attributes; rather their effects were captured by the MD- constants. Variables that should be included in the future are ‘affective’ factors such as ‘a sense of freedom’ or other functional factors such as ‘a degree of flexibility’ (Anable and Gatersleben 2005). The differences in trip context are more likely to manifest through these attributes of travel modes.

Results show that modal substitution is a source of conflict between LCCs and regional dispersal. There is evidence that a modal switch would occur from car to air even in situations when car may be the most suitable mode for the trip. The findings show tourists experience disutility from flying when the trip involves travel beyond the gateway regions, i.e. dispersal. This is shown by the negative MD-constant variables on air travel alternatives. In fact, the ‘increase in utility sourced from a decrease in airfare from $220 to $150’ is insufficient to offset the ‘loss in utility of air travel due to the need to disperse’ (or simply put, the influence of ‘context’). However, in situations when the price decreases from $220 to $80, the gain in utility is sufficient to offset the disutility of context, ceteris paribus. For instance, Virgin Blue’s utility increases by 1.16 when the price drops from $220 to $80 (see Results section), which is larger than the disutility of 1.03 caused by the shift in the choice scenarios from single- destination to multi-destination travel. This suggests that, in the presence of ‘low’ airfares, multi-destination trip arrivals by air will increase, because even if air travel may ‘inconvenience’ tourists’ travel upon arrival, tourists are willing to trade-off the ‘inconvenience’ for the low price, regardless of the trip context. Thus, from this we can learn how LCC can introduce a greater mixture of tourists arriving by air. This consequently has the effect of reducing the bias that air modes have in bringing greater single-destination than multi-destination travellers.

6-24 The results have a number of implications regarding the nature of the relationship between regional destinations and airlines, as well as for the subsequent challenges for destination managers. First, the level of airfare is an important factor that determines whether or not mode choices cause conflicts between affordable air-services and regional dispersal. When airfare levels are medium to high, the trip context effect dominates the utility gained from a decrease in fares. However, when airfares become low, tourists are much more likely to switch to air even in situations when car may be the most suitable form of transport for the trip. This implies a bypass of destinations (e.g. Port Macquarie in Figure 2) en route by those travellers making the switch from ground modes toward air. It is noted, however, that more research is needed in order to determine whether the accessing tourists who paid low airfares may use rental cars to visit the peripheral destinations, or limit their travels to the gateway only. The extent to which this occurs will determine the ‘net’ effects of affordable airfares on tourist dispersal, as well as on the region’s tourism economy.

Second, the results from this study have shown that in the presence of low airfares, multi-destination trip arrivals on air will increase and that these travellers should be identified and targeted to encourage dispersal from the gateway. The consequences of direct and cheap air travel on rapid urbanisation and congestions in tourism destinations have been documented in the tourism research literature (e.g. Papatheodorou 2002). In Australia, the spatial pattern of air travel demand is such that individual LCC services to peripheral regions within close proximity is not economically viable for the LCCs. Therefore, for those regions in the vicinity of the gateways, it is important to provide sufficient means of ground transport by which the demand for dispersal to the periphery is facilitated and enticed. Otherwise, increased congestion may appear in the gateway cities, causing the very problem that the Australian government aims to relieve (as outlined earlier in introduction).

Third, the results of this study have implications for cooperative marketing and the developments of niche markets. Through travel mode choice, marketing

6-25 promotion for multi-destination trips in one area may induce an unintended yet favourable impact for the destinations en-route. In our example, greater dispersal to the regions peripheral to Byron will induce more car travel along the corridor because driving the entire trip becomes a more attractive option. This increases the likelihood of planned or spontaneous stopovers en-route in regional centres such as Port Macquarie or Coffs Harbour, which belong to a different administrative boundary (for tourism) to Ballina and Byron (see Figure 2). Knowledge of the ‘natural partners’ among regional destinations can help regional tourism organizations to mobilise marketing resources more effectively. This research has shown that the greater understanding of mode choice can help to identify the linkage patterns between two regions belonging to different geo- political boundaries.

Fourth, the results from this study have demonstrated that the linkage patterns among regional destinations may change as a result of changes in airline services. It was shown previously that car travel benefits both the destinations en route and those peripheral to the gateway. However, when airfares are low, flying becomes a more attractive option, inducing tourists to bypass en route destinations while maintaining their visits to the periphery of the gateway. As a consequence of changes in airfares, what may have previously been a natural partnership between two regions may no longer be so, tilting towards that of competition than complementarity through modal substitution.

Fifth, the significance of the inertia effect indicates that there is a degree of rigidity in the willingness of tourists to switch modes. That is, tourists have the tendency to drive if they have driven to the destination before. Given that this study was undertaken in a static setting, the inertia effect can be interpreted as a short-run rigidity that draws parallel to the inelastic nature of demand for many goods and services in the short-run, but elastic in the long-run. LCC proliferation is seen as a crucial step towards the development of air travel in tourism much in the same way as the development of the charter sector and aviation deregulation (Bieger and Wittmer 2006). Among the many effects of the significant changes in

6-26 the air travel market, Quiggin (1997) argued that one effect of aviation deregulation in Australia was the ‘demonstration effect’ to travellers, that air travel was no longer a luxury reserved only for the affluent travellers. Hence, in the long-run, greater flexibility in substitutions between those two modes (car and air) can perhaps be expected.

6.9 Limitations and further research

One surprising result from this study was the lack of significance of the time attributes. The author proposes the following explanation. The utility function specified for each mode in the MNL model was made of each mode’s attributes i.e. cross effects were not estimated with a MNL model (Eq(2)). Thus, in specifying the time attribute in the experiment, the attribute levels varied in the time specific to that mode e.g. car’s time varied from 7 to 11 hours (a variation of up to four hours), whereas air modes varied from 1.5 hour to 3.5 hours (two hours variation). It is plausible that the study subjects were not responsive to differences in time because air is still the fastest mode by more than three hours when the upper and lower bounds for air and private car times are compared. This absolute time advantage of air travel holds even when out-of-vehicle time is added. As a matter of fact, inter-regional mode choice studies such as Hensher (1997), have shown that leisure tourists, compared to business travellers, are less responsive to time but much more in price. Thus, our result is not so surprising in this respect. Furthermore, this result may be a reflection of the differences in tourists’ behaviour compared to other choice contexts, as noted by Debbage (1991:266) in the study of tourists’ spatial behaviour in the Bahamas, “research in other fields (intra-urban commuting patterns, consumer shopping behaviour, and residential location decisions) may not be directly transferable to tourist behaviour”. Thus,

6-27 more empirical investigation into the sources that generate these differences is an important research issue for the future.

As for other attributes, the schedule-frequency nested attribute specification may be appropriate for future tests. This specifies a relationship between the two attributes, which will yield an output that is more amenable to interpretation relative to the case when they are independently specified. For instance, rather than an independent specification of ‘morning arrival’ and ‘three flights a day’, a nested schedule-frequency attribute have an interpretation that ‘a morning arrival flight of the three flights available’.

The lack of observed choices for alternatives such as coach, train and rental car are likely to have contributed to some inaccuracies in the respective alternatives’ parameter estimates. While a choice based sampling strategy was considered, this necessarily is a strategy for revealed preference data collection. Moreover, some modes on Sydney-Byron segment were favored by a particular group of tourists, e.g. the popularity of coach services by international backpackers, who were not the subjects of this study. Although tourist data at the level of Sydney-Byron is not readily available, recent statistics released by the Australian Federal government agency, Tourism Research Australia, shows that only 7% of domestic overnight visitors to Byron arrive on modes other than car (74%) or air (19%)vi (TRA 2008). Thus, small sample sizes for the other modes are consistent with the true market share of the population.

For future work, this research can be extended in a number of ways. For instance, the number of alternatives can be reduced to air and car, and specify a tree (nested) structure that examines the choice of transport mode at the destination, given the mode used to access the destination. Such specification will allow a comparison of a choice between ‘drive only’ and ‘fly and then drive’. Capitalising on fly-drive market is an important challenge for the destinations located peripheral to the gateway, and may also offer opportunities for the destinations en route as travellers may travel back home in the hired vehicle.

6-28

Our results show, in regards to dispersal, low airfares can increase the mix of tourists arriving by air. If the LCCs remain low-cost primarily to offer low-fares (e.g. abstain from providing ‘business’ class), and if Ballina-Byron is served by at least two competing airlines, presumably then ticket-discounting practices will continue on this corridor. Since the availability of ground travel modes at the destination is critical for tourists’ spatial behaviour at the destination, the provision of transport at the destination/gateway will become an important challenge for destination managers. Regional tourism organizations and government agencies responsible for the management and distribution of benefits from tourism for their respective tourism regions would require more information on the level of influence a better local transport system might have on the dispersal of tourists and the associated economic benefits. For these problems, the nested structure mentioned above can include other alternative travel modes such as public bus services, shuttle buses and rental cars. When the stated choice experiment is applied in such a context, we can generate information on the effect of ‘ground travel mode availability’ on the propensity of tourists arriving by LCCs to ‘venture beyond the gateways’, so as to evaluate the impact of regional transport infrastructure on tourist dispersal. Such line of research extension is discussed further in Chapter 7.

6.10 Conclusion

This chapter has analysed the relative importance of travel mode attributes and trip characteristics on mode choices of leisure tourists on the Sydney to Byron- Ballina travel corridor in Australia. The results empirically demonstrated that travel mode choice can be an avenue of conflict between LCC service

6-29 proliferation and tourists’ regional dispersal. The study found that when airfares become low, tourists are much more likely to switch to air even in situations where car may be the most suitable mode of dispersal for the trip. Thus, when airfares are low, the complementary relationship between two regional destinations that stem from the use of cars along the travel itinerary, may reverse to that of conflict, as a result of modal substitution from car towards air travel. The results have shown that trip context triggers a shift, but does not induce a change in the slope of the utility functions. It was argued that this supports the inclusion of qualitative and affective factors of travel mode choice in future studies.

Although Australian data was employed in this study, the results should be of interest to regional destinations worldwide. This is particularly the case for those destinations that are geographically large and where choice between a domestic flight and alternative ground transportation is a real option for potential travellers. The issues surrounding the implications of the growth of LCC for towns, which have traditionally relied on ground transportation for accessing tourists, have been under-researched despite their substantial importance to regional destination managers. The issues addressed in this chapter go at least some way to filling this research gap.

6-30 APPENDIX 6.1

Single Destination Model Multi-Destination Model t statistic Variables Coefficient P-value Variables Coefficient P-value

Constant CAR 0.739 CAR 1.141 ** -0.437 COACH 0.125 COACH -1.536 1.093 DJ 1.133 * DJ 0.421 0.760 GC 0.333 GC -0.961 1.196 JQ 1.288 ** JQ 0.252 1.104 REX 0.721 REX -0.057 0.793 RC 0.181 RC -0.312 0.586 Price ('High' price) CAR 0.062 CAR -0.068 0.612 COACH -0.804 COACH -0.620 -0.378 DJ -0.580 *** DJ -0.421 *** -0.746 GC -0.223 GC -0.781 ** 1.260 JQ -0.723 *** JQ -0.502 *** -0.979 REX -0.729 *** REX -0.617 *** -0.373 RC -0.007 RC -0.201 0.446 Price ('Medium' price) CAR 0.136 CAR -0.118 1.236 COACH -0.497 COACH 0.474 -1.090 DJ -0.052 DJ 0.034 -0.431 GC -0.586 ** GC -0.206 -0.903 JQ 0.020 JQ -0.220 1.164 REX -0.199 REX -0.068 -0.501 RC 0.056 RC 0.069 -0.032

6-31

APPENDIX 6.2

Single Destination Model (MNL) Single Destination Model (Nested) Variables Coefficient P-value Variables Coefficient P-value Base alternative = Rental car Constant CAR 0.920 CAR 0.922 ** COACH 0.235 COACH 0.221 DJ 1.318 * DJ 2.272 GC 0.521 GC 1.884 JQ 1.473 ** JQ 2.876 REX 0.907 REX 2.318 TRAIN 0.181 TRAIN 0.671 Price ('High' price) CAR 0.062 CAR -0.068 COACH -0.804 COACH -0.620 DJ -0.578 *** DJ -0.621 *** GC -0.223 GC -0.239 JQ -0.723 *** JQ -0.734 *** REX -0.729 *** REX -0.750 *** TRAIN 0.068 TRAIN 0.063 Price ('Medium' price) CAR 0.136 CAR -0.118 COACH -0.497 COACH 0.474 DJ -0.054 DJ -0.064 GC -0.586 * GC -0.579 * JQ 0.020 JQ 0.031 REX -0.199 REX -0.221 TRAIN -0.637 TRAIN -0.643

[*] Wald-test 10% level of significance; [**] 5%; [***] 1%

6-32 i Virgin Blue commenced Sydney – Port Macquarie services in early 2008. ii Tourism Research Australia; based on three-year average to June 2007 iii Australian Automobile Association Road Assessment Program 2005 iv Road Traffic Authority 2006 v The control variables are effects coded, e.g. 'Price' represents 'high' price ($220) and 'Price1' represents 'medium' level price ($150). The coefficient for the 'low' level price ($80) is obtained by {-(coefficient ‘price’ + coefficient ‘price1’)}.

Thus, if ‘price’ coefficient is (-0.58) and ‘price1’ coefficient is ‘zero’ then the coefficient of $80 is {-(-0.58 + 0)}, which is 0.58. vi Based on three-year average to June 2007

6-33

7. CONCLUSION, LIMITATIONS & FUTURE RESEARCH

7.1 Review

LCCs have stimulated air travel demand to the regions. It was shown in Chapter 2 that the LCCs have stimulated domestic dispersal, and increased the share of air travel over other modes of travel, which also had an effect of increasing the reliance of regional destinations on air transport. It was then proposed that the natural path to follow was to examine the effect of LCCs on regional dispersal of tourists. This was the general aim of this thesis (denoted G1).

G1. Examine the effects of LCCs on the regional dispersal of domestic visitors in Australia.

Altogether, there were five specific research aims. These are revisited below:

A1. Provide an interpretative survey of the aviation and tourism research literature and the secondary data sources relevant to understanding the link between LCCs and domestic dispersal (Chapter 2);

A2. Identify and explicate the relationships between regional dispersal and LCCs based on aviation, tourism and spatial behaviour research (Chapter 3);

7-1 A3. Build and test a causal model of regional dispersal and the intra-modal differences between LCCs and NCs (Chapter 4);

A4. Examine the trade-offs between destination transport factors and tourists’ travel characteristics in the choice of air arrivals’ regional dispersal (Chapter 5, ‘The Cairns experiment’);

A5. Examine inter-regional travel mode substitution as a source of conflict between low fare air services and regional dispersal (Chapter 6, ‘The Ballina-Byron experiment’);

A1 and A2 were interpretative surveys of the relevant literature and secondary data sources. The completion of A1 and A2 equipped us with the necessary contextual information and conceptual framework to derive the propositions for the empirical studies. The general research problem was framed in three inter- related research issues, which were individually examined in Chapters 4, 5 and 6.

This concluding Chapter is organised as follows. First, key findings from the empirical studies are briefly revisited. The subsequent section illuminates implications for the field’s theoretical development and government policy. Then, research limitations and some critical junctions for future research are outlined.

7.2 Key findings

Air arrivals increased to the regions as a result of the LCCs. NVS data indicated that the periphery’s share of the total air arrivals were between 29% and 33% over the last ten years. In other words, while it was clear that the LCCs contributed, in relative terms, more to the regions than the state capitals (incl. Gold Coast), there

7-2 is insufficient evidence to suggest a differential effect on the gateways and the periphery.

Traffic volume figures hide much of what is interesting about the LCCs, such as the differential characteristics of tourists, and the impact of the characteristics on dispersal propensity. The differences in the type of demand associated with the LCCs are documented in the aviation and tourism research literature. The results from the characteristics model have shown that some of these differences are empirically supported (please refer to Table 4.1). In particular, the following results and implications were highlighted. First, staying in one’s own property and friend and relatives’ property were important sources of dispersal for the LCC arrivals, implying that their economic impact may be lower due to the lower levels of expenditure injected. Second, risk and uncertainty reduction, and preference heterogeneity of the travel group, were particularly important motivating factors of dispersal for the air arrivals.

The results have provided evidence that, given the assumption of significant airfare differential between the NCs and LCCs, there will be discernible differences between the characteristics of LCC arrivals and NC arrivals. This is a significant finding because it provides a link between airline service types and dispersal impact. In particular, the evidence suggests that dispersal sourced from the LCC arrivals may inject much less expenditure than the NC arrivals. This explains why some destinations observe high growth in airport activity but the levels of tourism activity do not reflect the levels suggested by the airport activity. The analyses in this thesis has produced evidence that suggests affordable air arrivals tend to disperse for reasons that are different from the traditional air arrivals.

The trip factors examined in Chapter 4 were mostly exogenous to the destination, i.e. determined before arrival at the destination. In Australia, the geography of air travel demand is such that point-to-point LCC services to every peripheral regions

7-3 located within ‘close proximity’ to one another is not economically viable for the LCCs. Whether or not destination transport can influence the dispersal of air arrivals is relevant for the peripheral destinations looking to entice dispersal of the air arrivals. Obtaining an answer to this question, ‘can destination transportation policy stimulate the dispersal of the air arrivals, even in situations where the air arrivals exhibit trip characteristics that are dispersal-adverse?’ was the aim of the Cairns experiment.

The stated choice experiment in Chapter 5 has shown that appropriate ground travel mode attributes can offset some or all of the negative effects of trip characteristics on the choice of tourists to disperse. However, the extent to which this is feasible depends on destination context. In Chapter 5 it was shown that the dispersal to the North is easy to entice because northern destinations, which include Douglas and Daintree, are much more popular than southern destinations. The northern destinations are in fact the key attractions for the travellers flying to Cairns in the first place. To the less-popular destination region - the South - the importance of trip characteristics compared to modal attributes was strong, indicating that individual trip characteristics are binding constraints on dispersal to the South. Length of stay and travel party size were constraints that tended to reduce air arrivals in Cairns from reaching the southern destinations during their travel. Further, it was found that there are prospects for cheap public transport equipped with appropriate qualitative attributes to stimulate some demand to relatively unknown destinations. But this may be politically difficult to implement due to its potential conflict with regional tour operators who are likely to lose market share if the scheme is introduced.

For regional tourism destinations reliant mostly on ground travel modes, the extent of the low-airfare-induced-modal-substitution will determine the extent of the bypass effect. The final research question was ‘can low airfares induce tourists to switch from car to air, even in situations where the car may be the most suitable mode of dispersal for the trip?’ The Ballina-Byron experiment has shown that when airfares become low, tourists are more likely to switch to air even in

7-4 situations where the car may be the most suitable form of transport for the trip.

In many cases, peripheral destinations will be subject to a mixture of the two issues presented in Chapter 5 and 6. Consequently, better understanding of trip itineraries becomes an important task for destinations. While regional tourism destinations are clearly affected by the airlines’ conducts and performance, they have little influence over the airlines and airfares. Thus, continuous monitoring and understanding of the effects of airline strategies on destinations (such as airfare changes and flight frequency changes) are essential market intelligence that can benefit regional tourism.

7.3 Contribution to knowledge and implications for stakeholders

7.3.1 Contribution to theory

As shown by the literature review, the spatial behaviour of tourists in destinations and the proliferation of affordable air services are linked; for instance, the linkage between fly-and-drive patterns in the Mings and McHugh (1992) study and the proliferation of new-entrant-jet carriers in the U.S. Another example is the reduction in the length of stay of Western European travellers to the Mediterranean, which is associated with the LCCs’ growing share of traditional charter routes at the expense of charter carriers. In both instances, the link between spatial behaviour and airline business models has not been explicitly recognised. This thesis makes an original contribution to developing a theoretical link between the intra-modal transport choice (e.g., airline choice) and the spatial behaviour of tourists. This thesis shows that the effects of LCC (and more generally the effects of affordable air travel) on regional dispersal are trip characteristics oriented as much as traffic volume.

7-5

Research in tourism has largely neglected an analytical approach to assessing the trade-offs between travel mode choice and spatial behaviour. This thesis contributes by providing a utility compensation perspective on tourists’ choice of transport and the resulting spatial behaviour of tourists. The thesis also examined the trade-offs between ‘economic’ factors and ‘tourism’ factors of mode choice. It was also shown in this thesis that in long-distance leisure travels, trip characteristics vary widely across individuals and travel parties, and these have significant influence on the choice of travel modes. In some situations, trip characteristics offset the marginal utility gained from the changes in travel mode attributes. Therefore, the theoretical contribution of this thesis is in highlighting how our understanding of the relationship between long-distance leisure mode choice and spatial distribution of tourists can be improved by accommodating tourism variables and a wider range of trip characteristics in the discrete choice framework.

7.3.2 Implications for policy

The findings from this thesis should be of relevance to governments whose mandates may emphasise greater balance in the distribution of economic benefits from tourism. Cheap air transport can trigger a bypass effect of ground-mode- reliant destinations through inter-modal substitution. Cheap air transport can also stimulate tourists that are dispersal-averse. Thus, cheap air transport can contribute to the disparity in levels of growth between airports and tourism destinations. This illustrates some of the challenges in policy implementation because the dispersal of primary interest at the federal level - domestic dispersal - may conflict with the objective of greater (regional) dispersal at the state and local level.

7-6 Local transport issues are often at the centre of public policy agenda in state and local governments. As Gunn (1988) noted, local level tourism-transport planning and policy have a strong political dimension because the competition for funding tends to be greater at this level. This renders prioritisation an important task in the allocation of resources and policy making. One issue is that tourism is often at the lower end of the priority list behind social, environmental and other economic objectives (Ashworth 2009). This thesis highlights the growing importance of demand for local transportation by air leisure arrivals. In particular, the results have shown that public transportation can be an important mode of travel that meets the interests of a number of policy objectives, including environmental policies aimed at reducing car-usage and tourism policies aimed at greater regional dispersal. The findings from this thesis help bring the tourism and dispersal concerns to the forefront of regional and local transport policy appraisals.

7.3.3 Implications for destinations

To illustrate the relevance and value of these results, we discuss the results in the context of cooperative marketing of regional tourism destinations. Trip characteristics of the air arrivals are often determined prior to arrival, preventing tourists from dispersing to the peripheries. Thus, engineering greater dispersal is a formidable task because it requires the understanding of tourists’ trip planning stages. Being able to exert some sort of influence at this level by a single destination region is a difficult task because it is very expensive to do so (research and marketing costs), and also due to the free-rider problem of destination marketing. If air travel is the only real option for many accessing tourists, then it is probably appropriate that cooperative marketing arrangements take place because peripheral destinations will only collectively command an adequate demand for regular LCC services.

7-7

However in situations where both ground and air travel modes are real options for accessing tourists, the same conclusion no longer holds. The Ballina-Byron experiment presented in Chapter 6 has demonstrated that the linkage patterns among regional destinations may change as a result of LCCs and affordable air travel. The second experiment has shown that when airfares are low, flying becomes an attractive option, which induces tourists to bypass en-route destinations. As a consequence, what may have been a natural partnership between two regions, i.e. peripheral destinations located en route and those surrounding the gateway, may no longer be so, tilting the relationship between two regional destinations towards that of competition through modal substitution. This relationship should be considered for a more efficient allocation of regional tourism organisation’s funds.

7.4 Limitations and future research

7.4.1 Applicability of the results

The research issues are not restricted to any particular location in Australia, rather they stem from trip itinerary literature based on several international empirical work. The results from two choice experiments are most relevant for geographically large tourism regions; perhaps a rule of thumb indicator of a large tourism region may be a tourism region spanning at least 2-3 hours drive from one end to the other end of the tourism region boundary. For the Cairns experiment, the results are applicable to destinations with the mixture of following characteristics: (1) only one regional airport option for LCC services or alike in the tourism region; (2) the tourism region’s reliance on air services for incoming

7-8 tourists is significant; (3) there is a disparity in the popularity of peripheral destinations (e.g. hinterland vs. coastal, rural vs. cities). As for the Ballina-Byron experiment, the results should be of relevance for those destinations that are geographically large and where a choice between domestic flights and alternative ground transportations is a real option for potential travellers.

There are some limitations on the validity and generalisability of results on the alternatives with low sample choices. There was an under-representation of some choices in the experiments; for instance, the share of the train alternative in the Ballina-Byron study, or the share of the day-trip by public bus alternative to the northern destinations of Cairns. This has rendered model estimation for these alternatives and attributes difficult. While this reflects a common problem in many primary data based research, it is nonetheless a factor that limits our interpretation of the models for those alternatives. This also highlights a potential issue with stated choice methods. For this research, we brought closure to this issue by noting that we cannot control the number of ‘observed’ samples for all alternatives because they are the very choices that we aim to collect from the field. Instead, we controlled explanatory variables - and this is the key advantage of the stated choice method because we control the conditions under which choices are made. Choice-based sampling strategy, which solves the problem outlined here, could not be used because it is a revealed preference, not stated choice, sampling method.

7.4.2 Limitations of the MNL: utility compensation perspective and taste heterogeneity

This thesis contributes to tourism research by providing a utility compensation perspective on tourists’ choice of transport and spatial behaviour; the utility compensation perspective highlights the importance of trip characteristics and ‘contextual utility’ in a way that can be directly compared to the effects of travel

7-9 mode attributes. However, the utility compensation perspective has major drawbacks; for instance, there are interpretation issues of ‘utility’. Further, since the interpretation of utility can be meaningful only in a relative sense (relative utility), confusion can easily arise. In particular, there is a need for more research on the theoretically appropriate interpretation of trip characteristics and contextual utility in long-distance leisure travel context. Given the fact that discrete choice models include socio-economic characteristics information to account for taste and preference heterogeneity of individuals (Ben Akiva and Lerman 1985), the following questions may require further attention: how should we interpret the coefficients of trip characteristics? Should they be considered as having direct utility? Or are they ‘moderators’ akin to socio-economic characteristics? Is this approach consistent with utility maximisation?

The discrete choice models applied in this thesis estimated a coefficient vector assumed to be equal for all tourists in the sample. For instance, all sampled individuals were treated as having the same responsiveness to a unit change in airfares, or the same responsiveness to a change in the number of stopovers on public transport. However, these coefficients are likely to vary across market segments; for example, ‘general sightseeing’ tourists are likely to gain more utility than ‘activity-specific’ tourists when there is a stopover for sightseeing opportunities on a public transport service. Another example, in the context of air travel and dispersal, may be that ‘psycho-centric’ tourists exhibit different responsiveness to airfares to ‘allocentric’ tourists. As a result, the former may have a higher willingness-to-pay for a mode that can provide the required flexibility of a private vehicle, and consequently lower responsiveness to airfare discounting practices in choosing air travel. A similar line of reasoning was used in Chapter 6 to argue that airfares play a role in increasing the mixture of tourists to a destination (in terms of bringing a greater variety of spatial behaviour).

Such an issue – the heterogeneity in tastes and preferences - cannot be fully explored with the multinomial logit model because dichotomy of tourists such as

7-10 that described above (allocentric and psychocentric) are not easily observed by the analyst. The effect of such latent variables cannot be explicated in the standard multinomial logit models. The problem outlined above is essentially a problem of analysis and manipulation of the error structures of discrete choice models to capture the effects of latent variables such as tourists’ allocentricity. By extending the methodological boundary towards random probit or mixed logit models (or random coefficient models), such an issue can be explored in much greater detail. Future studies can test the effects of these latent variables on the relationships between tourists’ mode choice and spatial choice behaviour.

7.4.3 Operationalising ‘dispersal’

Alternative methods in operationalising dispersal should be considered in the future. The LGA boundaries as used in the first empirical study, while based on geo-politically salient boundaries, are arbitrary for tourists since they have little or no knowledge of the boundaries. An alternative may be to specify a measure of dispersal that is continuous, such as an index, or dependent variables that categorise dispersal in levels such as ‘high’, ‘medium’ or ‘low’ (ordered). Econometric investigation into the appropriate level of geographic delineation was done using multinomial logit models by Eymann and Ronning (1997). Similar application to delineating ‘dispersal’ boundaries will advance this field of research by providing a theoretical and empirical basis to measuring dispersal.

7.4.4 Integrating destination and mode choice

A number of extensions on the current research are desirable. The central premise of this research was the choice behaviour of tourists in intra-regional and inter- regional transport mode choice contexts. While both stated choice experiments accounted for some contextual information, e.g. trip contexts, this was largely

7-11 exogenous to the model of mode choice. For instance, the Ballina-Byron case study examined the moderating influence of single and multi-destination trips on long-distance travel mode choice, while the Cairns case study examined the moderating influence of northern and southern destinations on intra-regional travel mode choice, as well as the choice to disperse or not. While such designs enable us to extract the moderating influence of destination on mode choice, these designs cannot illuminate situations where tourists choose a distant leisure destination (1 hour flight) over closer destination (say, less than 3 hours drive) as a result of cheap airfares.

Models that endogenise destination choice within travel mode choice (or vice versa) can be used to examine the influence of transport modal attributes, namely cheap airfares, on the choice of destinations. Such research design will be capable of estimating an econometric model that may be able to explain ‘destination neutrality’ phenomenon observed by Mason (2005), where tourists tend to substitute destinations based on cheap airfares. Australian examples will be the effect of cheap air travel on the choice between the series of regional destinations on the Eastern Coast, which can be viewed broadly similar in their ‘sun, sand and sea’ attributes, or the choice between destinations in the outback and the coastal beaches. Such research design can also embed the model of ‘regional dispersal’ within ‘domestic dispersal’. Moreover, such an analytical approach has the potential to add to existing research on destination price competitiveness (e.g. Dwyer et.al. 2000) to help answer a question such as ‘how would a long-haul LCC or bilateral capacity relaxations impact on the competitiveness of Australia as a destination compared to other long-haul alternatives such as U.S. or Europe by international visitors?’

7.4.5 The time attribute in leisure and tourism

An interesting finding from this research was that the ‘time’ attribute was insignificant. The determining power of travel time in travel mode choice is

7-12 significant in the context of journey-to-work trips (e.g. Redmond and Mokhtarian 2001) and in long-distance inter-regional trips (e.g. Hensher 1997, Koppelman and Sethi 2005). The insignificant result may be a reflection of the relatively time- insensitive nature of leisure travellers, in particular when the range of travel time examined is between one to three hours. One implication of this result is that peripheral destinations are not significantly disadvantaged by the fact that they are an hour further from the gateway relative to another destination, at least within the travel time range examined here. In fact, the evidence supports Page’s (1994) argument that in tourism, transport is not only a cost to be minimised, but also an integral part of tourists’ overall travel experience. A positive utility could be attached to travel time, in which case we will not observe a significant negative relationship between utility and time. The positive utility in travel time is also illustrated, albeit to a small extent, in the intra-mode journey-to-work trips; for example, Redmond and Mokhtarian (2001) have shown that travellers to work prefer a short commuting time than none. By the same token, if this was the case then the choice experiments should have observed a positive significant value on time. Perhaps future studies can apply a similar approach with the aim of isolating the positive and the negative effect of time on utility.

Overall, the finding on time is in-line with the qualitative work of Lumdson (2006) and Eaton and Holding (1996) outlined in Chapter 5. In both studies, travel time was not mentioned as a key determinant for the demand of public transport for leisure travel to UK’s National Parks. The results from our experiments support the relative unimportance of travel time for leisure travellers. This is in line with Debbage (1991) who noted in the context of tourists’ spatial behaviour in the Bahamas, “research in other fields (intra-urban commuting patterns, consumer shopping behaviour, and residential location decisions) may not be directly transferable to tourist behaviour” (p.266). Thus, empirical work into the sources that generate these differences is an important research agenda for the future.

7-13

7.5 Towards an integrated model of tourists’ spatial choice and tourism yield

In this thesis, it is assumed that greater dispersal is equivalent to the greater visitations in the regions beyond capital and gateway cities. While this is a common measure used for decision-making in the industry, number of visitations has several shortcomings. A more comprehensive measure is tourism yield. Yield in tourism is variously defined; Dwyer et al. (2007) classified four types of yield: expenditure (tourists’ spend), financial (impact on firms’ profits and sales), economic (income and employment generated) and sustainable yield (environmental and social impact). Causal relationships between LCCs and dispersal should be developed with respect to various types of yield.

As discussed in Chapter 3 and Chapter 4, VFR travel purpose has increased in the share of air travel as a result of increased air travel affordability. Significant proportion of VFR saves on accommodation expenses by staying in ‘friends and relatives’ property’, which reduces the level of tourist expenditure. Dispersal arising from VFR may add to dispersal visits and nights, but comparatively little to expenditure and financial yield. Further to financial yield, given the traditionally labour-intensive nature of the accommodation industry (Dwyer, Forsyth and Spurr 2003), the marginal effect of a dollar spent by dispersing tourists may contribute little to the economic yield in the regions. Moreover, the level of leakages will be significant in peripheral regions because small regional economies tend to have a more homogenous industry base; consequently, significant share of tourists’ expenditure will leak-out as import payments to other regions and abroad.

Finally, sustainable yield is increasingly becoming of great import for the industry and governments. Sustainable yield includes the environmental and social impact (Dwyer et al. 2007) although not exclusively; Becken and Simmons (2008) added ‘regional dispersion’ in sustainable yield to account for the significant

7-14 contribution of tourism in ‘sustaining’ the well being of regional economies. In fact, one emerging theme from research on tourism yield is that there are trade- offs among yield types (Dwyer et al. 2007, Becken and Simmons 2008), and transport and aviation is an important part of the trade-offs. A recent study by Becken and Simmons (2008) on the yield of international visitors in New Zealand found that there are trade-offs between financial yield and sustainable yield; for instance, the ‘coach traveller’ (a tourist segment that uses air transport as a primary mode to travel within NZ) had high financial yield but performed poorly on dispersal, although the segment’s absence from road-based tourism meant that its carbon footprint was low as well.

Here we can appreciate the complex trade-offs in the context of LCCs - cheap air travel - and dispersal. For instance, air transport might be positively related to financial yield and sustainable yield (compared to car-based tourism) but it is negatively related to dispersal. The implication is that policy aimed at greater dispersal will be achieved, to an extent, at the expense of environmental and economic yields. The role of LCCs, or affordable air travel, is paramount in moderating these links and trade-offs.

7-15

References

Alegre, J. and Pou, L. (2006) “The length of stay in the demand for tourism” Tourism Management, 27(6), 1343-1355

Anable, J and Birgitta G. (2005) “All work and no play? The role of instrumental and affective factors in work and leisure journeys by different travel modes.” Transportation Research Part A: Policy and Practice, 39(2-3): 163-181

ANOP Research Services Pty Ltd (2005) "Motorists' Attitudes 2005 ANOP National Survey Prepared for the Australian Automobile Association." http://www.aaa.asn.au/publications/polls.php

Ashworth, G. (2009) “Heritage Management and Urban Tourism”. Keynote in CAUTHE, Proceedings of the 18th Annual Conference, Carlsen, J., Hughes, M, Holmes, K. and Jones, R. (ed), Curtin University of Technology

Australian Automobile Association, "Australian Road Assessment Program Risk Maps". http://www.ausrap.org/ausrap/riskmaps.htm

Australian Bureau of Statistics, “statistical local areas” http://abs.gov.au

Australian Bureau of Statistics, “Tourist Accommodation Statistics 2007”

Australian Bureau of Statistics, Tourism Regions Map release 2007

R-1 Barrett, S. D. (2004). "How do the demands for airport services differ between full-service carriers and low-cost carriers?" Journal of Air Transport Management 10(1): 33.

Becken, S. and Simmons, D. (2008). “Using the concept of yield to assess the sustainability of different tourist types” Ecological Economics, v.67 (3): 420-429

Ben-Akiva, M. and Lerman, S. R., 1985. Discrete Choice Analysis: Theory and Application to Travel Demand. MIT Press, Cambridge Masachusetts.

Bieger, T. and Wittmer, A. (2006) Air transport and tourism – Perspectives and challenges for destinations, airlines and governments, Journal of Air Transport Management, 12(1), p.40-46

Borenstein, S. and Rose, N. (1994). “Competition and Price Dispersion in the U.S. Airline Industry” Journal of Political Economy 102(4): 653-683

Borooah, V.K. (2002). Logit and Probit: Ordered and Multinomial Models. Thousand Oaks, CA, Sage University Papers Series on Quantitative Applications in the Social Sciences.

Bureau of Infrastructure (2008), Aviation Policy Green Paper – Flight Path to the Future, Department of Infrastructure, Transport, Regional Development and Local Government

Bureau of Transport and Communication Economics (1991) Domestic Aviation of Australia: First year after deregulation, BTCE

Bureau of Transport and Communication Economics (1993) Domestic Aviation of Australia: Three years after deregulation, BTCE

R-2 Bureau of Transport and Regional Economics (2007) "Transport Statistics." http://www.btre.gov.au

Bureau of Transport and Regional Economics (BTRE), Aviation Statistics, available from http://btre.gov.au

Burghouwt, G., Hakfoort, J. and Eck, J. (2003), The spatial configuration of airline networks in Europe, Journal of Air Transport Management 9: 309-323

Button, K. and Stough, R. (2000) “Air Transport Networks: Theory and Policy Implications” Edward Elgar Cheltenham

Centre for Asia Pacific Aviation (2006) Peanuts: the low cost airline weekly no.138 October 3rd

Centre for Asia Pacific Aviation (2007) Peanuts: the low cost airline weekly no.120 July 3rd

Chou, Y. (1993). “Airline deregulation and nodal accessibility” Journal of Transport Geography 1(1), p.36-46

Clippinger, M. and Strong, J. (1987) “Changes in Distribution Channels and the Travel Agency Business” in Meyer, J. and Oster, C. (ed) Deregulation and the future of intercity passenger travel, MIT Press

Cooper, C.P. (1981) Spatial and temporal patterns of tourist behaviour. Regional Studies, 15(5), 359-371

Crouch, G.I. (1995) A meta-analysis of tourism demand, Annals of Tourism Research, 22(1), p.103-118

R-3 Crouch, G.I. (1996) “Demand elasticities in international marketing: A meta- analytical application to tourism” Journal of Business Research, 36(2): 117-136

Crouch, G.I., Oppewal, H., Huybers,T., Dolnicar,S. Louviere, J.J. and Devinney. T. (2007) "Discretionary Expenditure and Tourism Consumption: Insights from a Choice Experiment." Journal of Travel Research, 45(3):247-258

Debbage, K. (1991) "Spatial Behavior in a Bahamian Resort." Annals of Tourism Research vol.18:251-268

Department of Industry, Tourism and Resources (2003) "Tourism White Paper - a Medium to Long Term Strategy for Tourism." Department of Industry, Tourism and Resources.

Dobruszkes, F. (2007). "An analysis of European low-cost airlines and their networks." Journal of Transport Geography In Press, Corrected Proof.

Doganis, R. (2006) The Airline Business, Routledge

Dresner, M. (2006) Leisure versus business passengers: Similarities, differences, and implications. Journal of Air Transport Management 12 (1), 28-32

Duval, T. (2008) Tourism and Transport: Modes, Networks and Flows, Channel View Publications

Dwyer, L. and Forsyth, P. (1992) The reform of air transport and its impact on tourism in Forsyth, P.(ed) Microeconomic Reform in Australia, Allen and Unwin

Dwyer, L. and Forsyth, P. (1993) “Assessing the Benefits and Costs of Inbound Tourism” Annals of Tourism Research, 20 (4) pp.751-768

R-4 Dwyer, L., Forsyth, P. and Rao, P.(2000) "The Price Competitiveness of Travel and Tourism: a comparison of nineteen destinations" Tourism Management, Special issue: the Competitive Destination, 21(1) pp 9-22.

Dwyer, L., Forsyth, P. and Spurr, R.(2003), ‘Inter-industry effects of tourism growth: implications for destination managers’, Tourism Economics, 9(2): 117- 132.

Dwyer, L., P. Forsyth, L. Fredline, L. Jago, M. Deery and S. Lundie (2007) “Yield Measures for Australia’s Special Interest Inbound Tourism Markets” Tourism Economics 13 (3), 421–440

Eaton, B. and D. Holding. (1996) “The evaluation of public transport alternatives to the car in British National Parks.” Journal of Transport Geography 4(1): 55-65

Eymann, A. Ronning, G. (1997) Microeconometric models of tourists’ destination choice, Regional Science and Urban Economics, 27(6), pp.735-761

Fennell, D.A. (1996) “A tourist space-time budget in the Shetland islands.” Annals of Tourism Research 23(4): 811-829

Forsyth, P. (1992) Microeconomic Reform in Australia, Allen and Unwin

Forsyth, P. (2003). "Low-cost carriers in Australia: experiences and impacts." Journal of Air Transport Management 9(5): 277.

Forsyth, P. (2006) Martin Kunz Memorial Lecture. Tourism benefits and aviation policy, Journal of Air Transport Management, 12(1): 3-13

Fourie, C & Lubbe, B 2006, “Determinants of selection of full-service airlines and low-cost carriers: A note on business travellers in South Africa”, Journal of Air Transport Management, vol. 12, issue 2, 98-102

R-5

Francis, G., A. Fidato, I. Humphreys (2003). "Airport-airline interaction: the impact of low-cost carriers on two European airports." Journal of Air Transport Management 9(4): 267.

Francis, G., I. Humphreys, S. Ison, M. Aicken (2006) "Where next for low cost airlines? A spatial and temporal comparative study." Journal of Transport Geography In Press, Corrected Proof.

Francis, G., Dennis, N., Ison, S., Humphreys, I. (2007) The transferability of the low-cost model to long-haul airline operations, Journal of Air Transport Management 28(2) p.391-398

Franke, M. (2004). "Competition between network carriers and low-cost carriers-- retreat battle or breakthrough to a new level of efficiency?" Journal of Air Transport Management 10(1): 15.

Gillen, D. and A. Lall (2004). "Competitive advantage of low-cost carriers: some implications for airports." Journal of Air Transport Management 10(1): 41.

Graham, A. (2000) “Demand for leisure air travel and limits to growth” Journal of Air Transport Management, 6(2), 109-118

Graham, A. (2006) “Have the major forces driving leisure airline traffic changed?” Journal of Air Transport Management, 12(1), 14-20

Graham, B. and Guyer, C. (2000) “The role of regional airports and air services in the United Kingdom” Journal of Transport Geography 8, 249-262

Greene, W. and D.A. Hensher. (2003) "A latent clas model for discrete choice analysis: constrasts with mixed logit." Transportation Research Part B: Methodological 37: 681-698.

R-6

Greene, W.H. (2002) LIMDEP Version 8.0 Reference Guide, Econometrics Software, Inc.

Grimm, C.M. and Miloy, H.B. (1993) Australian domestic aviation deregulation: impacts and implications, Logistics and Transportation Review 29

Gunn, C. A. (1988) Tourism Planning, Taylor & Francis

Hanlon, J.P. (1992) ‘Regional air services and airline competition’, Tourism Management, 13(2), p181-195

Hensher, D.A. (1997) "A practical approach to identifying the market potential for high speed rail: A case study in the Sydney-Canberra corridor." Transportation Research Part A: Policy and Practice 31(6): 431-446.

Hensher, D.A., Prioni, P. (2002) “A Service Quality Index for Area-wide Contract Performance Assessment.” Journal of Transport Economics and Policy 36(1): 93- 113

Hensher, D.A., John M. Rose and William Greene. (2005) Applied Choice Analysis: A Primer, Cambridge University Press.

Hooper, P. Findlay, C. (1998) Developments in Australia’s aviation policies and current concerns, Journal of Air Transport Management, 4(3) p.169-176 http://abs.gov.au

Huybers, T. (2003) ‘Modelling short-break holiday destination choices’, Tourism Economics 9(4): 389-405

R-7 Hwang, Y. and Fesenmaier, D. R. (2003), “Multidestination Pleasure Travel Patterns: Empirical Evidence from the American Travel Survey”. Journal of Travel Research 42(2): 166

Hwang, Y., Gretzel, U. and Fesenmaier, D. (2006) “MULTICITY TRIP PATTERNS Tourists to the United States.” Annals of Tourism Research 33(4): 1057-1078

Ito, H. and Lee, D. ‘Low Cost Carrier Growth in the U.S. Airline Industry: Past, Present, and Future’ (April 9, 2003). Brown University Department of Economics Paper No. 2003-12. Available at SSRN: http://ssrn.com/abstract=719741

Jara-Diaz, S.R. (1991) “Income and taste in mode choice models: Are they surrogates?” Transportation Research Part B: Methodological 25(5): 341-350

Johnston, R.J., Gregory, D., Pratt, G., Watts M. (2000) “The Dictionary of Human Geography”. Blackwell publishing.

Kain, J., Webb, R. (2003). Turbulent times: Australian airline industry issues. Report available online at http://www.aph.gov.au/library/pubs/rp/2002- 03/03RP10.pdf. Website last accessed 7th December 2005.

Kelly, Ian (ed.) (2001) Australian Regional Tourism Handbook: Industry Solutions 2001 Centre for Regional Tourism Research. CRC for Sustainable Tourism Pty. Ltd.

Koppelman, F.S. and Vaneet S. (2005) "Incorporating variance and covariance heterogeneity in the Generalised Nested Logit model: an application to modelling long distance travel choice behaviour." Transportation Research Part B: Methodological 39(9): 825-853.

R-8 Lawton, T.C. (2002) Cleared for take-off: Structure and strategy in the low fare airline business, Ashgate.

Lew, A. and McKercher, B. (2002) "Trip destinations, gateways and itineraries: the example of Hong Kong." Tourism Management 23(6): 609-621.

Lew, A. and McKercher, B. (2006) "Modelling Tourist Movements: A Local Destnation Analysis." Annals of Tourism Research 33(2): 403-423.

Li, X., Cheng, C., Kim, H., Petrick, J. (2008) “A systematic comparison of first- time and repeat visitors via a two-phase online survey” Tourism Management 29(3): 429-438

Limtanakool, N., Dijst, M. and Schwanen, T. (2006) “The influence of socio- economic characteristics, land use and travel time considerations on mode choice for medium –and longer-distance trips” Journal of Transport Geography, 14(5), 327-341

Louviere, J.J. and Hensher, D.A. (1983) "Using discrete choice models with experimental design data to forecast consumer demand for a unique cultural event." Journal of Consumer Research, 10(3):348-61.

Louviere, J.J. and Hensher, D.A., and Swait, J.D. (2000) Stated Choice Methods: Analysis and Application, Cambridege University Press.

Lue, C.C., J.L. Crompton and D.R. Fesenmaier. (1993) "Conceptualization of Multi-destination Pleasure Trips." Annals of Tourism Research, 20: 289-301

Lumsdon, L. (2006). “Factors affecting the design of tourism bus services.” Annals of Tourism Research 33(3): 748-766

R-9 Lumsdon, L. and Page, S. (2004). "Progress in Transport and Tourism Research: Reformulating the Transport-Tourism Interface and Future Research Agendas" in Tourism and Transport: Issues and Agenda for the New Millennium. edited by Lumsdon, Les and Stephen, J. Page, 1-28. Elsevier.

Maddala, G.S. (1986) Limited dependent and qualitative variables in econometrics, Econometric Society Monographs (No. 3)

Mansfeld, Y. (1990) "Spatial Patterns of International Tourist Flows: Towards a Theoretical Framework". Progress in Human Geography. 14 (3), 372-390

Mansfeld, Y.(1992) Tourism: Towards a behavioural approach, Progress in Planning, 38(1), 1-92

Marcus, B. and Anderson, C.K. (2008) Revenue management for low-cost providers, European Journal of Operational Research 188: 258-272

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

Mason, K. (2005) Observations of fundamental changes in the demand for aviation services, Journal of Air Transport Management, 11(1), 19-25

Mason, K (2006) The value and usage of ticket flexibility for short haul business travellers, , Journal of Air Transport Management 12: 92-97

Mason, K., Alamdari, F.(2007) EU network carriers, low cost carriers and consumer behaviour: A Delphi study of future trends, Journal of Air Transport Management, 13(5), p.299-310

Meyer, J.R. and Clinton V. O. (1987) Deregulation and the future of intercity passenger travel, MIT Press.

R-10

Mings, R.C. and McHugh, K.E. (1992) The Spatial Configuration of Travel to Yellowstone National Park, Journal of Travel Research, 30(4), p.38-46

Morley, C. (1994) "Discrete Choice Analysis of the Impact of Tourism Prices." Journal of Travel Research, 33(2): 8-14

Morrison, S. and Winston, C. (1986) The Economics Effects of Airline Deregulation, Brookings Institution Press

Moscardo, G., Saltzer, R., Norris, A., McCoy, A. (2004) Changing Patterns of Regional Tourism: Implications for tourism on the Great Barrier Reef, The Journal of Tourism Studies, Vol. 15, No. 1, p.34 – 50

Moscardo, G. and Phillip P. (2004) "Life Cycle, Tourist Motivation and Transport: Some Consequences for the Tourist Experience." in Tourism and Transport: Issues and Agenda for the New Millennium. edited by Lumsdon, Les and Stephen, J. Page, 1-28. Elsevier.

Mules, T. and Huybers, T. (2002) Substitution between tourism destinations: An application of discrete choice modelling, Sustainable Tourism Cooperative Research Centre Project 43001

Nelson, R. and G. Wall (1986) Transportation and accommodation: Changing interrelationships on Vancouver Island, Annals of Tourism Research, Vol. 13, 239-260

Nicolau, J.L. and Francisco J. M. (2006) "The influence of distance and prices on the choice of tourist destinations: The moderating role of motivations." Tourism Management, 27(5):982-996

R-11 Njegovan, N. (2006) “Elasticities of demand for leisure air travel: A system modelling approach” Journal of Air Transport Management, 12(1), 33-39

O’Halloran M., Cook S., Sbragi A. and Buchanan, I. (2000) BTR Occasional Paper Number 30, Rural Tourism in Australia: The visitor’s perspective, Bureau of Tourism Research, Canberra.

O'Connell, J. F. and G. Williams (2005). "Passengers' perceptions of low cost airlines and full service carriers: A case study involving Ryanair, Aer Lingus, Air Asia and Malaysia Airlines." Journal of Air Transport Management 11(4): 259.

Oppermann, M. (1994) “Regional aspects of tourism in New-Zealand” Regional Studies, 28(2), 155-167

Oppermann, M. (1995) "A Model of Travel Itineraries." Journal of Travel Research, 33(4): 57-61

Oppermann, M. (1997) “First-time and repeat visitors to New Zealand” Tourism Management, 18(3), 177-181

Oppewal, H. and Timmermans. H. (1991). "Context effects and decompositional choice modelling " Papers in Regional Science 70(2).

Page, Stephen. (1994). Transport and Tourism: Global Perspectives. Pearson Education

Page, Stephen. (2005). Transport and Tourism: Global Perspectives. Pearson Education, 2nd Edition

Pantazis, N. and I. Liefner "The impact of low-cost carriers on catchment areas of established international airports: The case of Hanover Airport, Germany." Journal of Transport Geography In Press, Corrected Proof.

R-12

Papatheorodou, A. (2002) "Civil Aviation Regime and Leisure Tourism in Europe." Journal of Air Transport Management 8(6): 381-388

Parolin, B. (2001) “Structure of Day Trips in the Illawarra Tourism Region of New South Wales” The Journal of Tourism Studies, 2(1), 11-27

Pearce, D.G. (1979) “Towards a geography of tourism” Annals of Tourism Research, 6(3), 443-454

Pearce, D.G. (1987) “Spatial patterns of package tourism in Europe” Annals of Tourism Research, 8(4), 183-201

Pina, A.I. and Diaz Delfa M.T. (2005) "Rural tourism demand by type of accommodation." Tourism Management 26(6): 951-959

Prideaux, B. (2000) “The role of the transport system in destination development” Tourism Management 21: 53-63

Quiggin, John. (1997) "Evaluating Airline Deregulation in Australia." The Australian Economic Review, 39(1): 45-56 Redmond, L., S. and Mokhtarian, P. (2001), “The positive utility of the commute: modelling ideal commute time and relative desired commute amount.” Transportation, 28: 179-205

Reynolds-Feighan, A. (2001) Traffic distribution in low-cost carriers and full- service carrier networks in the US air transportation market, Journal of Air Transport Management 7: 265-275

Rizzi, L.I. and Ortuzar, J. (2003) "Stated preference in the valuation of interurban road safety." Accident Analysis and Prevention 35: 9-22.

R-13 Road Traffic Authority. "Pacific Highway Upgrade." Road Traffic Authority. http://www.rta.nsw.gov.au/constructionmaintenance/majorconstructionprojectsreg ional/pacifichighwayupgrade/index.html

Sinclair, T.(1998) Tourism and economic development: A survey, Journal of Development Studies, 34(5), p.1-51

Sinha, D. (2001) Deregulation and liberalisation of the airline industry: Asia, Europe, North America and Oceania, Ashgate

Stewart, S. and Vogt, C. (1997), "Multi-destination Trip Patterns." Journal of Travel Research, 33:57-61

Swan, W. (2007) Misunderstandings about airline growth, Journal of Air Transport Management 13: 3-8

Taplin, J. and McGinley, Carmel (2000) "A linear program to model daily car touring choices." Annals of Tourism Research 27(2): 451-467.

Tideswell, C. and Faulkner, B. (1999), "Multidestination Travel Patterns of International Visitors to Queensland." Journal of Travel Research, 37:364-74

Timmermans, H.J.P. and R.G. Golledge. (1990) "Applications of behavioural research on spatial problems II: preference and choice." Progress in Human Geography. 14.(3): 311-354.

Tourism Australia. (2005). “Emerging Travel Patterns in Fly-Drive Interstate Leisure Tourism”, prepared by Tourism and Aviation Economics for Tourism Australia

Tourism Research Australia (2008) “Travel by Australians” Quarterly Results of the National Visitor Survey.

R-14

Tourism Research Australia. (2007) "Local Government Area Profiles." http://tra.australia.com/regional.asp

Tourism Transport Forum. (2002) “Keeping the bush in the game - New approaches to making regional tourism more competitive”, prepared by Hocking Research and Consulting for Tourism Transport Forum

Urry, J. (1991) The Tourist Gaze Sage

Wall, G. (1997) Tourist Attractions: Points, Lines and Areas, Annals of Tourism Research, 24/1:240-3

Walmsley, D.J. (2004), “Behavioural approaches in Tourism Research”, in Lew, A, Hall, C.M., Williams, A.M. (ed) A Companion to Tourism, Blackwell

Warnock-Smith, D. and Potter, A. (2005) ‘An exploratory study into airport choice factors for European low-cost airlines’ Journal of Air Transport Management, Volume 11, Issue 6, November 2005, Pages 388-392

Weaver, D. (2006) Sustainable Tourism: Theory and Practice, Butterworth- Heinemann

Whyte, R. and Prideaux, B. (2007) "Impact of Low Cost Carriers on Regional Tourism," refereed paper in CAUTHE, Proceedings of the 17th Annual Conference, McDonnell, I., Grabowski, S., March, R., (Eds) CD-ROM, University of Technology, Sydney.

Williams, G. (2001). "Will Europe's charter carriers be replaced by "no-frills" scheduled airlines?" Journal of Air Transport Management 7(5): 277

Williams, G. (2002) Airline Competition: Deregulation's Mixed Legacy, Ashgate.

R-15

Wu, C.L. and Carson, D. (2008) "Spatial and Temporal Tourist Dispersal Analysis in Multi Destination Travel." Journal of Travel Research, 46(3): 311-317

R-16