A Systematic Analysis of Retail Centre Distribution and Customer Travel Behaviour

Author Shobeiri Nejad, Seyedeh Maryam

Published 2016

Thesis Type Thesis (PhD Doctorate)

School Griffith School of Environment

DOI https://doi.org/10.25904/1912/2456

Downloaded from http://hdl.handle.net/10072/367354

Griffith Research Online https://research-repository.griffith.edu.au

A Systematic Analysis of Retail Centre Distribution and Customer Travel Behaviour

Seyedeh Maryam Shobeiri Nejad B.Arch. (Hons), M.Sc.

School of Environment Griffith Sciences Griffith University

Submitted in fulfilment of the requirements of the degree of Doctor of Philosophy

June 2015

Abstract

The important role of retail trips in the overall transport sustainability of cities in Australia has been broadly overlooked within travel demand management programs (TDMs). Brisbane, as the third largest city in Australia, has undergone fast and extensive urban development and become heavily reliant on an increasing number of shopping centres to provide its residents with products, services, leisure and lifestyle opportunities. Based on the 2009 Household Travel Survey data, shopping centres were the destination for more than half of weekly ‘retail trips’ in Brisbane. The evolving structure of a modern one-stop, multi-purpose retail format provides more options to customers who rely on their cars to travel to shopping destinations.

The evident economic success of these centres over recent decades has made their owners and operators powerful and wealthy and consequently hard to compete with. Planning policies such as zoning, required Environmental Impact Assessments, and transit-oriented developments (TODs) have tried to control the wide-ranging impacts of these shopping centres. However, these policies have not typically been accepted by developers and land owners and have not proved to be successful.

While a significant number of international studies have been conducted on retail, there is a clear lack of systematic research on the potential for reductions in trip numbers or length, and also on how, in the current car-reliant environment, shoppers can be encouraged to shift to other transport modes. Furthermore, very few studies have specifically address the impacts of shopping centres on customers’ retail travel behaviour.

This research addresses a broader range of elements which potentially affect customers’ travel behaviour and their choice of retail destinations. It aims to identify improved planning policies regarding the location and distribution of these centres to encourage the use of other transport modes except the cars. Given that shopping centres are sited at specific locations in the city, it is important to understand what influences shoppers’ destination choice, and also to consider how destination choice affects travel behaviour and mode choice. A wide range of potentially influential factors including current retail travel I

behaviour, customers’ socio-demographic characteristics and the spatial characteristics of shopping centres were therefore studied. A destination choice model was developed to quantify the impacts of these factors on customers’ choice of large, medium or small shopping centres as their trip destination. The preferences and concerns of significant retail players including developers and planners were then investigated through a number of interviews with professional practitioners. This informed a more comprehensive understanding of the likely implications of current and future retail trends and assisted in predicting the impacts of potential future policies on retail travel behaviour. The results from the destination choice model, together with findings from the interviews formed the basis for proposed future scenarios for retail development in Brisbane. An existing transport demand model (Brisbane Strategic Transport Model) was then used to predict mode choice and travel distances under future proposed policy and development scenarios for retail.

The results showed that large shopping centres will undoubtedly remain attractive destinations for purchasing clothes, household and personal goods. On the other hand, shoppers show a strong aversion to travelling longer distances to purchase groceries and food. Consequently, the transport implications of daily shopping for necessities should not be ignored. Planning approaches which provide these necessities more locally could succeed in modifying retail travel behaviour. For example, a larger number of big neighbourhood centres or village-type shopping centres distributed more evenly along transport corridors will not only support non-motorised and public transport trips, but will also reduce the distances travelled by cars and thus help to create a more sustainable environment. Encouragingly, this type of policy has started to attract considerable support from businesses and among planners.

Statement of Originality

This work has not previously been submitted for a degree or diploma in any university. To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.

Seyedeh Maryam Shobeiri Nejad

III

Dedicated to

my parents for their endless love, support and encouragement

Table of Contents

Abstract ...... I

Statement of Originality ...... III

Table of Contents ...... VI

List of Figures ...... XI

List of Tables ...... XIV

Abbreviations ...... XVI

Acknowledgements ...... XVII

Publications during candidature ...... XVIII

Keywords ...... XIX

Chapter 1 Introduction ...... 1

1.1 Statement of the problem ...... 3

1.2 Significance of the research ...... 6

1.3 Research Aims and Questions ...... 7

1.4 Justification of approach & structure of the thesis ...... 9

Chapter 2 Literature Review ...... 12

2.1 Introduction ...... 15

2.2 Economic geography and retailers’ concept of spatial distribution of retailing ...... 18

2.2.1 The concept of trade area and methods applied for its measurement ...... 19

2.3 Urban form, retail form and travel behaviour ...... 30

2.3.1 General methods of studying interactions between urban/retail form and travel behaviour ...... 35

2.4 Discussion of research gaps in the literature ...... 40

Chapter 3 The Case Study Settings ...... 45

3.1 Why Brisbane? ...... 47

3.2 Identifying Brisbane’s retail structure ...... 48

3.2.1 History of retail locations ...... 49

3.2.2 The beginning of the shifts ...... 49

3.2.3 The expansion of shopping centres ...... 52

3.2.4 Shopping centres’ catchment areas...... 53

3.3 Retail accessibility in Brisbane ...... 55

3.4 Regulatory concerns ...... 60

3.4.1 Commercial public policy and regulations in Brisbane ...... 60

3.4.2 Statutory documents and retail schemes ...... 63

3.4.3 Planning documents and their intended policies ...... 64

Chapter 4 Research Methodology ...... 71

4.1 Introduction ...... 73

4.2 Datasets available for research ...... 74

4.3 Methodology ...... 79

4.3.1 Analytical methods ...... 87

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane ...... 93

5.1 Introduction ...... 95

5.2 Literature review ...... 96

5.3 Methods ...... 98

5.4 Results and Discussion ...... 101

5.4.1 Customers’ travel behaviour to shopping centres & supermarkets ...... 108

5.5 Conclusion ...... 122

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour ...... 125

6.1 Introduction ...... 127

6.2 Literature review ...... 128

6.3 Methods ...... 130

6.4 Results and Discussion ...... 135

6.4.1 Mode Share ...... 138

6.4.2 Distance (km) travelled per capita by mode share ...... 139

6.4.3 Retail trip frequency by type of product ...... 140

6.5 Conclusion ...... 141

VII

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility ...... 145

7.1 Introduction ...... 147

7.2 Literature Review ...... 148

7.3 Methods ...... 151

7.3.1 Measuring retail density and population density around shopping centres ...... 152

7.3.2 Measuring the Land-Use Mix ...... 152

7.3.3 Accessibility of shopping centres via different modes of transport ...... 155

7.4 Results and Discussion ...... 159

7.4.1 Population density and the location of centres ...... 159

7.4.2 Retail density around the centres ...... 162

7.4.3 Mixed use analysis in the direct vicinity of shopping centres ...... 164

7.4.4 Accessibility of shopping centres by different modes of transport ...... 166

7.5 Conclusions ...... 171

Chapter 8 Shopping Centre Destination Preferences Modelling ...... 173

8.1 Introduction ...... 175

8.2 Literature Review ...... 177

8.2.1 Discrete choice models and the study of retail trips ...... 177

8.3 Methodology ...... 187

8.3.1 The Multinominal Logit Model ...... 187

8.3.2 Model estimations ...... 189

8.3.3 Data preparation and choice set construction ...... 193

8.4 Data analysis and results ...... 208

8.4.1 The MNL model developed for this study ...... 208

8.4.2 The model parameter estimates, utility and probability measurement ...... 214

8.5 Conclusion ...... 218

Chapter 9 Professionals’ Insights on Retail Accessibility ...... 221

9.1 Introduction ...... 223

9.2 Methods ...... 224

9.2.1 Interview questions ...... 225

VIII

9.3 Results and Discussion ...... 228

9.3.1 Planning Regulation and Processes for Retail Developments (interview questions 2 to 7) ...... 228

9.3.2 Views and perceptions about retail planning (interview questions 8 to 10) ...... 234

9.3.3 Future trends (interview questions 11 to 13) ...... 241

9.3.4 Sustainable transport for retail establishments (interview questions 14 & 15) ...... 246

9.4 Conclusion ...... 255

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour ...... 259

10.1 Introduction ...... 261

10.2 Literature review ...... 262

10.2.1 Scenario planning ...... 262

10.2.2 Land-use and transport scenario planning ...... 263

10.2.3 Land use and transport models/application of travel demand models ...... 265

10.2.4 Objective ...... 267

10.3 Methods ...... 268

10.3.1 Brisbane land use and transport model - BSTM_MM ...... 268

10.3.2 BSTM results for home-based shopping trips under anticipated future scenarios ...... 271

10.4 Results and Discussion ...... 278

10.4.1 Analysis of the results for city-wide scenarios ...... 278

10.4.2 Analysis of Regional-Scale Scenario Results ...... 280

10.4.3 Analysis of Neighbourhood Scale Scenario Results ...... 284

10.5 Conclusions ...... 289

Chapter 11 Summary, conclusion and policy recommendations ...... 291

11.1 Introduction ...... 293

11.2 Summary ...... 295

11.3 Contribution to knowledge ...... 305

11.4 Limitations ...... 308

11.5 Recommendations ...... 309

11.5.1 For Australian practice ...... 309

11.5.2 For international practice ...... 311

11.5.3 For further research ...... 312

References ...... 313

Appendix I ...... 328

Appendix Il ...... 338

List of Figures

Fig 1-1: Thesis framework ...... 11

Fig 2-1: Influential groups affecting the formation of retail structure in the city ...... 17

Fig 2-2: Reilly’s gravitation theory and the breaking-point concept ...... 21

Fig 2-3: Christaller’s ideal central place hierarchy showing nested hexagons. Each central place serves the market area of three smaller central places (Hanson, 1997) ...... 23

Fig 2-4: Diagram of a classic central place system (market area: K=3 hierarchy) (Downs and DeSouza, 2006) p.91 ...... 24

Fig 2-5: Walter Christaller, Central places in Southern Germany, Prentice-Hall, 1966 (Downs and DeSouza, 2006) ...... 24

Fig 2-6: Spatial models and methods applied to the design of retail trade areas (Yrigoyen and Otero, 1998a) ...... 29

Fig 2-7: Different categories of attributes which might affect retail travel behaviour ...... 42

Fig 3-1: Shopping centres’ locations in the Brisbane Statistical Division (BSD) ...... 52

Fig 3-2: The activity centre network hierarchy (Drechsler, 2014) P.274...... 54

Table 3-3: Passenger boarding across seven capital cities 2005-06 to 2007-08 (Australian Government, 2009) ...... 57

(e) Time of travel ...... 59

Fig 3-3: Trip attribute analysis of various trip types in SEQ Source (2012) P.19, 20 ,22 ...... 59

Fig 4-1: Datasets used in the research ...... 78

Fig 4-2: Three major participants with direct or indirect influence over the overall structure and form of retail ...... 80

Fig 4-3: Shopping centre locations and the zonal boundaries for the Census Collection District and the BSTM model ...... 83

Fig 4-4: Proposed research framework to address the major questions in this study ...... 86

Fig 4-5: Methods applied in each chapter to identify and quantify different aspects of the retail transport environment ...... 87

Fig 4-6: Methodologies applied in the thesis to study retail travel behaviour ...... 92

Fig 5-1: BSD boundary as the selected study area...... 100

Fig 5-2: Trip Frequency by Trip Purpose during Weekdays & Weekends ...... 101

Fig 5-3: Retail Trip Frequency by Mode-share for 18+ Residents-Weekdays ...... 103

Fig 5-4: Retail Trip Frequency by Mode-share for 18+ Residents-Weekends ...... 103

Fig 5-5: Retail Trip Frequency by Mode-share in Inner & Outer Brisbane - Weekdays ...... 105

Fig 5-6: Retail Trip Frequency by Mode-share in Inner & Outer Brisbane - Weekends ...... 105

Fig 5-7: Trip Frequency by Shop Type during Weekdays & Weekends ...... 106

Fig 5-8: Trip Frequency by Expenditure Code during Weekdays & Weekends ...... 108

Fig 5-9: Trips to Shopping Centres by the Expenditure code ...... 109

Fig 5-10: Trips to Supermarkets by the Expenditure code ...... 110

Fig 5-11: Trips to shopping centres by Mode-share & Distance during Weekends & Weekdays ...... 111

Fig 5-12: Trips to Supermarkets by Mode-share & Distance during Weekends & Weekdays.. 111

Fig 5-13: Brisbane super regional shopping centres and their customer originating zones ..... 112

Fig 5-14 Major Shopping Complexes’ share of the total Retail trips ...... 114

Fig 5-15: CCDs* including Regional and Major Shopping Complexes ...... 116

Fig 5-16: customers’ retail trip mode share during the Weekdays ...... 117

Fig 5-17: Percentage of trips during on-peak & off-peak hours on weekdays ...... 117

Fig 5-18: Weekdays’ trip percentage for on-peak and off-peak hours by centre’s category..... 118

Fig 5-19: Time spent in the retail destination ...... 119

Fig 5-20: Percentage of retail trips to various types of shopping centres based on the reported purchased items - Weekends ...... 121

Fig 5-21: Percentage of retail trips to various types of shopping centres based on the reported purchased items - Weekdays ...... 121

Fig 6-1: Trip Ratio for Retail and All-types of trips ...... 136

Fig 6-2: Retail Trip Frequency and Mode-share ...... 138

Fig 6-3: Distances Travelled for retail per capita by Mode-share ...... 139

Fig 7-1: Measuring land-use mix in the vicinity of shopping centres ...... 154

Fig 7-2: Population density and shopping centre locations in BSD in 2011, based on ABS data ...... 161

Fig 7-3: Retail density within BSD for 2016 ...... 163

Fig 7-4: The populated areas within the study boundary used for calculating the accessibility levels of the different categories of shopping centres ...... 167

Fig 7-5: LUPTAI public transport accessibiltiy map for regioal centres...... 168

XII

Fig 8-1: Parts of the thesis framework explaining the relation between the findings of previous chapters (answering the first research question) and how they feed into the discrete choice modelling section (answering the second research question) ...... 176

Fig 8-2: Location of shopping centres in the Brisbane Statistical Division (BSD) ...... 177

Fig 8-3: Probability density function for Gumbel and normal distributions (same mean and variance) (Koppelman and Bhat, 2006a) ...... 188

Fig 8-4: Mapping the pseudo- R2 to the linear R2 (Hensher et al., 2005) ...... 191

Fig 8-5: The zonal boundaries for the Census Collection District and the BSTM model ...... 198

Fig 8-6: Available attributes of the trip, destination centre and destination location, and characteristics of trip makers – for potential inclusion in the destination choice model ..... 199

Fig 10-1: Scenarios and forecasts / Source: (Ringland and Owen, 2007) ...... 262

Fig 10-2: A simple multinominal logit (MNL) model specification – Source (Ryan, 2008) ...... 269

Fig 10-3: BSTM-MM Study Area (subdivided by Collection Districts), showing constituent local government areas (LGAs) ...... 270

Fig 10-4: Location of shopping centres within the BSTM zonal system...... 272

Fig 10-5: City-wide, Regional and Neighbourhood study areas selected to measure the extent of travel behaviour change under alternative future scenarios for retail development ...... 277

Fig 10-6: Percentage change in mode for home-based shopping trips under the clustering and decentralizing scenarios: modelled at city-wide scale ...... 278

Fig 10-7: Percentage change in travel metrics for all trips under Clustering and Decentralising scenarios, compared with the baseline scenario: City-wide predictions ...... 280

Fig 10-8: Mode shifts in shopping trips, under different future scenarios, expressed as percentage changes from the 2016 baseline ...... 281

Fig 10-9: Percentage changes in overall trip characteristics, under different future scenarios, expressed as percentage changes relative to the 2013 baseline: modelled at a regional scale ...... 283

Fig 10-10: Percentage change in transport mode share and total number of shopping trips, relative to the 2016 baseline – Neighbourhood case study-1 [Garden City] ...... 285

Fig 10-11: Overall trip characteristics’ rate to the baseline - neighbourhood boundary-1[Garden City] ...... 286

Fig 10-12: Percentage change in transport mode share and total number of shopping trips, relative to the 2016 baseline – Neighbourhood case study-2 [Capalaba] ...... 287

Fig 10-13: Shopping trip characteristics’ rate to the baseline - neighbourhood boundary-2 [Capalaba] ...... 288

XIII

List of Tables

Table 2-1: Advantages and disadvantages of commonly applied methods to study the linkages between urban form and retail travel behaviour ...... 43

Table 3-1: Trips by trip purpose: CBD and non-CBD trips in 1992 (Ferreira, 1999) ...... 56

Table 3-2: Bus trips as percentage of all trips in 1992 (Ferreira, 1999) ...... 56

Table 3-3: Passenger boarding across seven capital cities 2005-06 to 2007-08 (2009) ...... Error! Bookmark not defined.

Table 3-4: Centre and Mixed Use Zone Categories - City Plan 2014 ...... 67

Table 5-1: Trip Frequency by Trip Purpose and Mode-share during Weekdays & Weekends * ...... 102

Table 5-2: Retail trip percentage to various shopping centres based on their trip origin ...... 120

Table 6-1: The most common socio-demographic groups in the 2009 SEQ-HTS ...... 135

Table 6-2: Retail-trips and All-trips made in one week ...... 137

Table 6-3: Percentage of retail trips for purchasing diverse products...... 141

Table 7-1: The number and percentage of centres of different hierarchy levels with High, Medium or Low levels of Mixed-use within a 1km radius ...... 164

Table 7-2: Entropy index measured within a 1-km radius of each shopping centre ...... 165

Table 7-3: Percentage accessibility of different levels of shopping centre by various mode types from 2,944 Brisbane collection districts with population densities >= 100 individuals/km2 169

Table 8-1: Summary of the literature on discrete choice models for shopping trips ...... 184

Table 8-2: Number of shopping centres in each condensed size category ...... 196

Table 8-3: Correlations among attributes of destination centres ...... 200

Table 8-4: Important attributes of Super Regional Centres (data from 2011) ...... 201

Table 8-5: Correlations among attributes describing the spatial vicinity around shopping centres ...... 202

Table 8-6: Attributes included in the MNL models, their defined categories and the selected baseline level ...... 211

Table 8-7: Results for model fit and significance of the 12 successive MNL models, as additional terms are added ...... 213

Table 9-1: Questions for public and private sectors interviewees ...... 226

Table 11-1: Main findings from Chapters 5 to 10 ...... 296

XIV

Abbreviations

Abbreviation Description

BSD Brisbane statistical division

CD Collection districts

CPT Central place theory

CDM Competing Destination Model

LUPTAI Land use and public transport accessibility index

PTKT Public transport kilometres travelled

RP Revealed preference

SCD Shopping centre directory

SEQHTS South East Queensland Household travel survey

TAZ traffic analysis zones

TDM Travel demand model

TOD Transit oriented development

VHT Vehicle hours travelled

VKT Vehicle kilometres travelled

XVI

Acknowledgements

This research would not have been possible without the advice, encouragement and support of many people. I would like to take this opportunity to extend my sincere gratitude and appreciation to all those who made this PhD thesis possible.

I would like to extend my sincere gratitude to my Principal Advisor, Dr Jim Smart for his endless patience, expert advice and unwavering encouragement throughout my Ph.D. My special gratitude and thanks is going to my initial Principal advisor Professor Neil Sipe, who I had his continuous support, invaluable guidance, constant enthusiasm and inspiration, even after his transition to the University of Queensland for the last year. I would also like to thanks Dr Matthew Burke, my associate advisor for his enlightening discussions.

I would also like to acknowledge Benjamin Pool and Peter English from the Department of Transport and Main Roads being so cooperative and providing data for this research. Their support has been critical in the fulfilment of this thesis.

I thank my fellow postgraduate students at the Urban Research Program (URP) at Griffith University and all friends particularly Lavinia Poruschi for being compassionate, thoughtful and making this journey pleasant. My special thanks to Professor Lex Brown and Dr Tooran Alizadeh for their constant support as well as the opportunities to work during my study.

I would like to express my sincere appreciations for my mother Sarieh Aboulian and for father Dr Ali Akbar Shobeirinejad, for their unconditional love, to whom I owe everything I have accomplished in my life.

I would like to extend my gratitude to my sister, Dr Ameneh and my brother in- law, Dr Ahmad Tavassoli for their affection and care and their constant inspiration and encouragement.

Finally, my appreciation also goes to my brother Dr Hamid and Dr Hadi and their lovely families for their inspiration, kind help and care throughout my PhD.

XVII

Publications during candidature

Journal paper - Shobeirinejad, M., Sipe, N., Burke, M., “Inductive Clustering of Customers In Brisbane - Investigating the Impacts of Socio-Demographic Characteristics on Retail Travel Behaviour”, Journal of Case Studies on Transport Policy, Accepted with minor changes

Book chapter - Shobeirinejad, M., Smart, J.C.R., Sipe, N., 2015. “A Polygonal Attitude Towards Customers’ Retail Travel Pattern and Their Spatial Destination Choice Decisions”, In: Zhao, P. ed. Planning for Sustainable Cities, Urban Challenges, Policy Responses and Research Agenda.

Conference papers - Shobeirinejad, M., Sipe, N., Burke, M., “Socio Demographic Groupings and Revealed Retail Travel Behaviour in Brisbane”, Washington, 93rd Transportation Research Board (TRB) Annual Meeting 2014 Proceedings – Presented and published

- Shobeirinejad, M., Sipe, N., Burke, M., “Retail Travel Behavior across Socio-Demographic Groups: A Cluster Analysis of Brisbane Household Travel Survey Data”, 13th World Conference on Transport Research (WCTR) 2013 Rio, Brazil - abridged version of the paper for the proceedings Presented and published

- Shobeirinejad, M., Veitch, T., Smart, J.C.R., Sipe, N., Burke, M., “Destination Choice Decisions of Retail Travellers: Results from Discrete Choice Modelling In Brisbane”, Australasian Transport Research Forum (ATRF), 2 - 4 October 2013, Brisbane, Australia - Presented and published in the proceedings

- Shobeirinejad, M., Smart, J.C.R., Sipe, N., Burke, M., “The Impact of Shopping Centre Attributes on the Destination Preferences of Trip Makers in Brisbane”, State of Australian Cities Conference, 26-29 November 2013, Sydney, Australia - Presented and published in the proceedings

- Shobeirinejad, M., Burke, M., Sipe, N., “Analysing Retail Travel Behaviour Using an Australian Data Set”, Australasian Transport Research Forum (ATRF), 26 - 28 September 2012, Perth, Australia - Presented and published in the proceedings

XVIII

Keywords

Retail travel behaviour, destination choice model, Travel demand modelling, Accessibility analysis, Cluster analysis, Customers’ socio-demographic characteristics, Transit oriented development

XIX

Chapter 1 Introduction

Chapter 1 Introduction 3

1.1 Statement of the problem

Today, sustainable transport strategies are pursued by hundreds of cities all around the world as part of their broader sustainability initiatives. A sustainable system is expected to simultaneously protect the natural environment, develop the economy, achieve equity (Jepson, 2001) and provide a better quality of life for all, now and into the future (Agyeman et al., 2003). Therefore ‘a sustainable transportation system is believed to focus on providing people access to different destinations while minimizing the negative effects of transport and maximizing economic prosperity and social equity’ (Rouhani, 2009).

Many studies have been undertaken during the last few decades investigating trip makers’ travel behaviour and the possible factors influencing their trip decisions. Much easier access to private cars and changes in the modern lifestyle have not only changed travel behaviour, but have also increased the overall number of everyday trips which people take to work and non-work destinations. Nelson et al. (2001) claimed that the increase in the number of non-work activities over the last three decades is so apparent that it accounts for three-quarters of all household vehicle trips and 80 percent of all trips in the US.

Work trips are typically less flexible than non-work trips because people need to be at work at a specific time and place. When it comes to shopping as a key component of non-work travel, there is more freedom because of the operating hours of shops. This lets the customers plan their trip duration, destination and mode of travel with fewer constraints and allows planners to have direct influence on these types of trips by changing the nature and distribution of shopping destinations in the city (Handy, 1996b).

Various studies define retail trips differently. Some studies have included all trips to retail destinations for a range of activities, such as shopping for goods and services, eating out, and participating in recreational, social and cultural activities (Niles and Nelson, 1999b). Other studies have considered retail trips as being only those trips undertaken in order to buy something, excluding all other recreational purposes (as defined – for example – in the South East

Chapter 1 Introduction 4

Queensland Household Travel Survey). Whatever the definition is, access to retail destinations can seriously affect diverse aspects of people’s travel behaviour. Issues such as trip frequency, destination choice and trip complexity (Hanson and Schwab, 1987), could all be extremely important in improving or diminishing trip makers’ quality of life (Iacono et al., 2010).

Australian cities, like many large cities around the world, are grappling with the adverse consequences of the rising number of car-reliant trips including traffic congestion and air pollution. A significant portion of these trips involves travel to retail destinations. The expansion of modern forms of retail outlets (such as regional shopping centres, warehouses, big box types of outlets and large supermarkets) began in the 1960s and continues today in close proximity to major roads and highways. This expansion was an attempt to provide a one- stop shopping opportunity for busy, two buyer modern households. Continuing through five decades of almost constant growth, and associated with more creative, complex and unpredictable shifts in retail environment, this expansion has made the owners of these retail businesses more powerful and less vulnerable to competition.

In accordance with Christaller’s central place theory, Australian shopping centres are developed and distributed following levels of hierarchy in order to provide similar products and services over large regional or metropolitan boundaries. These modern forms of retailing have significantly affected more walkable and easily accessible, traditional retail destinations, including high streets and corner shops, making it more difficult for these traditional outlets to function efficiently within smaller neighbourhoods. The large, dense undercover centres are particularly evident in the relatively newly established city of Brisbane, with its smaller population and a subtropical, humid climate, in comparison to Sydney or Melbourne. Retail expansion has provided the developers of these centres with the opportunity to create more pleasant, enjoyable and relaxing environments for customers.

Considering the large number of retail trips and their frequency in the overall trip sequence of many people, the transportation impacts of retail environments are substantial. In the case of almost 90 percent (Shobeirinejad et al., 2012) of trips

Chapter 1 Introduction 5 to retail centres, many factors have contributed to making the private car the only acceptable mode of transport. These factors include the large number of free parking spaces, proximity to major roads, large catchment areas, limited access to public and other transport options (changed slightly by provision of public transport stations at large regional centres within last few years), and expensive public transport compared to private car expenses. This is a significant threat to transport sustainability in Brisbane.

Australia’s travel demand management program has already directed a number of studies into peoples’ commuting and educational trips. However, regarding retail trips, to date there is a huge gap in the research. While retail accessibility, quality of service and transportation patterns are undoubtedly closely related to the retail form and structure of the cities, only a very limited number of studies address this issue from the urban planners’ perspective (Goodman and Coiacetto, 2009). To the author’s knowledge, no comprehensive, published, academic studies have analysed and measured different dimensions of retail environments and their impact on travel behaviour. There is also a lack of research into the current travel behaviour of retail customers in Australia, and the factors that influence their choice of destination. This is likely to be at least partly because accurate and comprehensive datasets on retail behaviour do not exist, or are not easily accessible. Many of the studies which have been undertaken have been carried out by big businesses to maximize their economic benefits (Baker and Wood, 2010). Typically, the datasets from these studies are not available to researchers.

In addition, planning regulations and policies, such as the Brisbane city plan ‘in- centre’ policy, (which has undergone some changes in the new city plan of 2014) do not support alternative distributions and forms of retail outlets. Thus, future expansion and location of car based retail forms in the city are likely to ‘lock-in’ motorised travel as the dominant transport mode for this type of trip. In short, the current retail environment and its on-going development trajectory, have failed to create and support sustainable retail travel behaviour in Australia. This is particularly true in Brisbane, the case study for this research.

Chapter 1 Introduction 6

1.2 Significance of the research

From the foregoing discussion, it is clear that there is a significant need to explore the impacts of the retail environment as an influential dimension of urban structure on peoples’ travel behaviour. Whilst there have been some attempts by transport modellers and engineers to consider and measure some of these impacts in their studies, many aspects of this complex environment remain neglected and so justify detailed investigation by urban planners. There is no comprehensive and reliable research in the Australian retail context that considers the current planning policies and their impacts on travellers’ behaviour. The existing literature has also failed to address likely trajectories towards a more sustainable retail transport environment in the future.

Given that retail trips typically account for almost a quarter of total trips (Shobeirinejad et al., 2012), any improvement in the sustainability of retail travel will significantly help to reduce traffic congestion, improve urban air quality and deliver other public benefits. The modification of retail form and land use pattern in Brisbane’s metropolitan area could provide a pathway towards a more sustainable environment, an objective that has been advocated in all legislated documents relating to the city.

The expected contributions of this research include the first in-depth exploration of contemporary retail travel behaviour in Australia, based on existing large travel diary datasets (HTS) and available retail datasets, such as shopping centre directories (SCD). This study intends to provide a rich understanding of the existing retail environment in Brisbane and to identify the factors which drive peoples’ shopping trip decisions. The study will also seek to identify current barriers and enablers for more sustainable retail travel behaviour. If, as expected, the study identifies key spatial factors that can be readily addressed via local and/or metropolitan scale planning, it will make some very significant contributions. The study will also aim to improve our understanding of how urban retail structure is implicated in the current transport problems of Australian cities, and will suggest how the current spatial settings can be modified to reduce these problems.

Chapter 1 Introduction 7

1.3 Research Aims and Questions

As previously explained, a sustainable transportation system is not only about providing more accessible destinations and reduction in the negative impacts of transport system on the environment but it also looks into other aspects of sustainability such as economic prosperity and social equity’ (Rouhani, 2009). While investigating all different aspects of sustainability is of crucial importance, the major goal of this research is to look at sustainability, defined as smaller number of vehicle kilometres travelled (VKT) and therefore lower number of fuel litres used to travel to the desired destination. The overarching research question of this thesis is:

- Can retail travel be made more sustainable, a case study of Brisbane?

Three sets of secondary research questions support the overarching question. The first set of questions looks at the existing retail environment and investigates the behavioural implications of this environment:

o What is the current form and structure of retail spaces in Brisbane? o What is the current travel behaviour of Brisbane’s retail shoppers? o Which transport modes are preferred for retail trips in Brisbane?

The second set of questions aims to model shopping destination choice and identify the factors that influence destination choice:

o How do Brisbane’s shoppers choose between available retail destinations? o Which factors affect shoppers’ destination choice? o Which types of destinations are generally preferred and why?

o What appears to be the likely future of retail spaces in Brisbane – given existing trends and probable future policies?

Chapter 1 Introduction 8

o How could planning policies potentially help to facilitate a shift towards a more sustainable future for retail environments in Brisbane?

To answer these sets of questions, this research will study the retail travel behaviour of trip makers to better understand customers’ mode choice and to identify factors which influence their destination preferences. The research will also aim to propose policies that could help to deliver a more sustainable retail travel future by reducing the vehicle kilometres travelled and increasing usage of greener modes such as public transport, walking and cycling.

Furthermore, the research will aim to challenge current trends in planning and policy making which have contributed to the present unsustainable retail trip pattern. It will also seek to provide a stronger platform for future development of retail structure which is more conducive to sustainability. The key objectives of the study can be summarised as:

 Examining current retail form and structure  Examining the current travel behaviour of retail customers  Modelling current shopping destination choice using actual retail travel data  Investigating and identifying factors which influence customers’ retail travel behaviour. These factors could include socio-demographic attributes of shoppers and/or spatial attributes of the retail environment and transport infrastructure which might be considered as barriers and enablers towards more sustainable retail travel behaviour  Predicting possible future trends that could affect retail trip patterns, such as additional expansion of existing large regional centres, an increase in the number of neighbourhood and small village centres, or the impacts of increased populations living within shopping centres’ catchments.  Modelling and forecasting subtle changes to retail structure, to identify what small but significant changes in planning policy may be delivered for low-density suburban areas, in terms of improved accessibility and travel behaviour

Chapter 1 Introduction 9

 Producing policy recommendations for future planning of retail environments which should support more sustainable forms of retail travel behaviour.

1.4 Justification of approach & structure of the thesis

Unfortunately, as the literature review will make clear, an absence of existing studies on retail function makes it difficult to identify a suitable starting point. A combination of qualitative and quantitative techniques will therefore be used to investigate various aspects of Brisbane’s retail environment in order to provide a foundation from which the research questions can be answered. These initial investigations will aim to clarify the current status quo in retailing, and identify existing problems embedded in the retail environment which affect peoples’ travel behaviour. The three sets of research questions will then be addressed in turn, following the structure shown in Figure 1.

Chapter 2 provides a literature review that identifies existing research gaps. While this research focuses on the sustainability of retail trips, the literature review begins by providing a comprehensive review of applicable existing studies in the field of economic geography and urban transport planning in retail environments. The literature review only focuses on studies that have implications for retail transport behaviours. It identifies existing methods of studying retail travel behaviour and factors addressed in existing studies, increases our understanding of the issue and identifies gaps in existing research.

Chapter 3 introduces Brisbane as a case study. The chapter explains why Brisbane provides an appropriate and relevant case study, and describes pertinent aspects of the city’s retail structure and regulatory planning.

The research framework and methodology are discussed in an overview in Chapter 4. The detailed methodological approach applied to answer each separate research question is discussed later in the relevant chapter, and considers the type and availability of data.

Chapter 1 Introduction 10

Chapters 5 to 10 describe the different techniques applied. Each of these chapters is written in the following format: introduction, technique-specific literature review, methods, results, discussion and conclusion. These chapters are intended to be self-contained.

Chapter 11 will review the findings of the previous chapters, identify the connections, possible interactions between them, and suggest future policies to shape a more sustainable retail structure, thereby returning to the overarching research question.

Chapter 1 Introduction 11

Fig 1-1: Thesis framework

Chapter 2 Literature Review

Chapter 2 Literature Review 15

2.1 Introduction

Drechsler in 2014 states that “Retailing is not only a major part of a property industry but it is also a major concern of the planning authorities. While planning’s attempts are to create more sustainable and effective urban structure which is being considerably affected by retail centres, property industry is looking for a more reliable market which will be able to return and benefit the investments” (Drechsler, 2014).

A large number of researchers have looked at different aspects of retail including its marketing aspects, trade area or the locational strategies adapted by investors, each of which contain various elements needed to be considered and studied in detail. For example, the location strategy of retailing is highly relevant to other factors including the competitive environment, closeness to other activity centres, mixed uses of land, ease of access to the city centre, socio demographic characteristics of customers (income, access to car, number of persons in household) and the level of services being delivered at each retail centre. Consequently, retail distribution has been the matter of investigation in a number of disciplines including economic geography, marketing and business, urban geography/planning and transport planning. Each of these areas has its own measuring criteria, desired priorities, and has therefore considered the retail environment from a different perspective.

‘Economic geography’ with its three fundamental economic activities, production, consumption and distribution of goods and services considers ‘location’ in the economy and the locational characteristics of production, consumption and the requirements that they have (Hanink, 1996). This approach is closely related to fundamental factors such as scale, level of function, level of accessibility and catchment populations that shape each of these destinations. When it comes to ‘Urban geography’ and ‘Urban Planning’, the focus is more on the residential spaces, transportation lines, economic activities, service infrastructure, commercial areas and public buildings and to study their characteristics and concerns (Pacione, 2005). ‘Urban planning’ is looking at the different aspects of the cities and communities, how they function

Chapter 2 Literature Review 16 and how the quality of lives within these cities or regions can be improved. The American Planning Association has defined planning as “a dynamic profession that works to improve the welfare of people and their communities by creating more convenient, equitable, healthful, efficient, and attractive places for present and future generations…Good planning helps create communities that offer better choices for where and how people live. Planning helps communities to envision their future. It helps them find the right balance of new development and essential services, environmental protection, and innovative change”. Considering these comprehensive definitions, retail, shops and services are a major element in both areas of economic geography and planning and play a major role in shaping or creating the environments and communities that can serve people and provide their everyday necessities, as well as creating mixed-use, liveable and vigorous communities.

Even though the literature shows different priorities, objectives, framework and methods applied for the study of each area, there are lots of interactions and communications between the two fields that need to be recognized and appreciated in order to enable researchers to better understand how the retail environment functions and where possible future shifts could be implemented. Besides, these approaches are unlikely to be effective if the customers’ desires and priorities have not been acknowledged within the context of the study.

One of the major problems that planners and developers are facing today is the lack of communication and understanding that results in the formation of policies or approaches that are not supported by all relevant groups. Planners put forward policies which business does not support, and developers are struggling with the existing regulations and waste lots of time and energy to get around them. Therefore, while the number of studies on retailing seems to be very substantial, there is hardly any research which simultaneously considers all the above-mentioned matters.

The focus of this research is on customers’ travel behaviour and those dimensions of the retail environment that can affect customers’ travel pattern, their trip mode choice and preferred travel distances. These issues require more serious attention by urban and transport planners. Therefore some

Chapter 2 Literature Review 17 dimensions including the spatial characteristics of the retail outlet, the availability of transport options, and socio demographic characteristics of the trip makers, will assume great importance, while other marketing factors such as price disparities and products’ brands which might also influence customers’ choice of destination, will not be investigated in this research.

Fig 2-1: Influential groups affecting the formation of retail structure in the city

This does not significantly affect the research findings since Brisbane, as the case study of this research, is experiencing a non-competitive retail environment and its market is highly monopolized by a few dominant enterprises which have created an environment in which only minor differences are discernable in terms of brands and prices across various locations in the city.

In conclusion, while the focus of this study is more on the transport dimensions of retail as a major element of city structure, different players significantly affect the formation and function of these destinations and, consequently, customers’ travel behaviour to those destinations. As a result, the literature will be studied

Chapter 2 Literature Review 18

in two major parts. The first section is dedicated to the major trends and studies which have been undertaken in the field of economic geography, considering how various businesses select their location and how they measure the number of the customers travelling to their establishments. The second section will look closely at how retailing has been dealt with in the area of urban and transport planning. The different urban form elements and their influences on travel behaviour will be investigated. The research gap will be then discussed considering the findings from the two major areas of retail studies in the literature.

2.2 Economic geography and retailers’ concept of spatial distribution of retailing

Since the early 20th century, different approaches have been applied and improved in order to measure retail catchment areas and understand the way people are attracted to, and move between, existing retail destinations located in different places. Examples include Reilly (1931) Gravitation law and Christaller’s Central Place Theory (1933) (Christaller, 1966). These studies were based on Newton’s Gravitation model and considered that factors such as distance and the population of catchment areas affect peoples’ destination choice (Bates, 2003; Roy and Thill, 2004). Since the 1960s, due to the growth in the number of large regional shopping centres, researchers have sought to explain competition between multiple urban regional centres and Central Business Districts (Roy and Thill, 2004). New techniques, such as Huff (1963) spatial interaction models, have developed to analyse retail market areas and to model consumers’ store choice processes. These models acknowledge the competitive and probabilistic nature of retail trips and consider the benefit that customers obtain by shopping at a specific retail centre (Yrigoyen and Otero, 1998a). About one decade later, discrete choice analysis based on disaggregated datasets, found its way into the study of retail outlet choice and made a considerable improvement in this area. These attraction models aimed to predict the probability of individuals choosing one alternative, among various discrete alternatives, based on the benefit that they derived from their visit.

Chapter 2 Literature Review 19

Discrete choice models can include additional factors, in comparison to earlier models, and are used when the dependent variable is discrete, such as travel destination or mode choice, activity participation, location choice, residential location choice and route choice (Athey and Imbens, 2007).

While the above mentioned studies can be named as some of the most significant turning points in the study of retail location, a broad range of other research also exists in the literature, either as an extension to these central models, or by proposing and establishing various new frameworks in this area. Some of the most important and influential approaches are briefly discussed here to frame our understanding of retail studies in the field.

2.2.1 The concept of trade area and methods applied for its measurement

2.2.1.1 Retail catchment area/Retail trade area

The ‘Trade area’ concept can be defined as a bridge between theory and practice which directs retailers’ strategies on the location and characteristics of retail centres (Kiel and Haberkern, n.d.). The concept is based on the similar idea in the field of geography, which tries to balance production, consumption and distribution elements and to maximize market benefits. Before dipping further into the methods of trade area measurement, a clear definition of ‘trade area’ is essential. The concept of a retail trade area has been used by analysts and practitioners in retail site evaluation and other market studies for a long time, and is a complementary procedure to site evaluation studies (Jones and Simmons, 1993). The term trade area attempts to define the most feasible locations for future businesses to be located, considering the actual visits to the centres by consumers. It would be of great help to the owner, manager or developer, to consider this key feature of their proposed location before developing their business. If the business has already been established, defining its trade area can be of great help to the owners, as a method of comparing their performance with competitors within the market; otherwise, it is of great help to decide future policies such as upgrading quality or strengthening advertising policies. Retail trade area analysis tries to find out if sufficient potential exists within the hypothesized trade areas to allow retailers to trade profitability (Kiel and Haberkern).

Chapter 2 Literature Review 20

The most common definition of a retail trade area is “that area, typically around the store, from which the store derives most of its patronage” (Dramowicz and Magazine, 2005). As mentioned above, different techniques have been used to analyse the market area for retail centres and to model consumers’ store choice. They range from simple ones, such as an application of rings, to more sophisticated ones, such as utilizing probabilistic trade area surfaces (Dramowicz and Magazine, 2005).

Various methods have been developed to measure the trade area of retail centres and the way they can be determined. Not all of them were developed within the inter-urban (within the city) context, but some were initially established to look at the intra-urban (between the cities) shopping movement and were later extended into inter-urban studies. Among these various methods, some rely on very simple assumptions of the function of the centres, and the perceptions of the costumers, while others apply very complex methods of representing retail trade areas’ formation. Some of these methods are discussed here.

Applebaum (1966) survey-based techniques, which have been used quite widely by planning and business people, rely on the real travel patterns of consumers. Applebaum uses surveys and locates the origin and distribution of respondents’ trips on a map, to determine a geographical pattern of destination centres’ trade areas. This method relies on real travel patterns instead of theoretical assumptions about travel behaviours of consumers. Primary catchment areas are typically defined as the areas where between 60 to 70 % of shoppers come from. ‘Simple radial trade area’ is another method presented which relies on behavioural assumptions about consumers’ travel time or travel distance. This approach assumes that the trade area is circular, with the radial distance from the store location predefined and ignores all the logistical barriers within this area that can cause significant changes in the destination choice of trip makers (Segal, 1999). While these methods seem to be quite simplistic, other methods have also developed over time, which tried to consider more influential factors such as Reilly’s gravitation model or the Central Place Theory by Walter Christaller.

Chapter 2 Literature Review 21

2.2.1.2 Reilly’s Gravitation Model - 1931

Going back to 1931, Reilly used Newton's gravitation law to look into the trade area of neighbouring cities at an inter-metropolitan scale rather than the trade areas of individual stores (Yrigoyen and Otero, 1998a). Reilly noted that: ‘the proportion of retail trade attracted from intermediate towns by two competing urban areas is in direct proportion to their population and in inverse proportion to the square of the distances from those cities to those intermediate towns’ (Bates, 2003). He presumed that the geography of the area was flat with no rivers, roads or mountains as barriers that affect consumers’ destination decisions, and consumers are assumed to be similar in all these cities. Reilly was a pioneer in promoting the gravity type of spatial choice models which are commonly used today (Yrigoyen and Otero, 1998a).

Fig 2-2: Reilly’s gravitation theory and the breaking-point concept

In his model, Reilly named the place of consumer demand as ‘i’ and the market centres or central places as ‘j’ and ‘k’ and came up with the following equation:

훼 훽 푝푖푗/푝푖푘= (푝푖/푝푘 ) (푑푖푘/푑푖푗)

Where 푝푖푗 and 푝푖푘 are the proportions of the whole of the trade from some place

of consumer demand i taken up by centres j and k respectively, 푝푗 and 푝푘 are

the populations of j and k, and 푑푖푘 and 푑푖푗 are the distances from i to the centres j and k; α and β are constants to be derived empirically. This equation defines the breakpoint between centres j and k, and is based on locating on which side of the breakpoint, the customers will pick centre j or k as their destination. “It gives us the breakpoint at which the proportion of trade flowing from i to j and k are equal” (Batty, 1978).

푝푖푗 = 푝푖푘

푑푗푘 = 푑푖푗 + 푑푖푘

Chapter 2 Literature Review 22

While α and β will be estimated as part of the model calibration, if we consider α equal to 1 and β equal to 2 then the equation would become:

(푝푗 푑푖푗 = 푑푗푘/(1 + √ ) 푝푘

Reilly’s gravitation model was the first approach to consider other elements, apart from distance, as factors which influence the attractiveness of a centre, and also to suggest that customers trade off the cost of travel against the attractiveness of alternative shopping opportunities (Yrigoyen and Otero, 1998a).

2.2.1.3 Central Place Theory (CPT) - 1933

The central place theory was proposed for the first time by Walter Christaller in 1933, to explain how some settlements of quite similar sizes and almost equal distances from each other are distributed in the flat landscape of southern Germany. Christaller applied geometric shapes to model this situation (Getis and Getis, 1966). This theory is still applied after almost 80 years as a major concept to support urban planning policies in several countries in the developed, developing and previously Communist worlds (Brown, 1995a). This theory is based on two major concepts; the distance that people are prepared to travel to get access specific goods or services, and also the threshold of demand which is an approximation for the minimum number of people required to elicit the supply of a certain type of good or services (Population and Distance). He believed that this system would result in the formation of a number of groups of central places with a similar hierarchy, based on their function. Each level of the hierarchy provides all the functions found in the lower levels of the hierarchy plus some additional functions that show the characteristics of that level of hierarchy, but not vice versa.

Chapter 2 Literature Review 23

Fig 2-3: Christaller’s ideal central place hierarchy showing nested hexagons. Each central place serves the market area of three smaller central places (Hanson, 1997)

Christaller noted that if each central place was assumed to serve a circular market area then this would result in either un-served areas or over-served areas. Therefore he proposed a hexagonal shape for the market area, thus ensuring that all parts of a region are served by central places. It is assumed that the price and comfort of trip is the same for everyone and that consumers would pick the closest centre as their destination, and that separate trips are made for each individual purchase (Brown, 1995b). Distance is the major focus of the model that highly influences customers’ choices while other factors, such as quality, choice, price, competitiveness, attractiveness of the centre, the centres' suitability for multipurpose trips, . are all ignored (Warnes and Daniels, 1979) (Dennis et al., 2002).

Chapter 2 Literature Review 24

Christaller’s theory is not capable of tracking other shapes of retail development which exist within urban areas such as ribbon developments (traditional shopping streets, urban arterial roads, and highway-oriented ribbon development), specialised functional areas (e.g. automobile rows, entertainment quarters, and concentrations of medical specialists) and big shopping centres and supermarkets in different form, size and function. It therefore finds it hard to describe the retail environment as a whole (Walmsley and Weinand, 1990).

Fig 2-4: Diagram of a classic central place system (market area: K=3 hierarchy) (Downs and DeSouza, 2006) p.91

Fig 2-5: Walter Christaller, Central places in Southern Germany, Prentice-Hall, 1966 (Downs and DeSouza, 2006)

Chapter 2 Literature Review 25

These limitations have widely restricted the capability, usefulness and application of the model, even though more recent authors like Dennis (2002) have suggested improvements which provide a more flexible and dynamic understanding of the location of central places. Dennis suggested that rather than defining the centres’ hierarchies based on measures of population or type of goods, measures of a centre’s attractiveness could be used (Dennis et al., 2002).

2.2.1.4 Huff Spatial Interaction Model - 1963

While the previous models consider the various factors influencing centres’ attractiveness and catchment area, they overlooked a significant element of ‘competition between the stores’ as one of the most influential factors in the retail environment. The number of people who travel to a specific centre is directly related to how successful that centre is compared to the other available ones (Dramowicz and Magazine, 2005). Huff, in 1963, for the first time argued that when consumers have several different options for their shopping destination, all available stores in the area will have a chance of being selected (Yrigoyen and Otero, 1998a).

Huff suggested that the market area of centres is more than a number of simple non-overlapping geometrical areas which was proposed in Central Place Theory. This was in contrast to the idea of market monopoly in the previous methods and considered the fact that different centres can share their market or catchment areas. Huff’s Spatial Interaction Model was developed to explain the complex, continuous and probabilistic trade area of the centres which affects customers’ behaviour. His proposed model could consider other factors such as floor-space and travel time rather than just population and distance and had the capacity to measure the probability of peoples’ choice between different stores (Bates, 2003). Thus the model could answer the question ‘what is the probability that a consumer will decide to shop at a specific store, given the presence of competing stores?’ by considering various factors including distance, attractiveness and competition (Yrigoyen and Otero, 1998a).

Chapter 2 Literature Review 26

∝ 훽 푈푖푗 푆푗 퐷푖푗 푃푖푗 = = ∑퐽 푈 퐽 ∝ 훽 푘=1 푖푘 ∑푘=1 푆푘 퐷푖푘

where:

Pij: probability of consumer at “i” visiting store j (or town j); J is the set of competing stores (or towns) in the region. Uij: utility of store (or town) j for individual at “i”. Sj: size (square metres) of outlet j (or set of outlets of town j) Dij: distance between consumer at “i” and store (or town) j. , : sensibility parameters; in line with Reilly’s Law, = 1 and = -2.

2.2.1.5 Later improvement in the store’s trade area model

A large amount of research has been tested and developed around the basic concepts of the Gravity model, Christaller’s CPT and the Huff interaction model, including Thiessen Polygons, the McFadden (1980) multinominal logit model or Fotheringham (1988) Competing Destination Model. About one decade after Huff proposed his interaction model, McFadden applied the multinominal logit model in his work to include the total benefit that one person can get from a specific store when considering his or her options. Using the discrete choice model to investigate customers’ trip mode share has been followed ever since by a large number of researchers applying various types of logit models to represent choice behaviour in a very wide variety of settings. These attraction models aimed to predict the probability of individuals choosing one alternative among various discrete alternatives based on the benefit that they derived from their visit. McFadden initially used these logit-based models in the 1970s and 1980s to understand the transportation mode choices of different types of trips, including retail trip-makers decisions (Athey and Imbens, 2007).

Later in 1983, the Competing Destination Model (CDM) was proposed by Fotheringham. Fotheringham’s model was an attempt to overcome the frequent criticism that gravity type models were biased in their estimation of distance decay parameters and in the way that distance was assumed to impact on people’s behaviours and spatial interactions. This issue has been referred to as the ‘spatial structure effect’. ‘The theory believes that spatial decision makers adopt a hierarchical strategy, that is, an individual who wants to make a trip first chooses a broad region with which to interact, and then chooses a specific

Chapter 2 Literature Review 27 destination within that broad region. Essentially, the hierarchical information processing approach is a compensatory decision strategy which means that good features of an alternative can compensate for bad features of another alternative’. Therefore, by removing the undesirable alternatives, decision makers limit their choice set. This suggests that the trip distribution of people will be directly influenced by the spatial arrangement of destinations in a geographical system (Hu and Pooler, 2002). Fotheringham believed that the negative weight that was allocated to the distance in previous models was incorrect. People are not always negatively affected by the length of their trips; instead they might pick a more distant destination instead of a closer destination for various reasons.

All these different spatial approaches, related to the design of the retail trade areas, were later classified by Yrigoyen (1997) into two major categories of descriptive-deterministic and explicative-sophisticated approaches (Fig 2-7) (Yrigoyen and Otero, 1998a). While ‘measuring the centre’s catchment area and attraction value’ are central themes in most of these studies, they do not seem to be a convincing approach that drive the consumers’ decision making process. There are still lots of restrictions which might reduce the reliability of the results. Measuring centre’s catchment area is a very biased process, considering that evaluating the attractiveness of a centre and defining its deterrence factors in each case, is not a simple and easy task. Qualitative and quantitative aspects of retail environment are not easy factors to estimate positively or negatively in many cases (Kiel and Haberkern).

While researchers attempt to find surrogates that can represent various dimensions of attractiveness, lack of data, or the differences arising from a different context make it a very difficult thing to do. Dennis et al. (2002) suggested that various factors could be applied as a proxy for the attractiveness of the centre. He refers to Guy’s retail sales floor area as the chosen proxy in his study on retail hierarchy in Cardiff. In addition, Bates (2003) discussed that while there are elements of attractiveness that cannot be easily approximated

Chapter 2 Literature Review 28 as an exact coefficient in an equation, it is possible to use the ‘turnover values’1 as the attractiveness factor instead of the complicated retail methods and models. He argued that turnover values can be (inversely) representative of the quality of the centre, the number of customers, the ease of access and many other influential factors that might not be easily recognized and measured in various models (Bates, 2003).

A wide range of approaches, discussed briefly here, have been developed to study the retail environment from an economic geography perspective regarding issues such as production, consumption and increasing business profit by delivering products and services to customers. What is important to consider is that, focusing on the travel implications of this retail distribution, there are many other aspects of the retail environment that have not been addressed in these studies, including urban form elements and their impacts on travel preferences of customers.

Considering the focus of this study, the transport impacts of retail structure and how this might affect customers’ travel behaviour, is a major issue that merits a careful review of the existing literature.

Therefore, the next section of the literature review will briefly look at retailing in the context of urban form and transport planning studies, to discuss existing approaches, their benefits and shortcomings.

(1) The annual sales volume net of all discounts and sales taxes

Chapter 2 Literature Review 29

Fig 2-6: Spatial models and methods applied to the design of retail trade areas (Yrigoyen and Otero, 1998a)

Chapter 2 Literature Review 30

2.3 Urban form, retail form and travel behaviour

As previously discussed, urban and transport planning can be viewed as a challenge to improve people’s quality of life, by improving their experiences of the urban environment and creating more sustainable communities. The genesis of sustainability goes back to the early seventies but the most frequently cited definition of ‘sustainable development’ was stated by the World Commission on Environment and Development in the “Brundtland Report “as “development that meets the needs of the present without compromising the ability of future generations to meet their own needs”. While this definition is quite broad, complex and multifaceted, “The future that we want”, report by the United Nations refers to transportation and mobility as central elements of any sustainable development. Improving transport and mobility can enhance economic growth by better integration of the economy, improved accessibility and at the same time respect the environment (UN General Assembly, 2012).

Doust and Black (1996) in their sustainability framework refer to ‘urban form’ and ‘Transportation’ and interactions with communities as the central drivers of sustainability in cities and raises them as the elements which define the structural configuration of the city. They believe that the ideal planning goal in order to achieve a sustainable community should be to have high accessibility with a low environmental footprint. This requires better management of interactions between urban form and transportation. Different approaches such as mixed-use development, transit oriented development and smart growth have developed from this concept. These approaches, have tried to reduce the number of car trips, to improve public and active transportation performance and, as a result, to reduce obesity, reduce the level of environmental pollution (air and noise) and improve individuals’ health.

So what is the urban form that is considered to be the central driver of sustainability in cities? Urban form at its simplest can be described as a city’s physical characteristics. Urban form is the combination of all urban elements, comprising building shape and design; it includes character and identity, built streetscape and urban structure, the physical layout through the subdivision pattern, public and private space, the interrelationship of activities and functions

Chapter 2 Literature Review 31 of centres or towns, open spaces and recreational areas. Urban form is also characterised by the density and spatial distribution of land-use (Doust and Black, 1996). The terms ‘Urban form’ and ‘land use’ have been used interchangeably by many researchers, although some researchers believe that land use is only one elements of urban form (Handy, 1996a).

Three major elements of urban form, including population density, design (street patterns) and diversity (land use mix), (commonly referred to as the 3Ds) (Litman, 2005), are believed to be the most important determinants of accessibility levels in different areas (Stead & Marshall, 2001). It is generally accepted by planners that the spatial distribution of land-use is a key factor influencing the travel behaviour of residents (Limanond and Niemeier, 2004). Accessibility, as the ease of reaching desirable destinations, is the concept that connects land-use planning with sustainable transport policies. It is a key concept in transportation planning that is made up of mobility and proximity (Cervero, 2005). Accessibility can occur through increased car access, cycling, walking and public transport or a mixture of all these options at the same time. Over the past several decades, while people in urban areas were becoming increasingly reliant on private motor cars to travel within the city, even for short distances, a number of different planning approaches (New Urbanism, Transit Oriented Development (TOD), Smart Growth, Neo-traditional development, Urban consolidation) have begun to halt this trend and its detrimental impacts on humans’ health and welfare. These approaches have advocated compact, mixed-use development in an attempt to shorten travel distances and to encourage travellers to cycle and walk instead of driving (Cervero, 2005). Studies of similar approaches, such as Aldous (1992); Calthorpe (1993); Ryan and McNally (1995); Urban Task Force, 1999, aimed to use land-use policy and urban design to promote more sustainable patterns of travel (Frank and Pivo, 1994a; Ewing, 1995; Cervero and Kockelman, 1997a; Meurs and Haaijer, 2001; Stead and Marshall, 2001; Cervero, 2002).

The number of studies dealing with the relationship between urban form and travel behaviour is huge. The spatial scale of these studies varies from strategic to local (neighbourhood level). They consider different aspects of urban form,

Chapter 2 Literature Review 32 such as distance of the residence from the urban centre, settlement size, mixing of land uses, provision of local facilities, density of development, proximity to transport networks, availability of residential parking, road network type and neighbourhood types (Stead and Marshall, 2001). Some of these studies are discussed briefly here to describe some of their key findings.

Boarnet and Crane (2001) studied the impacts of land use pattern and urban design on the travel behaviour of people in Orange County and San Diego, using a two-day travel diary. The results showed that while the geographical scale is always very important, land use and urban design can affect people’s travel behaviour by affecting the price of their trips. These authors, however, believe that the relationship between urban form and travel behaviour is far more complicated; many other factors such as individual preferences and the availability of other choices, can also affect people’s decisions (Boarnet and Crane, 2001). Handy and Clifton in their study of six neighbourhoods in Austin Texas, studied the value of the local shops. The finding showed while local shops do not show any significant change in reducing the total driving, it provides the residents with other options and increase that quality of life {Handy, 2001 #338}. In another study by Cervero (2002) in Montgomery County, Maryland, a high level of dependency between mixed land use settings and travel behaviour was found, to the extent that the level of mixed use can influence people’s decision to drive alone, share a ride, or use public transport (Cervero, 2002).

In contrast, Crane and Crepeau (1998) expressed a totally conflicting idea, finding that changes in the land use were not substantially related to car trips or to lower car mode splits (Leck, 2006). The commuting pattern, as another element of urban form, connected to job and residential locations in San Francisco showed a negative relationship between job to residential ratio and the proportion of journeys undertaken by foot and cycle (Cervero, 1989). Where there are many more jobs than houses, the proportion of journeys by foot or cycle falls (Stead and Marshall, 2001). Kulash et al. (1990) and McNally and Ryan (1992) believed that road pattern and moving the trip origins and destinations closer together, as achieved in a grid system, for example, , would

Chapter 2 Literature Review 33 reduce trip length, even though the effect on the number of trips was unclear (Crane, 1996).

It can be seen that the role of mixed land use or the availability and accessibility of destinations, have generally been found to be a significant factor in these studies; while there are conflicting ideas on how influential these factors appear to be on travel behaviour (Nelson et al., 2001).

Whereas the above mentioned studies are mainly focused on separate environmental measures, such as housing density, land use mix, street type, job numbers, or the presence or absence of pavements, the combined effects of groups, or ‘bundles’ of environmental measures on travel behaviour, has been addressed in recent studies using ‘Meta-Analysis’. These studies have mainly developed in order to respond to contradictory findings in the field. Meta- analysis, as a statistical research technique, combines the quantified results from several studies into one overall effect. It frequently does this by taking an average or a weighted average of the elasticities or effect sizes from individual studies. Only a few of these studies can be found in the field of built- environment and travel behaviour (Leck, 2006). A meta-analysis study carried out by Ewing to investigate the relationship between the built-environment and physical activity levels, showed that every 1% increase in density or design is generates a 0.45% increase in walking trips. The aspects of urban form which were considered in the study included land use density, diversity and design (Leck, 2006; Ewing and Cervero, 2010).

Ewing and Cervero (2001) tried to estimate the elasticity for Vehicle Kilometre Travelled (VMT) and vehicle trips based on the results of other published studies. They used four measurements of the built environment: density, diversity, design and regional accessibility. The results of the meta-analysis showed a statistically significant, but rather weak, connection between separate urban form variables and travel behaviour. Although the elasticity value seems small, they claimed that the cumulative effects of all these variables on urban form are relatively large (Leck, 2006; Boarnet and Handy, 2010).

It is noteworthy that although studies of the relationship between travel behaviour and the impacts of land use have been carried out for more than half

Chapter 2 Literature Review 34 a century, there is still very little agreement on the effect of built environment on travellers’ mode choices (Frank et al., 2008).

Among numerous studies on urban form and travel behaviour interactions, only a small portion of research directly focusses on shopping trips. This might be because researchers consider there to be few differences between factors that affect travel behaviour to retail destinations compared with travel behaviour to other types of destinations. There is a large body of literature that looks at the relationships between retail land use characteristics and how residents travel to access their required products and services. It is widely believed that access to retail destinations can impact trip frequency, destination choice, trip complexity (Hanson and Schwab, 1987) and improve the quality of life (Iacono, et al., 2010).

The literature has considered several spatial and socio-demographic variables that might affect retail travel behaviour. While various studies revealed that spatial factors such as the level of mixed use in the neighbourhood (Limanond and Niemeier, 2004) (Limanond et al., 2005), accessibility to retail destinations (Hanson and Schwab, 1987) (Robinson and Vickerman, 1976) (Recker and Kostyniuk, 1978) (Stead and Marshall, 2001), number of available retail opportunities (Handy, 1996b) (Niles and Nelson, 1999a), street layout (Crane and Crepeau, 1998), and travelled distances (Krizek and Johnson, 2006) significantly affect travel behaviour, other studies failed to find such a significant relationship (Frank and Pivo, 1994b; Crane, 1996; Crane and Crepeau, 1998).

Socio-demographic characteristics of individuals such as car ownership, household size, age, and customers’ background, are considered to be the most important factors influencing customers’ travel behaviour, or form an important part of their behavioural perspective (Hubbard, 1978; Hanson and Hanson, 1981; Carlsson-Kanyama and Linden, 1999; Cubukcu, 2001; Bromley and Thomas, 2002; Curtis and Perkins, 2006; Ryley, 2006)

While the role of the built environment, including retail structure is considered to be significant in some research, it is important to remember that the results are based on observations of the behaviour of people who are already living in a specific location, who have already made a decision to live in that

Chapter 2 Literature Review 35

environment. Therefore, a decision about whether they want to drive or not can be the result of their personal choice rather than the built environment’s impact. There are only a few studies on this topic such as the one carried out by Cao et al. It suggests that residential location plays a more important role in affecting driving behaviour than does residential self-selection in metropolitan areas - although the personal choice factor can still be significant by itself ((Cao et al., 2009),(Cao et al., 2010)).

2.3.1 General methods of studying interactions between urban/retail form and travel behaviour

The amount of research on the relationship between urban form and travel behaviour is quite considerable. This has led some studies to categorise this literature based on the methods applied. Crane (1966) suggested categorizing these studies by: their trip purposes (journey-to-work and shopping), the analytical method used (simulations, and regressions), the nature and level of detail in their applied data (aggregated or disaggregated), the measures of urban form used (street layout, composite measures of density, mixed use, and pedestrian features), the choice of explanatory variables (travel costs, travel opportunities, and socio-economic characteristics) and finally the Compact and dense residential and employment development (Leck, 2006).

Handy (1996) suggested to classify the studies on the relationship between urban form and travel pattern, based on their associated methodological approaches. She identified five categories of studies: aggregate and disaggregate analysis, choice models, activity-based analysis and simulation studies (Handy, 1996a). These categories cover almost all the existing methods applied in the field of urban and transport planning.

While each of these categories can be useful for the study of some aspects of the retail travel behaviour and its multidimensional nature, the advantages and disadvantages of each category should be considered, prior to apply in particular method for a specific study. The features, advantages and disadvantages of the different approaches are considered briefly in the following subsections.

Chapter 2 Literature Review 36

2.3.1.1 Aggregate and disaggregate analysis

Aggregate analysis is one of the most common approaches used to study trip characteristics in a region. These studies are often carried out at the scale of traffic analysis zones (TAZ) or other municipal divisions (Patterson et al., 2010) by creating an average number for the individual behavioural units in each of these divisions (Richards, 1974), such as the average level of income calculated for a zone, rather than handling individual income. In these types of studies, the location of the individual origins and destinations in each zone is usually ignored by the analyst (Richards, 1974). Since only the differences between various zones can be explained by the aggregated or zonal analysis, a large part of actual variation in travel demand behaviour will remain unnoticed (Richards, 1974) without understanding which urban form criteria most significantly affect people’s decisions to travel in that specific way (Handy, 1996a).

These types of analysis also lack precision in connecting the results with the location. Sometimes the zones are too big or heterogeneous and inconsistent, so that restricting data and analysis to the scale of the zones will not create meaningful results. It should be remembered that aggregate data are always easier and less expensive to collect and can include variables such as income that are often too sensitive to be collect as disaggregated data (Shaw & Wang, 2000).

In contrast, disaggregate analyses is based on individual observations and can be a far more reliable approach for analysing travel behaviour (Richards, 1974). Disaggregate data are usually collected for different members of different households and include information on socio-demographic characteristics such as age, sex, income, household size, and number of cars.

While aggregate analysis can be useful for testing the impacts of land use policies for issues such as decreasing automobile dependency, they are not capable of revealing the influences of urban form characteristics on individual behavioural decisions (Handy, 1996a). Aggregate analysis mainly tries to establish the strength of relationship between travel pattern (trip frequency, average trip length, mode split or total automobile travel percentage) and

Chapter 2 Literature Review 37

elements of urban form. Simple comparisons, correlations and regression procedures are generally used to establish these relationships. When it comes to disaggregate analysis, the focus of analytical regressions and logit models is the impacts of urban form elements on the travel behaviour of individuals with different socio-demographic characteristics (Handy, 1996a).

2.3.1.2 Choice models

One benefit of having access to the disaggregated data is to enable the use of more detailed and specific choice models. As explained previously, these models look at the probability of an individual choosing a particular alternative from the set of available alternatives based on the level of utility delivered to the choice maker. These models can address household or individual destination choice, mode choice or a mix of both. Handy remarks that “This approach has a stronger theoretical basis than the previous approaches and comes closer to directly testing causal relationships” (Handy, 1996a).

Disaggregate multivariate studies allow for simultaneous modelling of the varied factors thought to influence travel decisions where the decisions have already been made. “Choice models” take this approach a step further, by explicitly expressing individual-level decision models in a theoretical framework relating to the manner in which an individual goes about making such decisions. Discrete choice models view travel decisions, such as mode choice model and trip frequency, as efforts by an individual to maximise his or her utility by making trade-offs between available resources such as money, vehicle, time and the costs associated with a particular type of trip. Except for the commuting trips, transportation choice models have seldom included urban form factors. However, over the last two decades, microeconomic theory has been applied more frequently to model non-work trips, with an emphasis on possible influences of the built environment (Reilly and Landis, 2002).

While this approach has not been used very broadly in terms of revealing the link between urban form and travel behaviour, it could potentially be of great importance in this area. Such models are useful to policy makers because they can give an indication of how individuals might react to new sets of choices (Handy, 1996a).

Chapter 2 Literature Review 38

2.3.1.3 Activity-based analysis

Shifting from using aggregate data to disaggregate data was a big step forward in transportation studies, but the disaggregate approaches described so far have still been trip based (Handy, 1996a). In an activity-based approach, it is believed that people do not make separate decisions considering only one trip but instead they try to schedule their activities in a daily pattern. This type of approach is also termed tour-based analysis (Maat et al., 2005).

The activity models look at the total pattern for an entire day and the interdependencies between individual trips. This trend looked at travel choices based on the types of activity people wanted to participate in instead of regarding them as separate trips. Rather than focusing on particular travel characteristics, or even a set of simultaneous decisions, activity-based analysis often looks at the total pattern for an entire day. Relationships are not always tested statistically, but may be evaluated qualitatively.

Another fundamental difference between the trip-based approach and the activity-based approach is the way in which time is conceptualised and represented. In the trip-based approach, time is simply a ‘cost’ of making a trip. The activity-based approach, on the other hand, treats time as an all- encompassing continuous entity, within which individuals make activity and travel decisions {Bhat, 1999 #339}. Activity participation can thus be seen as a matter of time allocation (Pas, 1998). This means that individuals do not maximize utility for separate travel choices, but rather optimise their entire activity pattern (Maat et al., 2005).

Obtaining the required data for these types of models is very challenging and expensive. Urban form factors have usually only played a secondary role in these types of analyses.

2.3.1.4 Simulation Studies

Simulation models are usually used as a tool to forecast travel demand under different scenarios in the future. These different scenarios rely on specific assumptions about the relationship between urban form and travel behaviour (Handy, 2002). Therefore, these models involve a hypothetical testing of urban

Chapter 2 Literature Review 39

form according to the assumptions about development factors (Yigitcanlar, 2010). The focus of these models is on the overall structure of the city or metropolitan area in terms of the residential or employment distribution and the infrastructure network. While these models can be of great value in providing general understandings of future travel demand, they are not considered to be a reliable tool for identifying the relationship between urban form and travel behaviour.

These studies are helpful in broadening our understanding of the overall travel pattern of a city or a study area, but cannot directly address the relationship between urban form and travel behaviour (Handy, 1996a).

2.3.1.4.1 Travel Demand Models

Different types of transport models have been developed and are being applied to evaluate planning scenarios and to illuminate the effects of transport planning decisions. Transport models could predict the impacts of different possible changes in the network system, including the introduction of alternative transport models and price shifts. Some of these transport planning models are termed “Travel Demand Models”, “Trip and Parking Generation Models”, “Economic Evaluation Models”, “Integrated Transportation and Land Use Models”, “Energy and Emission Models”, “Simulation Models” and “TDM Program Models”.

Among these various models, the “Travel Demand Models” are applied to evaluate travel demands under specific existing or proposed land use policies, transportation prices and available transit services. Travel demand models aim to predict network traffic volumes and often pollution emissions. The ‘four step model’ is the most common model of this type. It, includes the four steps of trip generation, trip distribution, mode split and route assignment.

Four step models use travel surveys and census data to set their baseline conditions, and construct travel demands under separate categories such as work and shopping, which will later be aggregated into total trips on the network (Litman, 2014).

Chapter 2 Literature Review 40

While these models can be very useful for predicting the impacts of land use shifts and planning provisions on travel behaviour, and for studying mode choice percentage, a number of limitations influence their accuracy and reliability.

Litman (2014) described some of these limitations as: (1) no explicit modelling of trip chaining; (2) a primarily focus on vehicle trips; (3) limited or no modelling capability for transit, walking and cycling; (4) fixed vehicle trip rates by land-use type regardless of the design of the development; (5) zonal aggregation of traveller characteristics; (6) large traffic analysis zones; (7) a focus on primarily modelling peak periods, and (8) the traditional four-step model processes do not capture any increases in shorter intra-zonal automobile trips, bicycle trips and walking trips, that might be encouraged by smart-growth strategies. This last limitation arises from limitations of the four-step process in modelling intra-zonal trips and travel by modes other than automobiles. There is also the problem of lack of consideration for influential qualitative information in these types of models and the simplistic use of a travel time cost function to predict mode shifts (Litman, 2014).

2.4 Discussion of research gaps in the literature

The review of the existing literature, both within the fields of economic geography and transportation planning, reveals that it is not a straightforward task to assess retail travel behaviour. The number of factors which are involved in the form of retail destinations, how they function and the way consumers travel to them, is complicated. A host of tangible and intangible factors, including retailers’ preferred catchment area, the form of retail centres, the mixture and level of services being provided, the built environment, socio demographic characteristics of trip makers, matters of individual self-selection, location and scale of the study, etc. can influence how travellers choose between these destinations. Besides the type of data and methodology applied could significantly affect the studies’ findings.

The literature review reveals that there are significant number of studies on the impacts of single or multiple elements of the built environment on travel

Chapter 2 Literature Review 41 behaviour of people especially in U.S. including mixed of land uses in the residential developments, street design/pattern, number/location of the shopping districts, etc. or what is reffering to in general as the 3Ds in the urban palnning/design literature (Density, Diversity and Design) (for example studies by Handy (1996a), {Handy, 2001 #338} Schneider (2011), Boarnet and Sarmiento (1998), Ewing and Cervero (2010), Cervero (2002)). The objectives of these studies are mostly to understand how these different factors can affect the travel behaviour of people and therefore much of this research have been limited to specific neighbourhood shops or shopping centres rather than comprehensively assess of regional travel for shopping and in many cases ended up in proposing new urban design small scale recommendations. On the other hand, some more recent studies focused more on the suburban versus urban (or the traditional versus modern) environments. These studies compare examples of some existing type of traditional and modern neighbourhoods and investigate how travel behaviour of people is different within each contexts (see Handy (1996a)).

Prevailing studies in the field of economic geography on the other hand (see Christaller (1966), Fotheringham (1988), Huff (1963), Reilly (1931), Hu and Pooler (2002), Cubukcu (2001)), covers a wider number and variety of the aspects of shopping trips besided more complex methods to study the probability of people’s behaviour under varous circumstances (using logit choice models, etc) but the focus of the studies are on the supply-demand and the retail trade area of the shopping districts which in many cases in not in line with the goals of reducing car trips and encouraging other modes of transport.

Looking through the existing literature on both side, four major categories of influential factors have been identified and studied in various research (Fig 2-7):  Destination attributes of retail outlets - characteristics of retail establishments  Individual attributes of trip makers (customers)  Attributes of trips made to these retail destinations, reported in terms such as distance, time and cost

Chapter 2 Literature Review 42

 Spatial attributes of the retail destinations’ surroundings such as the distances between the centres and catchment population

Fig 2-7: Different categories of attributes which might affect retail travel behaviour

What is obvious from the literature review is the lack of research that attempts to deliver a comprehensive understanding of how and to what extent these different factors together can affect the way people travel to these destinations. Little research has been completed that systematically understand the potential to reduce the number, length or the mode of transport for retail trips, and could therefore result into future policies which encourage more active travel and reduce the VKT in general.

While different methods are required to study these broad range of factors, the application of any of these methods is accompanied with a number of benefits and shortcomings that needs to be considered and addressed. Some of these benefits and deficiencies have been summarized in table 2-1. Besides, the limitations and restrictions in regards to the availability of data, might also noticeably affect the scope and preciseness of research on retail environment. Considering the limitation and potentials of these methods and the available data, a number of different methods (which will be explained in chapter 4) haveare selected and applied to help to achieve the studies’ objectives.

Chapter 2 Literature Review 43

It is also important to mention that Australia is considerably lagging behind in the study of retail environment and its transport implications.

Table 2-1: Advantages and disadvantages of commonly applied methods to study the linkages between urban form and retail travel behaviour

Method Advantages Disadvantages Aggregate Uses data which can be collected Ignores exact trip origin and Analysis relatively easily and cheaply destination, specific trip attributes & Useful to look at overall travel also the characteristics of individuals patterns in different zonal area Lacks precision in connecting results with the locations Incapable of investigating actual variations in travel demand behaviour Incapable of revealing the influences of urban form characteristics on individual behavioural decisions Applies a trip based approach and ignores the complexity of trip chain Disaggregate Uses individual observations Applies a complex trip chain Analysis A reliable method for analysing approach travel behaviour and the impacts of urban form Includes information on characteristics of trip makers Choice Models Uses disaggregate data and is therefore capable of looking at the influence of more detailed individual characteristics Capable of estimating the probability of an individual choosing a particular alternative from a set of available alternatives. Stronger than the previous approaches in its ability to identify casual relationships Activity Based Look at the total pattern of an Complicated data gathering and models entire day and the modelling approach which makes it interdependencies between expensive to develop individual trips Looks at travel choices based on types of activity Simulation Can provide a general Not a reliable tool for identifying Models understanding of future travel relationships between urban form demand and the overall travel and travel behaviour due to pattern in the city imprecise assumptions regarding relationships between land use and travel behaviour

Chapter 3 The Case Study Settings

Chapter 3 The Case Study Settings 47

3.1 Why Brisbane?

Brisbane as the third big city in Australia is experiencing a sudden and rapid growth in its population and urban footprint. According to the Australian Bureau of Statistics (ABS) the current population of Australia is about 23 million and it has been projected to be more than 28 million by 20311. Due to the South East Queensland (SEQ) regional plan, in Brisbane, 156,000 more dwellings are going to be accommodated, raising the population from the 2006 level of 991,000 to almost 1,270,000 in 2031. This population will not only create many more trips for commuters and school children, but will also increase the population’s daily/weekly trips for retail services and recreation. This is happening while at the same time Brisbane is grappling with the domination of the private car.

Looking at the Travel Demand Management program (TDM), many studies have been undertaken to better manage commuting and educational trips but not enough attention has been paid to the trips for retail purposes. Brisbane is selected as the case study for this research for several critical reasons:  It is an Australian city, which is experiencing fast expansion in its population and urban footprint.  As a result of the urban sprawl, the modern retail components are glaringly obvious in this city.  There are a great many two purchaser households.  The shopping layout of the city is mainly based on big shopping centres, supermarkets and category killers, as elements of modern retail structures.  New forms of shopping outlets, such as DFOs (Direct Factory Outlets) and large standalone dedicated warehouses are expanding in the city.  The retail pattern and form of the city fits Christaller’s model (one of the major geographical models for retail environments developed in 1933) including a set of standalone malls. These malls are in different levels of

(1) The information was checked on June 19th 2015 from the following website: http://www.abs.gov.au/ausstats/[email protected]/0/1647509ef7e25faaca2568a900154b63?opendocument

Chapter 3 The Case Study Settings 48

hierarchy from super-regional down to regional and sub-regional centres, and all are located spatially in almost similar distances around the CBD.  The existence of the basic database required for the study of people’s travel behaviour (Household Travel Survey Dataset 2004/ 2009) as part of a cyclical data gathering process that occurs every few years in Brisbane.  The availability of strategic transport model (BSTM-MM) developed for South East Queensland to forecast future travel behaviour shifts, based on possible proposed scenarios. It covers various modes of private car, public transport, cycling and walking and is being updated almost every year.  The ease of access to key policies, officers and relevant agencies, and possible access to key public and private sector operators for the interviews planned later in the research process.

3.2 Identifying Brisbane’s retail structure

The major goal of this research is to look at sustainability, defined as smaller number of vehicle kilometres travelled (VKT) and litres of fuel used to travel to the desired destination. Therefore, finding and encouraging forms of retail space that decrease car dependency has become one of the major challenges that needs to be swiftly addressed by urban planners (Goodman and Coiacetto, 2009). A lack of foresight in regards to this challenge can cause serious deficiencies in achieving the ultimate future goal of sustainability in developed countries, including Australia.

Underestimating the number of regular car trips travelling to retail destinations, and giving in to the assumption that cars are the only way of accessing these destinations, will result in more congestion, a more polluted environment and a less healthy and more obese population, as well as a high level of waste in the communities’ time and resources.

Besides the socio-demographic, economic, and trip comfort factors which might influence people’s travel behaviour for both shoppers and employees, the literature review (Chapter 2) showed that there is significant interest in the form,

Chapter 3 The Case Study Settings 49 structure and distribution of retail establishments. Therefore, it is essential to have a comprehensive understanding of the status of retail environments in our case study to make sense of current and possible future trends.

The first step towards modifying the currently unsustainable structure is to understand the retail situation and the various phases that it has gone through during the last few decades, which has resulted in the current car reliant environment. The next step is to reveal more about the possible approaches that can make changes in the future structure of the retail environment.

3.2.1 History of retail locations

Historically, most of the urban form in Australian cities was shaped on accessibility routes including the rail and tram systems. Brisbane’s metropolitan retail network has over the past few decades extended radially from the central business district (CBD) in response to population and transportation changes (1990). Initially, retail and shopping developments were located at the centres of the towns or villages. In the 1800s, the city started to experience suburban expansion along the main routes coming into and out of the city. This was followed by expansion of general stores along these transport routes which provided highly accessible public transport options or gave people the opportunity to walk to their destinations (Warner, 2013).

The tram system that operated between 1885 to 1969 in Brisbane played an influential role in directing the retail locations before cars become a popular means of transport in the 1950s and 1960s (Warner, 2013). In the mid-1900s, as the use of the private car emerged as a popular and easy accessible mode of transportation for every household, the shape of the cities began to change and roads/corridors and highways became the main tools for measuring distances in the cities. New, scattered development started to appear and simultaneously, new types of retail spaces and shopping centres were built to service people’s needs.

3.2.2 The beginning of the shifts

The decentralization process and the transformation of retail spaces into planned shopping centres began in the US. Between 1950 and 1980 the

Chapter 3 The Case Study Settings 50 number of these centres increased from 100 to 22,000 in a process known as the “Malling of America” (Bromley and Thomas, 2002). This trend was followed in Europe between 1972 and 1974, most noticeably in France, with a large increase in the number of planned shopping centres. This period was known as the “Golden age” by some (Bromley and Thomas, 2002).

Changes started to happen in the retail environment in Australia, including in Brisbane, in about the 1950s not with a considerable delay from its pioneer, US. The geography and form of retail land uses in Australian cities, specifically between the periods of 1960s to 1990s, shows the considerable changes in this area and reveals the supportive planning policies behind the development of car-based shopping centres, which expanded widely in number within this period (Scott, 2002 ) . In considering the trends happening in the US, Australia and France, in 1985, Dawson concluded that “the shopping centre over the past thirty years has become an established feature of urban structure in countries with widely divergent urban policies”. Other countries with rigorous and restrictive land use planning systems, such as the UK experienced a slower and more regulated trend (Bromley and Thomas, 2002).

Factors influencing the retail form of Australian cities changed over time with transformations in the overall urban form of these cities. The period of shift in socio economic characteristics of the population, household mobility expenses, and technological changes in the nature of flows, economic growth, etc. all had a serious impact on people’s attitudes toward shopping (Baker and Wood, 2010).

In the new environment, more affluent double purchaser households could afford to buy automobiles and spend more money. The developing technology gave them access to white goods, including refrigerators to keep perishable products for a longer period and to buy these products in larger quantities. These affluent households were also looking for bigger houses with gardens and abundant space for their children, which was not possible within the urban core of the city. Brisbane, with its large amount of available land for growth and easy access to cars, gave these households the chance to live further out of the city centre. Today, Brisbane is experiencing rapid growth in its urban fringe and

Chapter 3 The Case Study Settings 51 many outer suburbs are comprised of large lots of up to 500/600 m2 with detached houses.

The sudden and extensive accessibility and ownership of private cars, the increased level of mobility and the large boost in the construction of new roads and highways in the mid-1900s changed the face of the city and the function of retail environments. It provided the opportunity for larger retail outlets to spread. This changed the ‘centralized retail model’ accessible by foot for the majority of people to a shopping centre model mainly accessible by car (Bromley and Thomas, 2002; Baker and Wood, 2010; Warner, 2013).

This new widespread growth led to low density housing areas incapable of supporting a high street and corner shops. Instead, this combination of housing development, transport mode, transport infrastructure and Brisbane’s climate created a demand for huge, enclosed, car-based suburban shopping centres. Walking to shops or using public transport was no longer a feasible option. Over the past several decades, shopping centres and big boxes have continued to expand in both in number and market share, while traditional retail forms have experienced a sharp decline (Fig 3-1) (Baker and Wood, 2010).

The first purpose built shopping centres in Australia were Chermside shopping centre (1957) in Brisbane and Chadstone in Melbourne (1960). By 2007, the number of shopping centres based on the Shopping Centre Council of Australia’s statistics was 1,338. This included various sizes from regional centres down to neighbourhood centres. These centres generated about 41% of retail sales while comprising only 28% of the retail space. 1.75 billion shoppers visited these shopping centres each year across Australia, which showed that the average Australian visited a shopping centre twice a week (Property Council of Australia, 2007). This number later increased to 1,450 comprehensively planned shopping centres in Australia (Shopping Centre Council of Australia, 2011) (Drechsler, 2014).

While shopping centres play a significant role in all metropolitan Australian cities, some disparities are traceable between cities such as Brisbane, Sydney and Melbourne. Melbourne has retained a greater number of traditional strip shopping centres, while Brisbane appears to have a greater number of

Chapter 3 The Case Study Settings 52 corporately owned shopping malls. Sydney and Melbourne are experiencing a more diverse retail environment, which is very much related to the denser urban areas and older history of settlement in the larger areas of these cities. This has formed a more diverse retail environment that still serves the population and is dynamically active.

Fig 3-1: Shopping centres’ locations in the Brisbane Statistical Division (BSD)

3.2.3 The expansion of shopping centres

From their beginning, shopping malls became the dominant element of the retail hierarchy in Australian cities at the expense of rail based strip shopping centres. Today, some of these centres offer a challenge to the conventional central business district (CBD) in terms of their retail offerings (Davison and Yelland, 2004).

The history of the retail environment in Brisbane shows that in 1965 the most dominated retail area in Brisbane was the CBD with almost 150,000 m2 of retail floor space. Chermside was the only regional centre with 20,000 m2 of floor space, as well as a number of smaller planned centres at Newmarket, Stones Corner, Inala and Jindalee. Several established strip retail centres were also

Chapter 3 The Case Study Settings 53 active in that time including, but not limited to, Moorooka, Stones Corner, Mt. Gravatt, Coorparoo, Nundah, Stafford, Wynnum, Sandgate and Redcliffe. From 1965 to 1975 other centres were established, mostly located within 12 km of the CBD. These included four regional centres at Toombul, Indooroopilly, Garden City, and Brookside. The greatest expansionary growth in areas more distant from the CBD occurred between 1975 and 1985 in centres such as Capalaba, Alexandra Hills, Kippa Ring, Kuraby, Centenary, Aspley Hypermarket Chermside, Toombul, Brookside, and Indooroopilly. From 1985 to 1990, this trend was followed by the expansion of other existing centres. At the same time, new centres were developed in the northern extreme of the Brisbane metropolitan area, in Caboolture and in the southern extreme, Logan City 1(1990).

This continuous trend in expansion of shopping centres is still happening due to the evolutionary nature of retail and the constant shifts and innovation in the field of marketing. While shopping centres form a considerable part of retail destinations in Brisbane, there are other very common types of retail space that need to be considered. For example, internet shopping is a significant shift which has made its way into the retail industry and it is affecting the existing retail structure, forcing developers to start finding other options to compete and attract people to their stores.

3.2.4 Shopping centres’ catchment areas

Planners in Australia have used survey-based techniques to define the 'catchment area' of specific retail centres (Bates, 2003). However, previous planning documents and the way these centres have formed around the city, show that Australian shopping centres have the same functions as central places in Christaller’s theory (Yamashita et al., 2006). Australian metropolitan activity centre networks typically follow a functional hierarchy along the lines supported by Christaller’s theory. These activity centres are being supported

(1) from “The Brisbane plan : a city strategy”. Strategy paper, published by Brisbane City Council 1990 / Accessed date: http://trove.nla.gov.au/work/34912556?selectedversion=NBD42285714

Chapter 3 The Case Study Settings 54 and implemented by the metropolitan spatial plans in all states within Australia, apart from Tasmania and the Northern Territory (Drechsler, 2014).

Fig 3-2: The activity centre network hierarchy (Drechsler, 2014) P.274.

The central place theory has been explicitly mentioned by the National Capital Development Commission (NCDC) (1981) in reference to the planning of retail conditions in Canberra, although what was actually developed was somewhat different from what the theory might predict (Walmsley and Weinand, 1990). As part of the statutory planning tool, the Western Australian government’s Activity Centres Policy (2010) has clearly executed the activity centre hierarchy proposed by Christaller (Drechsler, 2014).

Higher order shopping centres in Australia are providing lower order centres’ products and services while they are delivering some additional levels as well. The central theory is based on the hierarchical grouping of functions and the number of functions being proposed in each centre. This theory was applied by the planners up until the 1970s to establish the relationship between shop numbers and population (Fig. 3-2) (Walmsley and Weinand, 1990). Christaller’s central place theory (CPT) model is easily recognized in younger cities such as Canberra and Brisbane in comparison to Sydney and Melbourne, which have a more complex system of commercial locations. Brisbane accepted Christaller’s model in its first town plan in 1978 and found it to be a reliable method of assessing the suburban proposals, because at that time, before the arrival of the large supermarkets, there was not much variation between shop sizes.

Chapter 3 The Case Study Settings 55

Other states such as Victoria found it to be an imperfect, inadequate approach and never accepted it (Corporation et al., 1992).

3.3 Retail accessibility in Brisbane

In Brisbane during the 1950s, people mostly relied on trams and walking as ways of accessing retail destinations. The population was settled close to tram routes where stores developed along the major transit corridors, which were served by frequent tram services. The significance of tram and bus services in Brisbane can be shown by its high passenger numbers (about 152 million) in 1946 . Around the middle of the 20th century, the construction of new roads and highways was rapidly increasing, and at the same time, private cars were becoming the most convenient and affordable option for many households, so the new types of retail outlets had the chance to expand. The Riverside expressway and the Captain Cook Bridge were completed in 1973. In 1982, the South East freeway was completed to Logan Road, Nathan. In 1986 the world’s longest cantilevered box-girder bridge – the Gateway Bridge – was officially opened, linking the Sunshine and Gold Coasts. Much of the development along these routes had no other accessibility options, apart from these highways and roads. Consequently, the number of car trips was rapidly increasing.

Ferreira (1999) studied the different trips relating to trip purpose taking place in Brisbane in 1992. Shopping trips were the second largest group among the other home-based trips in the non-CBD area, but only 0.3 % of all these trips were made by bus. His study showed that the percentage of bus trips experienced a decline compared with those in 1986 with only 2.4% of bus trips. As table 3-1 and 3-2 show, the percentage of bus trips for home based shopping purposes had decreased from 0.7% and 2.5% in 1986 to 0.3% and 2% in 1992 for non-CBD and CBD areas, respectively.

Chapter 3 The Case Study Settings 56

Table 3-1: Trips by trip purpose: CBD and non-CBD trips in 1992, source ARUP (1996) as found in (Ferreira, 1999)

Trip Purpose CBD (%) Non-CBD (%) HBWork 28 14 HBShoping 10 18 HBEducation 14 4 HBSocial 5 11 HBOther 13 3 NHB 30 50 All Trips 100 100

Table 3-2: Bus trips as percentage of all trips in 1992 source ARUP (1996) as found in (Ferreira, 1999) Trip Purpose CBD Non-CBD 1992 1986 1992 1986 HBWork 3.8 5.4 0.2 0.2 HBShopping 2 2.5 0.3 0.7 HBEducation 0.8 1.2 1.3 2.1 HBSocial 0.5 1 0.1 0.2 HBOther 0.1 0.1 0.1 0 NHB 1.7 3.6 0.5 0.9 All Trips 9 13.8 2.4 4.1

*HB refers to home based and NHB refers to non-home based

From the 1990s on, it is possible to see more focus on public transport in order to make it easier for people to travel inside the city. From this time, public transport usage was targeted in various strategic and transport plans, both for Brisbane and for larger areas like SEQ. As an example, in 1996 the first Brisbane River City Cat service was opened. In 2001, the first section of the busway network for South East Queensland, the South East Busway, was opened (Ferreira, 1999). Today, Brisbane commuters significantly rely on an expansive bus network in addition to the train and ferry routes. Prior to 2004, growth in passenger numbers on Brisbane Transport buses averaged 1.9 percent from 2001 to 2004, but Brisbane Transport has recorded strong patronage growth since the introduction of integrated ticketing on 1st July 2004,

Chapter 3 The Case Study Settings 57

averaging 8.9 percent (2009). This allows people to switch from buses to trains to ferries and City Cats throughout their travelling day, and therefore improves convenience for commuters.

Table 3-3: Passenger boarding across seven capital cities 2005-06 to 2007-08 (2009)

2005–06 2006–07 2007–08 2006–2007 2007–2008 Total boardings (million) Percentage change Total reported Total reported Total reported Reported Reported boardings boardings boardings boardings boardings Sydney 226.8 233.4 241.3 – 3.4 Melbourne 79.1 85 91.3 7.5 7.4 Brisbane 129.7 140.5 149.4 8.3 6.3 Adelaide 50.2 51 52 1.6 2 Perth 63.9 64.6 65.7 1.1 1.7 Hobart 7.6 7.7 7.4 1.3 -3.9 Canberra 16.8 16.8 16.9 0 0.6 Total 574.1 599 624.1 – 4.2

The Department of Transport and Main Roads in their “Travel in South-East Queensland – May 2012” report (Fig 3-3), claimed that although the private vehicle still dominates as the primary mode of travel, there has been a continued increase in the mode share of public transport since 1992, and a modest increase in active transport mode share in 2009. This, show an opposite trajectory compared to the unrelenting decline from 1992. Their report shows 307,000 less private vehicle trips per day, 193,300 more public transport trips per day and 94,000 more active transport trips per day between the periods of 2007 to 2009. Looking more closely, for the trips with the purpose of shopping/personal activities, within the same periods there is a constant decrease in the percentage of public and active transport mode share.

The report also shows that shopping/personal trips are experiencing the lowest average of time and distance in comparison to the other seven defined categories such as work, education, etc. People appear to prefer closer locations and lower trip length when it comes to shopping/personal trips. Shopping/personal business trips remain at fairly consistent levels from 9am throughout the day to the peak hour at 3pm.

Chapter 3 The Case Study Settings 58

Although all of the abovementioned studies show a considerable increase in the percentage of use of public and active transport, it seems that shopping trips are not experiencing an extensive difference in their trip mode share.

(a) Percentage of trips which happened by each mode type for different years

(b) Public and active transport modeshare by trip purpose

Chapter 3 The Case Study Settings 59

(e) Time of travel

Fig 3-3: Trip attribute analysis of various trip types in SEQ Source (2012) P.19, 20 ,22

Chapter 3 The Case Study Settings 60

3.4 Regulatory concerns

Simmon et al. (1998) discussed the theory that consumer service activities need to be successful within individual markets with distinct types of competitors, and therefore they are very much local entities. He compared commercial structure and public policy implications in a number of metropolitan cities including, Barcelona, Nagoya, Toronto, Dallas-Fort Worth, Mexico City, Hong Kong, Melbourne, and Munich, and concluded that the types of markets, local commercial structures, competitive situations, and regulatory environments are major elements. These elements need to be understood correctly in order to prepare for a discussion about commercial public policy (Simmons et al., 1998).

Simmons et al. (1998)argues that while the impacts of governmental regulations such as those relating to store location and size limits, store hours, restrictions on pricing and “who sells what” and market size on the commercial structure of the cities are considerable, these factors are not as powerful as some other indirect regulatory actions. These indirect regulatory actions might include income redistribution, transportation policies on Transit Oriented Development (TOD), or building more highways or rapid transit, land use regulations on planning and zoning matters showing the future densified residential areas or future development directions, which can affect the catchment area of the retail outlets, the degree to which local government can influence the retail location and operation, labor regulations and collective bargaining, or even competition policy. He concluded that commercial structure can be directly related to urban form, and that local and/or national regulatory environments shape the urban form. Therefore noticeable disparities in metropolitan commercial structures throughout the world are due to different general public policies and specific regional regulations which have been discussed above (Simmons et al., 1998).

In this section, this regulatory environment is assessed to see to what extent these policies are affecting the commercial structure of Brisbane.

3.4.1 Commercial public policy and regulations in Brisbane

In Brisbane, while some of these policies and regulations, including land regulations, trading hours, preparation and submission of Environmental Impact

Chapter 3 The Case Study Settings 61

Assessment (EIA), seem to be very effective in the formation of the current retail structure, other factors such as sales tax revenue (GST) does not seem have a significant impact. This is due to the fact that Australian cities do not have the power to levy sales tax and the revenues from the sales tax rate are an important source of income for the government of Australia (Lee and McCracken, 2012).

Extensive expansion of shopping centres in the 1970s and 1980s caused the formation of many of these regulations. The provision and submission of an Environmental Impact Assessment (EIA) was first brought into force by the Queensland State Government in mid-1980, at the same time as strategic plans, development control plans and new advertising requirements for development applications were introduced. EIA was introduced in response to the business concerns from small centres regarding the increasing number of larger centres and the resulting competition. It was intended to ensure that approvals of proposed new shopping centres took into account public need and demand for the centre and its potential impacts upon existing development. The economic impact assessments for shopping centre developments were enforced for the developments of more than 4,000 m2 of gross floor area (GFA) or sites exceeding 1.5 hectares, which later changed to 6,000 m2 of GFA or a site of 2.5 ha (Queensland Local Government Act, section 33[1]). The EIA was the first of its type in Australia and is defined by the Local Government Act as:

A study report including an assessment of the public need and demand for a major shopping development and a statement of the likely economic impact upon existing development of a similar nature or involving similar activities in the locality and in the estimated area of influence of the proposed development if such proposal were implemented (Corporation et al., 1992; Lee and McCracken, 2012).

For these types of developments, after preparing the documents, the re-zoning proposal and its relevant EIA should be available for assessment both by the public and the local authority. This gives everyone the opportunity to present their opinions. The EIA will be checked against the statutory urban planning

Chapter 3 The Case Study Settings 62 regulations. These reports usually include information about the existing retail network, population growth and economic overview of the area, trade area analysis of the proposed shopping centre, demand analysis to indicate the importance of the new establishment, and finally the impacts and influences of this new shopping precinct on the existing ones.

Land price is another factor that significantly influences the location and characteristics of the retail environment. Despite the abundant amount of land in Australia, there are strict land regulations because only 0.3 percent of the land is urbanized. These restraints have resulted in increases in land prices for different activities, including costs of building shopping centres, followed by a reduction in competitive pressures and an increase in retail prices (Moran, 2007). Due to the high price of land in the retail market, many new businesses cannot afford to enter the competition. Many small businesses will never be able to afford to grow and will eventually leave the competition. Westfield, Coles and Woolworths and their powerful market reach is a good example of this impact, eventhough the rising of some other competitors such as Aldi has started to challeng them over the last few years. Even if a small business can afford the business cost, the land value is so high that it will be prevented from the competition in the first place. This is to the extent that Baker (2004) claims Australia is under-provided with shops compared with countries such as U.S. (Moran, 2007).

Finally, the Queensland Government legislation concerning trading hours has had more influence on the retail environment by providing longer trading hours for independent retailers outside shopping centres. It allows them to source customers and make a profit, since they do not have the power to compete with the large shopping centres, which have supermarkets, large stores, chain- operated specialty shops and a variety of stores.

While the abovementioned policies directly or indirectly influence the retail structure and its functions in Brisbane, it is also important to look into the existing statutory regulations which directly shape the city’s urban form and its various elements, including the retail environment and the anticipated direction of future development.

Chapter 3 The Case Study Settings 63

3.4.2 Statutory documents and retail schemes

A number of strategic plans have been developed in response to the ongoing, rapid development of shopping centres in most Australian cities. These plans were made in an attempt to regulate the size, function and location of the centres (Lee and McCracken, 2012). In Queensland, the development of shopping centres is being regulated by the local authorities through the town planning schemes, which direct the location and the nature of development. There are disparities within the ways that these local governments behave and their control documents for these activities (Corporation et al., 1992).

It is believed, mostly by planners, that commercial forces alone cannot ensure the broader community objectives and there is a need for local authorities to develop other restrictive regulatory frameworks that refer to further community interests and objectives, including locational criteria.

In most cases, shopping centre locations are controlled by “zoning regulations”. This can be determined either through the “pre-zoning” or “re-zoning” process. In practice, pre-zoning is not usually an option (apart from for the CBD related lands) regarding problems attached to this policy, such as high land prices. Making decisions about shopping centres’ locations in the earlier development stages may result in increasing the value of the site and therefore more allocation of infrastructure requirements to these sites by local authorities. At the same time, the developers would no longer have any interest in applying for rezoning in other sites, since this may result in an oversupply of needs and would not be profitable (Corporation et al., 1992).

Queensland’s planning process is not based on a pre-zoning concept and this makes it different from other states in Australia. Developers are responsible for identifying suitable sites and putting their applications for rezoning the land through the ‘applicant-initiates rezoning’ process. This is to consider the right of every person to apply for a specific type of land use.

Developers should propose the sites where land is available and which satisfy commercial criteria highly related to the location. Such criteria should be good accessibility to a strong trading area, high visibility, presence of market

Chapter 3 The Case Study Settings 64 opportunities and transport considerations such as traffic and available public transport options. In the case of applications for rezoning land, this should also satisfy the criteria set out by the local authority in its planning documents (Corporation et al.). These planning documents could be a simple zoning table showing the compatibility/non-compatibility of different land uses, or more explicit provisions in the form of other planning documents including strategic or structure plan, DCPs (development control plans), etc. These plans could present the aims, goals, objectives and implementation criteria about how the development (including retail development) should happen in the city, their preferred hierarchy, location’s criteria, and retail floor-space ratio. Strategic plans contain the maps that show the existing fabrics and preferred areas for future development. They sometimes indicate the notional sites for shopping facilities. These planning documents are statutory and should always be followed.

Most of metropolitan Brisbane’s local government areas (LGAs) have articulated, to varying degrees, a preferred hierarchy or network of activity centres as part of their planning schemes, dating back to the 1970s and 1980s. Most of these hierarchies or networks simply reflect historical patterns of centre development (Lee and McCracken, 2012).

3.4.3 Planning documents and their intended policies

3.4.3.1 City Plan 1978

The city plan, which was approved for Brisbane city in 1978, was the first example of these documents, and included objectives and assessment criteria that represent the retail hierarchy and retail floor-space ratio.

3.4.3.2 City Plan 2000

Brisbane City Plan 2000 was the next planning document that directly controlled and set policies on the city land use, including commercial activities. This city plan, which is a comprehensive statement declaring the council’s intentions for the future developments in Brisbane, includes the citywide desired environmental outcomes and supporting strategies which provide the builders, developers, solicitors and others with useful guidelines, ensuring the

Chapter 3 The Case Study Settings 65 achievement of the vision of Brisbane. Every new development should be assessed against this plan. The general spatial land use of the city and the interrelationship between the land uses have been revealed in this plan. It considers a number of elements representing the city’s directions for strategic issues including residential neighbourhoods, green space systems, industrial locations, activity centres, movement system, native title and heritage. Strategic directions for these various elements are generally used as a basis for assessing the impacts of any proposed plan.

The plan focuses on maintaining the city centre’s dominance for higher order activities in SEQ and supports it with a well-defined network of multi-purpose and special purpose centres. It is seeking integrated retailing, service, cultural, and community facilities allowing for different community uses, services and facilities within the same site or facility in centres linking with pedestrian ways, bikeways and public transport. The city plan highlights the in-centre development and discourages any ribbon development. It proposes higher density housing and mixed land use development in centres and other locations that are served by public transport. The location of shops and offices out of the suburban and convenience centres is only permitted where they are sufficiently removed from centres and other small shops and offices, are located on a district access route or suburban route, and serve local community needs.

The plan proposes a “system of centres” including multi-purpose and special- purpose centres. These multi-purpose centres, which incorporate most of the traditional strip shopping centres, should be properly serviced by pedestrian, bike and public transport (to improve people’s access, mobility and mode choice) and include the categories of city centre (incorporating the Brisbane Central Business District), major centres, suburban centres and convenience centres. These multipurpose centres must be surrounded by higher density residential housing.

The plan strongly encourages in-centre and discourages out-of-centre development, excluding the communities with “overwhelming needs” which must be verified by the provision of a Commercial Impact Assessment (CIA). It also includes retail warehouses, which might seem to be more attractive in a

Chapter 3 The Case Study Settings 66 standalone location. In the case of the approval of the assessment, the preferences will be given to “a development at the edge of an existing centre rather than a standalone or more isolated location if it is possible”. For the communities with an approved overwhelming need, small-scale shop or office activities should be provided with 250m2 or smaller ground floor area (GFA) which serves the local community and is only accessible through district or suburban routes.

The plan is basically strengthening the inside looking design, in-centre development and commercial character buildings for the centre by emphasizing limitations for all the future shops and offices to front the city’s major road system to increase the carrying capacity, safety and visual amenity of the major roads.

3.4.3.3 City Plan 2014

While there are many similarities between the City Plan 2000 and 2014 in terms of the number and location of the major regional centres, some major differences are also traceable in the way these plans look at the distribution of smaller scale centres around the city. Table 3-4 shows the different centre and mixed-use classifications both for the 2000 and the 2014 city plans.

As can be seen, the major centre classifications are very similar but when it comes to the suburban centres (district centres) and mixed-use retail centres, a new category of corridor type of centres (rather than the previous dot type of centres) has emerged. It shows the new understanding of the role of the retail centres followed by new distribution and form of these centres in the city. The new city plan has started to consider proposals outside of the limited idea of the in-centre policy and into a more evenly distributed corridor type of retail centre which can expand alongside transport routes. This type of development is taking place more in combination with other types of activity, for example having houses on the top levels and retail outlets on the ground level.

Chapter 3 The Case Study Settings 67

Table 3-4: Centre and Mixed Use Zone Categories – (City Plan 2014)

*Mixed Use definition: Building that integrates residential activities with commercial, retail or low-impact industry activities.

Furthermore, the new city plan omits the process of EIA submission for getting approval for new or major extensions in shopping centres. This plan now only requires a number of basic specifications such as good level of accessibility, a minimum number of parking spaces, etc., as being preliminary requirements

Chapter 3 The Case Study Settings 68 from the developers in order to reduce the limitations and increase the level of the competition between the businesses.

3.4.3.4 SEQ Regional Plan 2009-2031

Under the Sustainable Planning Act 2009 (SPA) planning schemes must be consistent with the intent of desired regional outcomes, principles and policies set out in any applicable regional plan. The local government is required to assess a development application against the regional plan if the regional plan is not identified as being appropriately reflected in the particular planning scheme. Where there is conflict, the regional plan will prevail to the extent of any inconsistency (Lamb, 2010 ).

The statutory Southeast Queensland Regional Plan 2005–26 was introduced in the 2000s by the Queensland State Government in order to establish an urban footprint to guide future sustainable urban development and more efficient use of infrastructure in the region. The plan designates Brisbane’s CBD as the region’s primary activity centre, followed by a number of principal and major regional centres (almost always anchored by a major shopping centre) located throughout the metropolitan area. However, all the activity centres apart from the future major centre at Ripley pre-existed the introduction of the regional plan, revealing the pattern of development largely prearranged by the region’s respective LGAs’ individual centre planning policies and major land and centre developers (Lee and McCracken, 2012).

3.4.3.5 Summary

An inspection of the existing planning documents and the way in which the commercial structure of Brisbane is being directed, clearly shows that the substantial role of shopping centres is becoming more noticeable. In many cases, these retail centres are considered to be major activity centres that should be supported by highly accessible and frequent public transport and a high level of population density.

A system of centres in various hierarchical levels is planned which will be considered on the basis of any future proposed retail establishments in their

Chapter 3 The Case Study Settings 69 vicinity or on their existing edge. Other types of retail establishments are mostly restricted within the proximity of these centres.

No restrictions have been applied on the expansion and the level of function of large centres, and the issue of competition is clearly ignored within these planning frameworks.

A study of the direct and indirect policies influencing the commercial structure of Brisbane, from land price and trading hours to the pre-planned and in-centre development of retail or the required EIA for any newly proposed location, shows that these policies are all highly supportive of the growth and empowerment of the shopping centers.

Apart from the new city plan which is currently experiencing some modifications, none of these planning documents seems to be mindful of the developers’ concerns or to consider the feasibility aspects of their plans, such as the supportive population density, the issues of blocking the required lands for high density development, etc.

Chapter 4 Research Methodology

Chapter 4 Research Methodology 73

4.1 Introduction

The literature review showed that a considerable number of techniques from the fields of economic geography and urban/transport planning have been used to investigate customers’ choices of destinations and transport modes, various aspects of their travel behaviour and to identify attributes which influence retail developers’ decisions. Some of these techniques have been applied quite widely by developers to boost the profitability of their businesses by finding the right locations for the right type of businesses in terms of their characteristics and the types of services and products provided.

However, a gap was identified in the field of retail transport sustainability, with regard to comprehensive approaches that reflect different dimensions of the retail environment and the factors that shape it, before proposing future modifications of policy directions. Furthermore, these methods have hardly been applied in an Australian retail context. Also, even though shopping centres have been mentioned quite extensively in various planning documents, there is an obvious gap with regard to the real impacts which these types of land uses have on the way people travel in the city.

When it comes to retail, customers, retail developers and urban/transport planners all have significant opportunities to influence its form and structure. This is a key reason why the study of retail trips is more complicated than the study of other trip types, such as commuting. Identifying the spatial distribution and set of characteristics for retail environments that simultaneously considers the desires and interests of these different participants appears to be a considerable challenge. This challenge therefore requires a more comprehensive and complicated methodological framework.

Due to the number and complexity of questions addressed in this study, both quantitative and qualitative approaches are considered within the overall methodological framework. The methods not only aim to answer the three major research questions, but also to bridge the gap between existing approaches, find relationships between them and identify a methodological framework to

Chapter 4 Research Methodology 74 facilitate a comprehensive understanding of retail structure and its future form within the city.

The chapter begins with a detailed description of the datasets available for this research. The research methodology will then be explained and the research framework will be described. A discussion of the applied methods follows, considering the reasons for their selection and possible limitations of their application. Each method will be discussed briefly to construct a picture of the overall steps in the research. The methodology chapter will focus on the overall methodological concept, without going into the detailed practical implications of each method. Practical implementation issues will be discussed in the relevant chapters later in the thesis.

4.2 Datasets available for research

While potentially a wide range of methods can be applied for these types of studies, the nature and availability of datasets may significantly limit the scope of research. Therefore, as the first step in defining the research methodology, existing datasets on retail environments and customers’ travel behaviour in Brisbane were investigated.

Two major categories of data have been identified and collected for subsequent analysis. The first category of datasets addresses the current locational and spatial characteristics of retail in the city and comprises:

1- Land use dataset (2011)

Brisbane land use distribution dataset (2011) provided by the Queensland Government, indicating the distribution and location of different land uses in the city.

2- ABS dataset (2011)

The Australian Bureau of Statistics (ABS) Census datasets (2011), containing information on the population distribution and characteristics within ‘mesh blocks’ (MBs), as the smallest geographical areas defined by the ABS.

Chapter 4 Research Methodology 75

3- Shopping Centre Directory (SCD) (2011)

The Shopping Centre Directory (SCD) dataset provided by the Shopping Centre Council of Australia (SCCA), containing data on the exact location and some of the characteristics of existing shopping centres in Queensland. Based on the ‘Directory of Shopping Centres/Queensland 2011 report’ provided by the SCCA, this dataset follows the standard classification of shopping centres employed in all the Property Council’s shopping centre directories in Australia. Shopping centres are categorized in descending order of size as city centre, super-regional, major-regional, regional, sub-regional, and neighbourhood shopping centres, with separate categories for themed, market, outlet centres and bulky goods shopping destinations1. While these data are useful, they are

(1) 1- City Centre: Based on the SCCA’s definition, city centre is the retail premises within an arcade or mall development owned by one company, firm or person and promoted as an entity within a major Central Business District. Total gross lettable area for retail space exceeds 1’000 square metres. The key features of these centres are: dominated by specialty shops; likely to have frontage on a mall or major CBD road; often coexists with large department stores. 2- Super Regional Centre: A major shopping centre typically incorporating two full line department stores, one or more full line discount department stores, two supermarkets and approximately 250 specialty shops. Total gross lettable area for retail space exceeds 85,000 square metres. The key features of these centres are: one-stop shopping for all needs; comprehensive coverage of the full range of retail needs (including specialised retail), containing a combination of full line department stores, full line discount department stores, supermarkets, services, chain and other specialty retailers; typically includes a number of entertainment and leisure attractions such as cinemas, arcade games and soft play centres; provides a broad range of shopper facilities (car parking, food court) and amenities (rest rooms, seating). 3- Major Regional Centre: A major regional centre typically incorporating at least one full line department store, one or more full line discount department stores, one or more supermarkets and approximately 150 specialty shops. Total gross lettable area for retail space ranges between 50,000 and 85,000 square metres. The key features of these centres are: one-stop shopping for all needs; extensive coverage of the full range of needs (including specialised retail), containing a combination of full line department stores, full line discount department stores, supermarkets, chain and other specialty retailers; typically includes a number of entertainment and leisure attractions such as cinemas, arcade games and soft play centres; provides a broad range of shoppers’ facilities (car parking, food court) and amenities (rest room, seating). 4- Regional Centre: A shopping centre typically incorporates one full line department store, a full line discount department store, one or more supermarkets and approximately 100 specialty shops. Total gross lettable area for retail space ranges between 30,000 and 50,000 square metres. In some instances, all other characteristics being equal, a centre with two full line discount department stores, without a department store, serves as a regional centre. The key features of these centres are: extensive coverage of a broad range of retail needs (including specialised retail), however, not as extensive as major regional centres; contains a combination of full line department stores, full line discount department stores, supermarkets, banks chain and other specialty retailers; provides a broad range of shopper facilities and amenities. 5- Sub Regional Centre: A medium sized shopping centre typically incorporating at least one full line discount department store, a major supermarket and approximately 40 specialty shops. Total gross lettable area for retail space ranges between 10,000 and 30,000 square metres. The key features of these centres are: provides a broad range of sub-regional retail needs; typically dominated by a full line discount department store or major supermarket. 6- Neighbourhood Centre: A local shopping centre comprising a supermarket and approximately 35 specialty shops. Total gross lettable area for retail is less than 10,000 square metres. The key features of these centres are: typically located in residential areas; services immediate residential neighbourhood; usually has extended trading hours; caters for basic day-to-day retail needs.

Chapter 4 Research Methodology 76 not comprehensive and reliable for some categories of information because of the missing data from some fields. Besides, while the dataset comprises about 200 shopping centres within the Brisbane statistical division (BSD), some neighbourhood or small centres are not included.

4- Socio-demographic data from the BSTM model

Socio-demographic data which can be extracted from the Brisbane Strategic Transport Model (BSTM), include the current and anticipated number of retail jobs, retail job density, employment density, etc., for almost 1,500 transport zones within the BSD.

Detailed information is also available from the city plans and priority infrastructure plans in the Brisbane Local Government Area such as the available detailing housing type, development corridors, etc. Locational data for major supermarkets such as Coles, Woolworths, IGA and Aldi as ‘points of interest’ are also accessible from their websites.

While some of these retail-focused datasets seems to be very informative, they are only available for the limited area within Brisbane’s council boundary. Issues related to data mismatch are also quite common and will be addressed in the relevant chapters.

The second category of datasets contains data on the current transport network in the city. These include information on the current transport system and available transport modes as well as a dataset on the travel behaviour of the resident population.

1- Household travel survey data

The Household Travel Survey data applied in this study was provided by the Queensland Department of Transport and Main Roads. The 2009 South East Queensland Travel Survey (SEQTS) data detailed the 7-day travel of residents in South East Queensland, Australia, between 20 April and 28 June 2009. The

7- Bulky Goods Centre: A medium to large sized shopping centre dominated by bulky goods retailers (furniture, white goods and other homewares), occupying large areas to display merchandise. Typically contains a small number of specialty shops. Generally located adjacent to large regional centres or in non-traditional retail locations (i.e. greenfield sites and industrial areas); purpose designed, built and operated, generally with a layout of outlets around a central, landscaped area and an overall design and colour theme to promote the appearance of an integrated development; generally greater than 5,000 square metres (GLAR) in size.

Chapter 4 Research Methodology 77

SEQTS used multi-stage, variable-proportion, clustered sampling of households. The survey achieved a 52% response rate and obtained information on the travel behaviour of 27,213 respondents living in 10,335 households. To account for non-reporting, weightings for both non-response and selection bias, derived from household characteristics and Australian Bureau of Statistics census data, were included within the SEQTS dataset. These weightings were applied to the sample results to estimate the travel behaviour of a total of 641,061 households in the SEQ region (The Urban Transport Institute, 2010).

Week-long diaries were completed by respondents aged five and over, and the travel of persons aged zero to four years was reconstructed from diaries provided by other household members. The survey covered residential households within the Brisbane Statistical Division (BSD), the Gold Coast City Council and Sunshine Coast Regional Council areas. However, the scope of this study only includes the BSD, with a sample size of 4,240 households (The Urban Transport Institute, 2010).

The SEQTS recorded all trips made by respondents during their survey week. The data includes individual characteristics, household characteristics and trip attributes self-reported by the participants. Each trip is divided into ‘trip stages’, defined as all parts of a trip that may be made by different transport modes (for example, a public transport trip from home to a shop may involve three stages: a walking stage to a bus stop, a bus stage and then a final walking stage from the bus to the shop). In total 86,549 trip stages were recorded for the BSD, out of a reported total of 79,790 trips (The Urban Transport Institute, 2010).

The exact routes travelled by respondents were not captured within the SEQTS. Trip distances were calculated using geographic information systems (GIS) to determine the shortest path on the street and path network. Within the SEQTS, motor vehicles were defined as either a car, 4WD, van or truck (The Urban Transport Institute, 2010).

The dataset includes all the information about the different trips that the people who were tracked started and ended on specific days, such as their origins and destinations, means of transport, travel time, etc. Shopping trips as major parts

Chapter 4 Research Methodology 78 of these trips are also detailed in the dataset. These data give us information about where people usually go to shop, how often they go shopping, what types of goods they usually buy and what mode of transport they choose for accessing these destinations. This dataset also provides information on the personal characteristics of trip makers such as age, gender, income, place of residence, etc.

2- Network data

Railway, busway and road network data were obtained from the public transport operator, Translink, and the Queensland government website (2014).

Fig 4-1: Datasets used in the research

Chapter 4 Research Methodology 79

4.3 Methodology

As previously discussed, this research attempts to answer three sets of questions as a prerequisite for responding to the overarching question of how to encourage more sustainable retail travel behaviour in Brisbane in the future. Specific methods were identified and developed to answer these subsidiary research questions (Fig 4-4), based on the availability of datasets and their ability to address the research questions. The overall methodology will be described briefly in this chapter. The methods to be applied will be outlined briefly and the justifications for their selection will be explained, while their practical implementation will be discussed in relevant chapters later in the thesis.

Three sets of subsidiary research questions will be addressed to support the overarching research question of how to facilitate/encourage more sustainable retail travel.

The first set of questions looks at the existing retail environment and investigates the behavioural implications of this environment: o What is the current form and structure of retail spaces in Brisbane? o What is the current travel behaviour of Brisbane’s retail shoppers? o Which transport modes are preferred for retail trips in Brisbane?

The second set of questions aims to model shopping destination choice and identify the factors that influence destination choice: o How do Brisbane’s shoppers choose between available retail destinations? o Which factors affect shoppers’ destination choice? o Which types of destinations are generally preferred and why?

The third set of questions focuses on the potential future of retail environments. They ask whether these spaces will follow their current trajectory, and whether policies which could help to facilitate a shift towards a more sustainable retail transport future can be identified: o What appears to be the likely future of retail spaces in Brisbane – given existing trends and probable future policies?

Chapter 4 Research Methodology 80

o How could planning policies potentially help to facilitate a shift towards a more sustainable future for retail environments in Brisbane?

The first set of questions that relates to current planners’ understanding of the issue, raises important questions regarding the current retail environment and travel behaviour responses of customers, preferred mode types, selected travel distances, etc. The second set of questions aims to develop an understanding of customers’ desires for their choice of trip destination. The third set of questions aims to identify potential turning points in the form or distribution of the retail environment (mostly dictated by developers and planners) which might affect customers’ travel behaviour. In this way, the research aims to address various features of retail related to the planners, developers and customers that might directly or indirectly affect customers’ travel behaviour.

Fig 4-2: Three major participants with direct or indirect influence over the overall structure and form of retail

Addressing the first set of subsidiary questions

The Household Travel Survey (HTS) dataset will be used to answer the first subsidiary question on the existing retail setting and its implications for current transport behaviours (Chapter 5). This analysis will aim to reveal various aspects of customers’ trip decisions; their transport mode choice preferences, mode share percentages, chosen distances of travel, chosen destination types, types of items purchased, etc., whilst also identifying individual characteristics

Chapter 4 Research Methodology 81 of trip makers. The HTS recorded almost 3,350 shopping trips in the Brisbane statistical division (BSD). These trips will be extracted and investigated to illuminate existing customers’ trip patterns in Brisbane.

The same dataset will also be used in Chapter 6 to identify major socio- demographic groups of retail customers. Chapter 6 will use cluster analysis to recognize and classify trip makers into a number of major categories, rather than assigning particular travel behaviours to particular socio-demographic groups. Once groups have been identified by cluster analysis, transport mode share, travel distances, and types of product purchased, etc., will be examined for each group. The results are expected to identify significant differences in travel behaviour between the different groups of retail trip makers. Identifying such differences in travel behaviour can help inform planning policies and facilitate more sustainable travel patterns in the future.

While the important role and influence of shopping centres has been discussed previously in Chapters 2 and 3, the analysis of customers’ travel behaviour which will be undertaken in Chapter 5 recognises that shopping centres are the most important element of Brisbane’s retail environment, attracting more than sixty percent of retail trips every week. From Chapter 7 onwards, these findings, in addition to limitations in the availability of information on retail destinations, will result in redefinition and restriction of the research scope only to the trips that have ended in a shopping centre.

The decision to restrict analysis of retail trips to shopping centres as destinations enables the characteristics of these shopping centres and their localities to be considered in more detail. The features which will be examined in detail in Chapter 7 include: population density in catchment areas around centres, retail-job density within a specific distance of centres, accessibility level for each type of centre from trip origins around the city, etc. Analyses of shopping centre characteristics is limited by the available datasets. While the SCD provides consistent information about most existing centres within the study boundary, other information regarding trip attributes (from the HTS) or spatial attributes in the immediate vicinity of shopping centres (such as population density, land use, retail density or accessibility) were obtained from

Chapter 4 Research Methodology 82 other sources and are based on different zonal divisions. A number of assumptions and a careful process of amalgamation are therefore required to allow these datasets to be used concurrently. Appropriate methods for dataset matching were devised using ArcGIS software (Fig 4-3).

Addressing the second set of subsidiary questions

The second set of subsidiary research questions regarding factors which influence customers’ destination choices will be addressed in Chapter 8. The HTS data and the analyses of Chapters 5, 6 and 7 should provide relevant information on trip makers’ transport mode choice and travel preferences and the attributes of the destinations, which can feed into future planning policies. However, these initial analyses are unlikely to identify how trip makers select particular destinations among the broad set of available options.

Chapter 8 will aim to identify significant driving factors of retail shoppers’ destination choice by undertaking a discrete choice modelling analysis of the HTS retail trip destinations. While it is hoped that the discrete choice modelling results should reveal some interesting aspects of customers’ destination choice preferences, important limitations may also restrict the usefulness of these results. These restrictions arise from the number of assumptions required to assemble the input data for the discrete choice model and, surprisingly, the absence of some potentially important data – even when the various datasets are combined. These limitations – and attempts to circumvent them – will be addressed in their relevant chapters later in the thesis.

The results from the destination choice model of Chapter 8 should provide an understanding of what currently is happening within the retail transportation environment in Brisbane, and will hopefully also identify influential factors which affect how shoppers’ respond to that retail environment, as evidenced by their choice of retail destinations. While this analysis may not reveal all relevant drivers – given the limitations inherent in the combined datasets – it should identify some key variables that will need to be considered when planning future changes in the sustainability of retail travel. Examples of such attributes which could be amenable to future changes in planning policy are trip distance

Chapter 4 Research Methodology 83

(affected by planning zoning regulations) and accessibility (affected by public transport policy).

Fig 4-3: Shopping centre locations and the zonal boundaries for the Census Collection District and the BSTM model

Chapter 4 Research Methodology 84

Addressing the third set of subsidiary questions

Once Chapters 5 to 8 have characterised Brisbane’s retail environment, categorised shoppers in terms of retail travel behaviour and transport mode preferences, and identified significant factors affecting destination choice, Chapters 9 and 10 will consider the future retail trajectories predicted by planners, developers and economists. These two chapters will incorporate anticipated shifts in the market and changes in planning policies and regulations. Chapters 9 and 10 will address issues such as: the predicted future performance of retail environments (in terms of size, distribution and spatial attributes) and pathways towards this future, proposed planning approaches and predicted swings in retail planning and ensuing travel behaviour. These chapters will also consider impacts of possible future trends on transport mode shares and total kilometres travelled as proxies for the future sustainability of the transport environment. All of these issues address the third set of subsidiary research questions.

Chapters 9 and 10 will employ two different methods to address these questions. Chapter 9 will describe how semi-structured interviews were undertaken to collect experts’ ideas on future retail trends in Brisbane, drawing expertise from professionals in both private and public sectors. A number of developers, planners and economists will be interviewed for these purposes. Chapter 9 thus directly targets the interest of the developers and planners. The semi-structured interviews will be important for clarifying how leading participants in the field intend to plan for the future of the retail environment in the city.

Chapter 10 will then use a transport modelling approach to identify how potential future retail and transport scenarios might affect shoppers’ travel behaviour. This analysis will use Brisbane’s existing strategic transport model (BSTM) to predict retail transport mode share and mode shifts, under future retail and transport scenarios. The BSTM will predict possible impacts of future scenarios on shoppers’ travel behaviour and the potential sustainability of retail transport in terms of the share of motorized and non-motorized modes. The BSTM model already contains various proposed scenarios for future years

Chapter 4 Research Methodology 85 including 2016, 2021, 2026, and 2031. The 2031 scenario included is that which the Department of Transport and Main Road propose for a more sustainable transport and land use system in the city. The BSTM model includes a pre- existing, validated mode choice model for different types of trip, including shopping trips. While there are some known and documented limitations with the BSTM’s mode choice model, its use will still be helpful for developing an understanding of the outcomes and impacts of retail travel under future development scenarios.

The results from Chapters 5 to 10 will be discussed and drawn together in Chapter 11 to form a set of guidelines, policies and recommendations to assist planners in devising more sustainable policies for the future of retail in Brisbane. Interrelationships between the separate chapters of the thesis are illustrated in Fig 4-4.

Chapter 4 Research Methodology 86

Fig 4-4: Proposed research framework to address the major questions in this study

Chapter 4 Research Methodology 87

4.3.1 Analytical methods

The literature review found that a large number of trip-specific, individual- specific, destination-specific and spatial-specific attributes are associated with shoppers’ trip patterns. As explained above, a number of different analytical methods including disaggregated analysis, cluster analysis, choice modelling, semi-structured interviews and simulation modelling will be applied in this study to better understand, explain, measure and model the impacts of different direct and indirect factors which influence the travel behaviour of customers (Fig 4-5).

This section will briefly clarify how each of these methods will be applied in other chapters. Full implementation details are provided in relevant chapters later in the thesis.

Fig 4-5: Methods applied in each chapter to identify and quantify different aspects of the retail transport environment

4.3.1.1 Disaggregated analysis of shoppers’ travel behaviour - (Chapter 5)

Different methods have been applied to study travel behaviour at different scales from region down to local neighbourhood. Based on datasets available, these studies can be divided into those which have used disaggregated vs. aggregated analysis.

Chapter 4 Research Methodology 88

As used in this research, the SEQHTS is a disaggregated dataset which provides information on 3,350 individual retail trips which were undertaken in BSD. The HTS dataset does not identify exact origins and destinations for each trip, but only identifies origins and destinations at ‘collection district’ resolution. However, for trips between origins and destinations in these collection districts, detailed information is provided on trip length, transport mode type, trip purpose and the individual characteristics of the trip makers. This enables a detailed disaggregated analysis of individual travel behaviour, which is preferable to relying on imprecise aggregated and averaged zonal analysis.

Chapter 5 analyses shoppers’ travel behaviour by using SPSS and Excel software to implement descriptive analyses in terms of transport mode choice, trip frequency, distance travelled, types of products purchased, type of destination chosen, age and gender of customers. These analyses will aim to reveal the shoppers’ mode preferences and habits in the existing transport and retail environments, as well as revealing potentially hidden factors which might be related to retail travel behaviour.

4.3.1.2 Inductive analysis of shoppers’ travel behaviour considering their socio-demographic characteristics using cluster analysis techniques- (Chapter 6)

An improved understanding of the socio-demographic characteristics of shoppers who make particular types of shopping trips will also be useful for addressing the main and subsidiary research questions. Earlier studies in the literature concluded that people with different socio-demographic and economic characteristics displayed different tendencies in their mode choice decisions. Shoppers from different age groups or income ranges might, for example, show higher usage of public or active transport on their shopping trips.

One simple method of scrutinizing the impact of shoppers’ socio-demographic characteristics would be to predefine shopper groups based on differences in socio-demographics and then attempt to identify differences in travel behaviour between these pre-defined groupings. However, it would be even more enlightening if the prevailed socio-demographic groups within the dataset are

Chapter 4 Research Methodology 89 identified inductively and the characteristics of the trip makers are then measured in more detail.

Cluster analysis is applied in Chapter 6 to identify distinct socio-demographic groups of retail trip makers from the trips recorded in the HTS dataset. Trip maker clusters are identified by considering differences in age, gender, household type, etc. The results will, hopefully, form a good platform from which the travel behaviour and mode choice of different individuals can be studied in subsequent chapters, based on their socio-demographic characteristics.

4.3.1.3 Spatial analysis of the retail destinations and their geographical setting by application of various ArcGIS techniques (Chapter 7)

Travel is a spatial activity. Thus, a range of spatial measures can be used to inform observed travel behaviours. These measures range from very general indicators such as population density, job density and housing ratio at particular trip origins and/or destinations to more detailed indicators of land-use mix and accessibility. For retail trips, other spatial factors such as the distribution of retail job numbers and the shopping centres’ catchments areas could also be important.

In Chapter 7, ArcGIS software will be used to explore spatial characteristics which could influence the retail environment. GIS analysis helps to identify and understand the impacts of space on travel behaviour. It also helps to quantify spatial characteristics of trip origins or destinations by exploring multivariate relationships between different characteristics – such as land use, accessibility and employment density (Srinivasan, 2002).

While some of these characteristics and attributes can be measured easily in ArcGIS, other characteristics such as accessibility require more complicated techniques. Accessibility is believed to be one of the major elements of spatial structure which exerts a significant effect on retail trip decisions (Handy, 1992). Different models have been developed to estimate/predict the level of accessibility in the city. This study applies the Land Use and Public Transport Accessibility Index (LUPTAI) developed for SEQ based on GIS measurement of travel time to measure the shopping destination accessibility by different mode

Chapter 4 Research Methodology 90 types from various locations in the BSD.

4.3.1.4 Modelling the destination preferences of customers using discrete choice models (Chapter 8)

Chapter 8 will develop discrete choice models to analyze trip maker’s choice of destination from a limited set of alternatives. Discrete choice models have been used in numerous applications, since many behavioural responses are discrete in nature, and a single outcome is selected from a set of alternatives. For every retail trip, a customer chooses a destination from the set of available alternatives based on a number of factors that make destinations more or less attractive.

The types of factors which will be considered in the discrete choice model in Chapter 8 include relevant trip-specific, spatial-specific, destination-specific attributes and individual-specific socio-demographic characteristics which will have been determined by the analyses in Chapters 5 – 7. The significance for these different attributes and the attractiveness of different types of destinations will be estimated using the discrete choice model developed in Chapter 8.

4.3.1.5 Semi-structured interviews conducted to understand current and future trends in form and geography of retail (Chapter 9)

Analyzing revealed preference data on destination choice, together with characteristics of retail form and structure should significantly improve understanding of customers’ travel behaviour. However, revealed preference data from shoppers’ choice decisions is unable to reveal the full range of factors which influence developers’ decisions on the form of retail outlets, and the influence of planning policies and schemes that steer developments to particular locations and/or formats.

Therefore, Chapter 9 will describe a series of semi-structured interviews designed and conducted in order to gather professionals’ views on these matters. Semi-structured interviews were chosen for this task because, compared to structured interviews, they make it easier for the interviewer to communicate with the respondent and ask questions about different aspects of the retail environment.

Chapter 4 Research Methodology 91

Interviews will be conducted with planning professionals and experts in the public and private sectors who specialise in retail and commercial planning. Interviews with owners, managers and developers of shopping centres and major retail outlets will focus on identifying crucial factors which affect the form or expansion of a shopping centre and the limitations or advantages that planning policies impose on the process. Interviews with members of planning bodies (urban/transport planers) will aim to identify the objective and role of planning policies and regulations, as well as identifying potential shortcomings and deficiencies that need to be addressed.

4.3.1.6 Measuring the impacts of possible future scenarios on travel behaviour of customers using the travel demand modelling techniques (Chapter 10)

Interviews with professionals should enable an understanding of the possible future form and distribution of retail in the city to be constructed. However, the impacts which future retail form and distribution will have on travel behaviour and transport mode share also require serious examination.

A transport model can be used to predict how travel behaviour and transport mode share will change in response to changing attributes in the retail environment. The Brisbane Strategic Transport Model-Multi Model (BSTM-MM) has been developed by Brisbane City Council and the Department of Transport and Main Roads specifically for this purpose. The BSTM-MM model is generally available and commonly used for this purpose. The BSTM-MM model uses a four step modelling (FSM) pattern of trip attraction, trip distribution, mode choice and trip assignment to provide a ‘more holistic’ assessment of shifts in trip pattern, based on the derived data from real-world datasets. The model is widely used to model and predict changes in the transportation system, and is an important planning and decision support system.

Models similar to the BSTM-MM are commonly used for urban planning forecasts. However, there are also some acknowledged deficiencies with these types of model. The BSTM model relies on the fact that land use generates activities such as housing, working, shopping and leisure. Individuals’ desire to participate in these activities and the spatial distribution of these activities then

Chapter 4 Research Methodology 92 generates the need to travel. These projected travel needs then interact with other influential factors such as the cost of travel (time or money) and socio- economic factors (such as household income, size, car affordability, etc.) to affect the preferred travel behaviour of people in the city (Gachanja, 2010). The FSM is one of the major models that transport planners rely on to predict the effects of modifications in the transport system of cities. The FSM model can reveal how some trends might encourage people to use more sustainable means of transport for accessing various destinations including retail/shopping in the city. Burke et al. (2010) used the BSTM model to investigate the impacts of moving government workers and employment decentralization in Brisbane.

Fig 4-6: Methodologies applied in the thesis to study retail travel behaviour

Chapter 10 will use BSTM_MM to analyse future retail travel behaviour in Brisbane under likely future scenarios for the retail and transport environment. The model will assess the impacts of policies which encourage the development of different urban planning scenarios.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 95

5.1 Introduction

During the last few decades, as discussed in the literature review chapter, substantial transport planning research has been done on the impacts of substituting daily/weekly car trips with more sustainable alternatives, considering both groups of working and non-working trips. In an Australian context, despite the fact that a large volume of personal travel directly or indirectly originates from retail activities, there has been little attention given to this trip type in the travel demand management programs (TDM), an example being the last two decades of Travel-smart programs. Australian cities are challenged by the expansion and changing form/structure of the retail environments and their potential impact on customers’ travel behaviour, air pollution and the amount of fuel consumed in accessing these destinations.

Current urban planning regulations typically do not support alternative distribution and forms of retail outlets. The expansion and locations of car- based retail forms in the city are likely to ‘lock-in’ continuation of motorised travel for this trip-task in future. Influencing Australian retail travel behaviour is challenging, due to the decades long expansion of car-based retail environment types located on major roads and highways, which rely on, and in turn reinforce, car ownership and use. Today some of these centres challenge the conventional central business district in terms of their retail offerings.

This chapter aims to contribute to this debate by exploring travel to retail destinations within the Brisbane Statistical Division (BSD) using 7-day South East Queensland Household Travel Survey (SEQ-HTS) data from 2009. As a starting point in understanding the retail travel pattern and how it can be influenced by future policies, customers’ travel behaviour should be investigated in more detail. Statistical analysis is performed to examine trip frequency, trip complexity, destination choice, and mode of transport for this trip type. Travel differences between weekday and weekend trips are explored, as are differences between trips made by males and females and trips made to purchase different products. Shopping centres and supermarkets as two major destination choices of trip makers (based on the analyses results) are studied in

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 96 more detail. The results have the potential for planning initiatives to achieve more sustainable travel behaviour by influencing the geography and form of retail outlets.

The chapter begins with a brief literature review on the household travel survey and how it is being applied, in order to explore various aspects of the transport environment in Australia. This section is followed by a description of applied data and methods. The following section is dedicated to the data analysis on investigating customers’ travel behaviour for various types of retail trips and the relevant findings. The chapter concludes with a summary of the significant findings, and forms the basis of the next stage of the research.

5.2 Literature review

As previously mentioned in the literature review chapter, the development of the disaggregated data made a substantial difference in terms of transportation studies, especially those which focused on travel behaviour.

Household travel survey data usually provides information on current demographic, employment and travel patterns. These data are used as inputs to transport and land use planning and policy making forecasting models. They can be collected in various measures, from the smaller scale of the cities to those at the state or national level. Surveys were held to look at the individual (socio-economic, demographic, etc.) and household characteristics (HH size, HH structure, etc.) of the interviewees and their travel behaviour. As well, a diary of their journeys on given day/days (origin and destination of the trip, start and end time of the trip, mode of travel, purpose and place of the trip, etc.) was kept. Some of these surveys collected data for each household member or each person sampled for one day of travel, while others brought together the data on each individual during one full week.

In Australia, travel surveys are frequently undertaken in each of the major metropolitan areas. No national travel survey data exists in this country, and there are differences between different states in terms of the methodology and the detailed information that is collected.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 97

While many of the travel surveys taken around the world are being collected to provide a description of the way people travel, determine the long-term trends in travel and provide information for planning and sustainable development, the major goals of most metropolitan travel surveys in Australia are principally to serve the needs of modelling and policy formulation (Stopher et al., 2011).

The disaggregated information available in these types of datasets enables the researchers to investigate various aspects of travel patterns and behavioural aspects of the trips being made in the area, such as the distances being travelled and the selected mode or trip time (on-peak, off-peak), etc. This information enables analysts to measure the general vehicle kilometres travelled, public transport kilometres travelled (PTKT), and cycling or walking distances travelled, and allows them to measure the level of emission or fuel consumption and evaluate the total price of trips made by various modes. These data are basic requirements for developing more precise forecasting models (compared to zone based models) to help planners and policy makers be aware of the influences of their decisions on the future environment.

Various studies have investigated the travel behaviour of people in Australian cities and metropolitan areas using these types of travel surveys. In 2011, McGeoch investigated the travel behaviour of Melbourne people, as reported in the 1978/79 Melbourne Home Interview Travel Survey. The Victorian Integrated Survey of Travel and Activity, conducted some 30 years later in 2007/08, was undertaken in order to gain an understanding of the shifts in travel patterns and behaviour of Melbournians during this period (McGeoch, 2011). In another study by Olaru et al. (2005), the relationships between daily travel, urban facilities and activity spaces for individuals/households were studied using Sydney Household Travel Survey (SHTSD) data from 1997 to 2002. The SHHTS data provided information about the travel activities and the individual/household characteristics of the trip makers.

Raimond and Milthorpe (2010) studied trends in people’s travel behaviour and the factor of holding a driving licence for younger drivers. He applied the Sydney Strategic Travel Model (STM) with a licence holding sub-model for the prediction of future levels of licence holding. The model was developed based

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 98 on Sydney Household Travel Survey (HTS) data. His research aims to explore the reasons why young people under 35 are becoming less likely to hold a driver’s licence .

Shaz and Corpuz (2012) used the SHTSD, which is the largest and most comprehensive source of personal travel data for the Sydney Greater Metropolitan Area (GMA), to study the trips made across the entire region on an average weekday by workers who reside in the GMA. The same dataset, comprising information on people’s travel behaviour, parking behaviour, mode of travel and mode choice, parking location and costs, was also applied by this researcher in 2012 to compare the driving trips and parking behaviour of drivers to parking space levy and non-parking space levy locations across the city (Hay and Shaz, 2012).

Zhao et al. (2013) investigated the relationship between fuel price, land use characteristics and the level of household transport greenhouse gas (GHG) emission in Sydney Metropolitan Area. They researched ways in which the application of new policies might reduce the level of GHG emissions in a short- term period. This study was also based on the Sydney Household Travel Survey (HTS) data, which has been conducted every few years since 1997/98.

Apart from the simple analysis that can be done using these datasets and which are quite considerable in number, HTS are also used to develop a large number of disaggregated transport and planning models including the choice demand logit models, generally look at the mode-choice, destination-choice of trips makers and also a large number of other transportation forecasting models. While the number of studies in Australia is significant and growing every day, hardly any study in the existing literature focuses on retail trips.

5.3 Methods

As previously explained in Chapter 4, the Household Travel Survey (SEQTS) data (2009) covering the 7-day travel of people in South East Queensland and provided by the Queensland Department of Transport and Main Roads, was applied in this study. The survey comprises 4,240 residential households within

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 99 the Brisbane Statistical Division (BSD) area (The Urban Transport Institute, 2010).

For the purpose of this study, following the analysis of all trips, the dataset was manipulated to identify all retail trips, excluding trips to purchase petrol. Retail trips were defined as those trips which had a ‘destination place’ of ‘shopping’ and a ‘destination purpose’ of ‘buying something’. This excluded those trips made to shopping locations that were for non-retail purposes, such as trips to shopping centres for personal services (banking, mail collection), libraries or to dine out1.

A total of 3,354 retail trips were identified within the sample. The data was weighted for weekdays and weekends trips separately, considering 11 regions in the BSD area for weighting and expansion. Therefore, the 3,354 trips in the original sample represent 1,436,533 trips made by the regional population after the weighting is applied. Final retail trips include different types, namely, home- based shopping (1596 trips - 47%), work-based shopping (283 trips - 9%), shopping-based shopping (880 trips - 26%) and others-based shopping (595 trips -18%) for both weekdays and weekends.

After the data was collected the analyses were done using the SPSS software (Statistical Package for the Social Sciences) and the results were examined in order to illustrate a better understanding of the current retail travel behaviour of people in Brisbane.

Statistical analyses were undertaken to investigate various characteristics of retail trips, including trip frequency, trip complexity, destination choice, trip time and mode of transport for different days of the week and for various gender and age groups. Various types of purchased products and the destination types selected by trip makers for each were also investigated.

1There is one limitation in that which is an unknown but presumably small proportion of all retail trips that may not actually involve a purchase, and it is not certain that these are correctly captured either in respondent’s diaries or the coding provided by the data providers.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 100

Fig 5-1: BSD boundary as the selected study area

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 101

5.4 Results and Discussion

As explained above, the BSD travel survey was carried out for a sample of 4,240 BSD households. After applying the weightings, the total number of trips taken by these households increased from 20,440 to 7,519,823 trips1. Among ten categories of trips undertaken for various purposes, retail trips represent a very important part of everyday travel (Fig. 2). Retail trips comprise almost 16 and 29 percent of total trips on weekdays and weekends, respectively, accounting for the second biggest category on weekdays (after the journey to work) and the largest one for weekend trips.

Weekends 25 Weekdays

Trip Percentage Trip 10

Social

Other

Buy

Business

Personal

Education

Something

Accompany

Recreational

Pickup/Drop

offSomeone

Pickup/Deliv Work-related er Somethinger

Fig 5-2: Trip Frequency by Trip Purpose during Weekdays & Weekends

As table 5-1 shows, for the trips with the purpose of “buy something”, 68 percent and 16.5 percent of the trips (almost 85 percent) were taken by a driver or a passenger, respectively, while the percentage of walking trips was 10 percent. The trip mode-share on weekends is even higher, with car trips accounting for 90 percent of both vehicle driver and passenger followed by just 8 percent of walking trips. “Buy something” trips taken by private cars are among the highest after “pickup/drop-off someone”. Other means of transport

1 It represents all the trips excluding the ones with the purpose of survey home and changing mode in the regional scale.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 102 such as cycling and taxi have a very low share for this trip type, except in the case of public transport which counts for approximately 4 percent of all shopping trips on weekdays. Public transport mode share decreases to 1.22 percent on weekends for retail trips, giving it the lowest score among all the other trips that are made by public transport.

Other types of trips, including the work-related and education trips, show a higher percentage of public transport usage. For the active modes of transport, education trips experience a more sustainable situation compared to buy something trips, while work-related trips lag even further behind.

Table 5-1: Trip Frequency by Trip Purpose and Mode-share during Weekdays & Weekends *

Vehicle Vehicle Public Walking Bicycle Taxi Other Total Driver Passenger Transport1 9.02 76.99 11.04 0.61 0.25 1.72 0.37 100 Accompany (6.68) (84.36) (7.00) (0.49) (0.00) (1.47) (0.00) (100)

Buy 68.08 16.51 10.11 0.53 0.29 3.87 0.61 100 Something (60.00) (29.49) (7.93) (0.54) (0.07) (1.22) (0.75) (100)

Pickup/Deliv 60.48 24.93 8.75 1.59 0.00 4.24 0.00 100 er Something (63.35) (29.19) (6.21) (1.24) (0.00) (0.00) (0.00) (100)

Pickup/Drop- 93.58 4.25 1.27 0.00 0.00 0.80 0.09 100 off Someone (86.56) (11.29) (0.27) (0.00) (0.00) (0.54) (1.34) (100) 6.08 53.44 15.01 2.86 0.22 12.96 9.44 100 Education (40.00) (20.00) (15.00) (5.00) (0.00) (20.00) (0.00) (100) 71.62 6.24 7.75 1.47 0.17 11.06 1.69 100 Work-related (68.49) (16.89) (5.94) (0.91) (0.46) (6.39) (0.91) (100)

Personal 46.31 32.20 13.11 0.62 0.78 5.59 1.40 100 Business (46.9) (39.61) (9.85) (0.64) (0.43) (1.28) (1.28) (100) 48.34 26.83 17.29 0.41 1.18 5.67 0.28 100 Social (43.16) (43.48) (8.03) (0.72) (1.27) (2.62) (0.72) (100) 43.73 27.35 21.23 1.85 0.00 3.28 2.56 100 Recreational (37.77) (34.24) (18.75) (6.25) (0.00) (2.45) (0.54) (100) 40.00 20.00 10.00 0.00 0.00 30.00 0.00 100 Other (66.67) (33.33) (0.00) (0.00) (0.00) (0.00) (0.00) (100) 1 Public Bus, Ferry, Train * Weekends percentage are presented in parenthesis

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 103

In relation to the gender based travel behaviour differences, during the working week, 195,872 and 302,056 trips are made by women and men over 18, respectively, to shopping destinations. These trips almost double in number for women, increasing to 411,316 while rising to 466,819 for men on the weekends. This shows that men, in comparison to women, travel to retail spaces 1.5 times more on weekdays and 1.13 times more on weekends.

Male 70 Female 60

Retail trip percentage percentage trip Retail 30

0 Vehicle Vehicle Walking Bicycle Taxi Public Other Driver Passenger Transport

Fig 5-3: Retail Trip Frequency by Mode-share for 18+ Residents-Weekdays

80 Male 70 Female 60

50 percentage percentage 40

30 Retail trip trip Retail 20

0 Vehicle Vehicle Walking Bicycle Taxi Public Other Driver Passenger Transport

Fig 5-4: Retail Trip Frequency by Mode-share for 18+ Residents-Weekends

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 104

On weekends, a higher rate of car usage (either as a driver or a passenger) is apparent among both women and men, but the noticeable point is that about 17 percent of female drivers shift to becoming passengers on the weekends, giving a total rise of 20 percent.

There is about a 3 percent increase in male walking trips during the working week in comparison to weekends, but this percentage does not change for females. This is due to the shopping trips that occur when males are on the journey to or from work. In terms of using public transport for retail trips, there is a rise of up to 3.5 percent during the weekdays, while women have a greater share in these types of trips. Taxis comprise less than 0.3 percent for both groups, which is somewhat at odds with the role of taxis for retail trips. Considering the total number of car trips on weekdays and weekends, it is proved that men and women do not show a great deal of difference (in terms of their retail mode choice).

Going one step further in realizing shopping travel behaviour in Brisbane, we need to understand people’s travel behaviour in different urban contexts based on the level of accessibility to retail destinations. The inner Brisbane area, inner north suburbs and inner south suburbs (regions 1 & 2) including the Central Business District (CBD), experience higher density of retail establishments and is believed to be more sustainable in terms of using active means of transport (walking and cycling) in comparison to the outer Brisbane suburbs, formed on the base of scattered distribution of land use.

A simple examination of weekday mode-share for retail trips shows that private cars are used for about 87 percent of trips in outer Brisbane, but almost 71 percent of trips in inner Brisbane. While both percentages still show a high level of car dependency, there is a 16 percent difference between inner and outer Brisbane in terms of car based retail trips.

On weekends, total auto trips account for 94 percent of trips in outer Brisbane and 74 percent in inner Brisbane. This shows a 20 percent higher level of car- reliability for outer Brisbane compared to inner Brisbane, which is even worse than during the working days. During weekdays, walking has a considerable mode share of 19 percent in shopping trips for inner Brisbane, which increases

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 105 to an even higher number of almost 25 percent during the weekends. In outer Brisbane, walking drops to less than 10 percent during the week. Figure 5-5 shows that public transport plays a minor role in weekday shopping trips, accounting for only 7 percent of trips in inner Brisbane and 3 percent in outer Brisbane. Based on Figures 5-5 and 5-6, the use of other means of transport such as taxis and bicycles are so insignificant that they can be ignored, especially on weekends.

70 Inner Brisbane Outer Brisbane

Retail trip percentage percentage trip Retail 20

0 Vehicle Vehicle Walking Bicycle Taxi Public Other Driver Passenger Transport

Fig 5-5: Retail Trip Frequency by Mode-share in Inner & Outer Brisbane - Weekdays

70 Inner Brisbane Outer Brisbane 60

30 Retail trip percentage percentage trip Retail 20

0 Vehicle Vehicle Walking Bicycle Taxi Public Other Driver Passenger Transport

Fig 5-6: Retail Trip Frequency by Mode-share in Inner & Outer Brisbane - Weekends

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 106

The share of shopping centres and supermarkets as retail trip destinations in Brisbane is an endorsement of the key role these centres play, as discussed previously in the literature review. Figure 5-7 provides a summary of the trip frequency by shop type. Among all shopping trips with the purpose of “buy something”, supermarkets and shopping centres are the destination for almost 60 percent of weekday trips and nearly 50 percent of weekend trips, based on the HTS data for destination place. The next most important destination is food stores, accounting for approximately 7 percent of trips, both on weekdays and weekends.

The results are a good indicator of the important role of large shopping centres and supermarkets in the typically two purchasers’ household lifestyle of Brisbane people, trying to provide them with one stop inclusive destinations for their daily or weekly multipurpose shopping.

Weekdays 30 Weekends 25

Percentage of all Retail trips Retail all of Percentage 5

Video

others

Chemist

Markets

Furniture

FastFood

Hardware

Food Store Food

Homewares

Liquor Store Liquor

Vehicle Sales Vehicle

Supermarket

Petrol Station Petrol

Car Accessories Car

Shopping Centre Shopping

Clothing & Shoes & Clothing

Sports & Outdoor & Sports

ConvenienceStore

Computers & Music & Computers Music & Computers

Domestic Appliances Domestic Newsagency & Bookstore & Newsagency

Department & Discount Store Department& Fig 5-7: Trip Frequency by Shop Type during Weekdays & Weekends

As explained in the SEQTR report, “multipurpose stops within regional undercover shopping centres have been simplified to a single trip to the shopping centre, irrespective of the number of different activities undertaken while at the shopping centre” (The Urban Transport Institute, 2010). Therefore,

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 107 we do not have access to the information about the shop type that people visited inside the shopping centre. However, the large percentage of trips to these destinations, in comparison to the number of trips to the other shop types outside the shopping centre and the variety of goods purchased, attest to the diversity of goods available in these regional shopping centres.

Hardware, fast food, petrol stations, excluding trips to buy petrol (convenience shopping) and department and discount stores, as well as chemists, newsagencies and bookshops constitute the next most popular shopping destinations. While the same trend is perceptible in shopping centres and supermarkets on weekends, the numbers of shopping trips for hardware, domestic appliances and furniture that primarily take place to big boxes and warehouses (e.g., Bunnings, IKEA, Kmart, Amart) suggest that this is clearly a weekend activity.

Markets, as would be expected, experience an increase of 2 percent on weekends in comparison to weekdays, since most people go to these places at weekends to buy their necessities more cheaply for the whole week.

The SEQTS 2009 has made it possible to provide information about the types of products being paid for as a result of the retail trips to different destinations, from a shopping centre to a supermarket, convenience store or a small local shop (Fig. 5-8). As expected, groceries and food are the most significant products shopped for during weekdays and weekends, facing a reduction of about 18 percent and 13 percent on weekends, respectively. Other categories, including alcohol, tools and hardware, newspapers and tobacco, household furnishings and fruit and vegetables that are mostly supplied at warehouses and markets, are often purchased at weekends.

It is interesting to see that daily household needs, such as groceries, food, milk, bread, confectionary, non-alcoholic drinks and pharmaceutical products create a higher percentage of trips during weekdays rather than weekends, since there is an everyday need for these products. It seems that people are happy to make frequent trips to purchase these items rather than buying them all at the weekends.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 108

35 Weekdays 30 Weekends

10 Retail trip percentage percentage trip Retail 5

Milk

Food

other

Bread

Fruit &…

Clothes

Tools&…

Books&…

Personal…

Groceries

Take-away…

Household… Household… Household… Household…

Meat&Fish

Recreational…

Audio-Visuals

Non-alcoholic…

Confectionary

Newspapers&…

Pharmaceutica…

Vehicle Runnin…Vehicle AlcoholicDrinks Animal Expenses

Fig 5-8: Trip Frequency by Expenditure Code during Weekdays & Weekends

5.4.1 Customers’ travel behaviour to shopping centres & supermarkets

Regarding the abovementioned analysis, between 50 to 60 percent of the trips made during the week take place to a shopping centre or a supermarket, and these are the two major categories of retail destinations. Most of these trips are mostly considered to happen for multipurpose and one stop shopping. They are mostly made by car, to give consumers the opportunity to buy goods in quantity. Consequently, walking and public transport are considered by many people not to be an appropriate substitute for car trips, and therefore, these trip types are investigated further in this section.

In regards to the HTS data, the type of products mostly bought in these destinations and the type of mode share for travelling to these destinations can be investigated. It is also possible to see whether there are differences in people’s preferences for mode share when travelling for various distances, and to investigate whether having a supermarket or shopping centre close to the customers’ house will affect the mode share they use.

The analysis of the purchased products at these centres (Fig. 5-9) shows that of all the trips to shopping centres, about 46 percent (weekdays) and 40 percent

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 109

(weekends) are to buy groceries. The next major category of products bought is food and clothes.

50 Weekdays Weekends

10 Percentage of Retail trips trips Retail of Percentage

Milk

Toys

Food

Bread

Shoes

Others

Clothes

Groceries

Household…

Audio-visuals

Meatand fish

Clothing(NEC)

Pharmaceutical…

Personal Goods…

HouseholdGoods… Non-alcoholicdrinks Booksnewspapers &

Fig 5-9: Trips to Shopping Centres by the Expenditure code

For supermarkets, grocery shopping increases to 63 percent and 59 percent, respectively for weekdays and weekends (Fig 5-10). Food consists of about 15 percent of trips made during the week, and is followed by other edible products such as bread, milk, confectionary, etc. as the next important shopping trip category. These results suggest that grocery shopping should be a focus of any future efforts to reduce car-based retail travel.

A more detailed examination of Brisbane shopping trips shows that travel mode is dependent on trip length. Figure 5-11 shows that 8 percent and 16.5 percent of total retail trips made to shopping centres on weekends and weekdays, respectively, are within less than 1 km, which is clearly a walkable distance. Of all these trips, 6.6 percent are made by walking on weekdays.

The number of trips undertaken by walking drops by one-sixth (about 1.6 percent) on weekends, which shows people’s high preference for cars for retail travel. For trips of more than 1 km, travel is primarily limited to cars with little use of other modes. The role of public transport is more noticeable by the

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 110 increase in trip distance and comprises 4 percent of shopping trips on weekdays for trips greater than 5 km.

60 Weekdays

Weekends 50

Percentage of Retail trips Retail of Percentage 10

Milk

Toys

Food

Bread

Others

Groceries

products

Meatand fish

Confectionary Pharmaceutical

Non-alcoholicdrinks Fig 5-10: Trips to Supermarkets by the Expenditure code

In the case of supermarkets (Fig 5-12), shopping destinations within distances less than 3 km comprise more than 55 percent of retail trips. There is a large difference for the percentage of retail trips’ mode share to distances less than 1 km compare to shopping centres. While the number of trips within distances of less than 1 km is increasing to 28 and 24.5 percent for weekends and weekdays, respectively, walking is playing a more evident role in this type of trip. For distances of less than 1 km, walking constitutes about 7.5 percent of trips made during the week and 9 percent on weekends. For trips of 1 to 3 km, walking accounts for almost 4 percent of weekday trips and 1 percent of weekend trips.

Public transport is not an important factor, except in trips that were longer than 5 km, where there is a slight rise to 2.2 percent for the mode share. The results show that for distances of more than 1 km, people are mostly using other mode shares such as private cars, and to a much less extent, public transport.

Figure 5-13 shows the three super regional centres in Brisbane, including

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 111

Westfield Garden City, Westfield Carindale and Westfield Chermside and the Euclidean buffer with 1, 3 and 5 km radius around them. It includes the collection districts (CDs) that the trips have originated from. As can be seen, most of the origin zones for these trip types are located out of the 1 km buffer area and within the 5 km distance from the centres. It shows that walking to these centres has not been an option according to the analysis on the mode choice of retail travellers.

40 35

25 20 15 10

Retail trip percentage trip Retail 5 0 WE WD WE WD WE WD WE WD 1 km >1 km & 3 >3 km & 5 >5 km Vehicle Driver Vehicle Passenger Walking Public Transport Bicycle Taxi Other

Fig 5-11: Trips to shopping centres by Mode-share & Distance during Weekends & Weekdays

Retail trip percentage trip Retail 10

0 WE WD WE WD WE WD WE WD 1 km >1 km & 3 km >3 km & 5 km >5 km Vehicle Driver Vehicle Passenger Walking Public Transport Bicycle Taxi Other

Fig 5-12: Trips to Supermarkets by Mode-share & Distance during Weekends & Weekdays

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 112

Fig 5-13: Brisbane super regional shopping centres (with over 85,000 square metres gross lettable area for retail) and their customer originating zones

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 113

5.4.1.1 Shopping centres’ hierarchy share in retail trips

The Shopping Centre Council of Australia (SCCA) in its 2011 report “Productivity Commission Inquiry into the Economic Structure and Performance of the Australian Retail Industry” alleged that “shopping centres comprise only 38% of total retail space; make up only 19% of all retail locations and around 35% of all retail shops; and generate 40% of total retail sales in Australia” (Shopping Centre Council of Australia, 2011). This includes not only the regional and major shopping malls but also smaller shopping centres made up of a collection of small shops and a major supermarket.

When it comes to Brisbane as one of the cities experiencing a large growth in the number and size of the shopping centres, this number would be more considerable.

Using a similar category for shopping centres as that employed in the SCD, the mode choice decisions of customers for travelling to each of these destinations’ hierarchy was investigated. The major regional and regional centres are considered in one group as major regional centres. The themed, outlet and market centres are not considered in this research since there is only 1 market and 3 themed centre reported in the Brisbane Statistical Division (BSD) in the SCD. The total number of other categories of centres (super-regional, major- regional, regional, sub-regional and neighbourhood centres) reported for BSD by SCCD for 2011 is 194.

In Brisbane, based on the SEQTS 2009, the trips to large shopping centres (super and major regional) comprise about 16 percent and 14 percent of retail trips made during weekdays and weekends, not including trips to the CBD which stands for 3 and 2 percent of trips separately. It is worth noting that only the three super regional shopping malls, namely, Westfield Garden City, Westfield Carindale, and Westfield Chermside comprise 4-5 percent of the total number of retail trips, which is about one third of the share of the weekday trips to major shopping complexes and more than half of the weekend trips.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 114

(a) weekdays (b) weekends

4% 2% 3% 5% 12% 9%

11% 12% 6% 65% 67% 4%

City Centre Super Regional Major Regional Sub Regional Bulky Goods Others

Fig 5-14 Major Shopping Complexes’ share of the total Retail trips

Among all the shopping trips made on the weekends, about 2 percent end in the city centre as the shopping destination, while 5 percent and 9 percent are going to the super regional and major regional/regional centres, respectively. During the weekdays the trip numbers change to 12 percent for the major regional/regional centres, while 4 percent are allocated to super regional centres and there is 1 percent increase in the number of trips to the city centre (3 percent). While there is only 1 percent decrease or increase in the percentage of trips to the city centre and to super centres on weekdays and weekends, major regional/regional centres are experiencing a 3 percent increase during the weekdays. This shows that the super-regional centres are preferred on the weekends, since people have to travel farther but they can go with their family and spend more time there, enjoying themselves, dining or going to the cinema. Sub-regional medium sized centres including a major supermarket play an important part in attracting customers during the week. Around 11-12 percent of retail trips are being made to these destinations. The 18 bulky goods centres reported in the SCCD are the trip destinations of customers at a ratio of about 4 and 6 percent on weekdays and weekends, respectively.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 115

The major shopping complexes considered in this analysis and within the BSD include: Brookside Shopping Centre (Mitchelton); Capalaba Central Shopping Centre (Capalaba); Capalaba Park Shopping Centre (Redland Bay); Centro Toombul (Nundah); Logan Hyperdome (Shailer Park); Morayfield Shopping Centre (Morayfield); Mount Ommaney Shopping Centre (Mount Ommaney); Redbank Plaza (Redbank), Riverlink Shopping Centre (North Ipswich); Toowong Village (Toowong); Westfield North Lakes (North Lakes) and Westfield Strathpine (Strathpine); Grand Plaza (Browns Plains); and Aspley Hypermarket.

The sub-regional centres include Alexandra Hills Shopping Centre, Arana Hills Kmart Plaza, Arndale Shopping Centre (QLD), Booval Fair Shopping Centre, Bribie Island Shopping Centre, Caboolture Park Shopping Centre, Cannon Hill K-Mart Plaza, Centro Lutwyche, Centro Springwood, Centro Taigum, Deception Bay Shopping Centre, Great Western Super Centre, Inala Plaza Shopping Centre, Kenmore Village Shopping Centre, Logan Central Plaza, Logan City Centre, Mt Gravatt Plaza, Orion Springfield, Peninsula Fair, Stafford City Shopping Centre, Stockland Cleveland, Sunnybank Hills Shoppingtown, Sunnybank Plaza Shopping Centre, Underwood Marketplace, Victoria Point Shopping Centre, Windsor Homezone and Wynnum Plaza Shopping Centre.

A close inspection of the travel behaviour of the centres’ customers during the week shows that trips to the city centre are the most sustainable trips in terms of using active and public transport modes (Fig. 5-16). Around 54 percent of trips are undertaken by walking, 35 percent by public transport while the other 11 percent take place by car, either as a driver or as a passenger. This can be due to the higher price of parking in the city centre and the more dense and mixed-use urban form in this area. For the super-regional centres, the share of public and active transport suddenly drops to around 11 percent, and the rest is allocated to private cars. Major-regional and regional centres experience an even worse situation, with around 95 percent of travel happening by car.

The higher level of use of public transport at super regional centres can be related to availability of public transport (bus stations) at two out of three of these centres. In other types of retail spaces including other levels of shopping

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 116 centres, convenience stores, supermarkets, bulky goods stores, strip shopping areas, access by about 85 percent of car trips shows a better but still unsustainable situation for these centres, in comparison to the previous two categories.

Fig 5-15: CCDs* including Regional and Major Shopping Complexes

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 117

100 90

70 60 50 40 30 20 10 Retail Retail Percentage Trip 0 WDS WEs WDS WEs WDS WEs WDS WEs WDS WEs WDS WEs City Centre Super Major Sub Regional Bulky Goods Others Regional Regional & Regional

Vehicle Driver Vehicle Passenger Walking Public Transport Taxi Bicycle Others

WeekDays (WDs), WeekEnds (WEs)

Fig 5-16: customers’ retail trip mode share during the Weekdays

During the weekends, city centre shopping relies on car trips by about 60 percent while there is hardly any sign of public or active transport for the super- regional and major-regional centres. Other mode share available for trip makers service only 10 percent of non-private-car based trips. If this trend continues there will be little to no chance of a more sustainable transport future for the large amount of these weekly trips.

It is important to note that different from the trips to school and work, almost half of these retail trips are initiated within off-peak hours (between 9am and 3pm and between 7pm and 5am), while the other half of the peak hour trips considerably impact traffic congestion in Brisbane (Fig 5-17 and Fig 5-18).

42%

58%

Peak Hours Off Peak Hours

Fig 5-17: Percentage of trips during on-peak & off-peak hours on weekdays

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 118

70 Peak Hours

60 Off-Peak 50 40 30 20

Retail Retail PercentageTrip 10 0 City Centre Super Regional Major Sub Regional Bulky Goods Others Regional & Regional

Fig 5-18: Weekdays’ trip percentage for on-peak and off-peak hours by centre’s category

Furthermore, more time is spent on shopping within the larger centres (Fig. 5- 19). People spend more hours in super-regional and major-regional centres, and this might be related to the further distance of these centres from people’s homes. It could also be related to the recreational identity of these centres, giving people the opportunity to shop and to socialize.

In terms of trips to the super regional centres, about one fourth of trips are longer than 3 hours duration, while around 50 percent are between 1 to 3 hours. In the case of the weekends trips to major regional and regional centres, the percentage of trips with a duration of 1 to 3 hours, and trips longer than 3 hours both drop by about 2 and 5 percent, respectively. This can be explained by the fact that people prefer to go to larger, super-regional centres when they have a full day off.

The city centre is slightly different. Because it is close to offices and working areas, many people can still travel here during off peak hours. It is not easy for working people to travel to a major or super regional centre unless it is close to their workplace. About 12 percent of trips to the city centre are of more than one hour in duration, which maintains it as a destination for short trips by more than 75 percent. The high cost of parking in the city centre can discourage people from spending a long time shopping in this area.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 119

100 90

70 60 50 40 30 20 Retail Trip Percentage Trip Retail 10 0 WDs WEs WDs WEs WDs WEs WDs WEs WDs WEs WDs WEs

City centre Super regional Major Sub-regional Bulky goods Others centre regional/Regional centre

centre Less than 1 hour Between 1 to 3 hours More than 3 hours

WeekDays (WDs), WeekEnds (WEs)

Fig 5-19: Time spent in the retail destination

Trips of more than 3 hours duration rarely happen in the smaller type of centres such as sub-regional or neighbourhood centres. For the smaller shopping destinations, around 90 percent of trips are shorter than 1 hour. There is only a small shift in the number of hours spent in these types of destinations. This shift is unnoticeable and is mainly within the 1 hour time limit.

More than 50 percent of the weekdays trips to the super-regional and major- regional centres and bulky goods stores are home-based shopping, while work- based shopping trips are limited to around 8, 11 and 15 percent of trips, respectively (Table 5-2). During the weekend, work-based shopping trips are mostly replaced by the shopping-based shopping trips.

About 40 percent of shopping trips in the city centre during weekdays are work- based shopping trips. This highlights the different nature of these types of trips compared to those to the other locations in the city. During the weekends, this number reduces almost to 2 percent while the percentage of trips with a period of 1 to 2 hours duration increases by about 40 percent (Fig. 5-19). This trend can be related to the fall in the cost of parking and the rise in the number of shopping-based shopping trips taking place on weekends.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 120

Table 5-2: Retail trip percentage to various shopping centres based on their trip origin

City Super Major Regional Sub Bulky Other Centre Regional & Regional Regional Goods s Home WDs 30.58 51.17 52.80 42.72 48.60 41.74 Based WEs 41.51 48.81 61.09 54.29 46.45 49.13 Shopping Shopping WDs 12.35 22.90 15.77 25.93 17.69 18.70 Based WEs 20.51 22.42 16.28 15.97 17.94 18.28 Other Shopping WDs 17.20 17.67 19.96 25.21 18.72 26.10 Based WEs 36.35 28.76 22.25 29.74 35.61 30.70 Shopping Work WDs 39.87 8.25 11.47 6.14 14.98 13.46 Based WEs 1.63 0.00 0.39 0.00 0.00 1.89 Shopping These results can be of great significance to the transport planners in setting policies such as determining the cost of parking at the destinations. Recently all of the super-regional centres have begun a new policy, charging customers who park for longer than 3 hours in centres’ parking areas. It is important to see what percentage of trip makers will be affected by these policies, and whether or not they change their trip pattern.

Categorizing shopping trips based on reported purchased items will clarify the role that is expected of these centres associated with their function, level of utility/attraction for the customers or any other unrecognized influential factor. Nine groups of products are identified and the share of trips going to any of these shopping centres is then measured (Fig. 5-20, Fig. 5-21). The graphs show that a high percentage of trips to different levels of centres are related to shopping for groceries, clothes, household goods, food and personal goods.

The most popular locations for clothes shopping are the city centre and super- regional centres, followed by the major-regional and regional centres. During the weekdays the city centre and super regional centres have almost the same share of clothes shopping, there is a considerable increase in the percentage of trips to super-regional centres on weekends. Grocery shopping is a major part of all trip destinations, but it is obvious that it comprises the biggest portion of shopping trips at sub-regional centres during the week, followed by the food category. In the category of bulky goods, household goods are the most prevalent item. For the major regional and regional centres, clothes, food and

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 121 groceries seem to be of a similar high level of importance. It should be noted that trips to buy food do not include dining or drinking trips, since these have been categorised as recreational trips in HTS, which makes understanding this type of trip difficult.

100 90

80 70 60 50 40 30 20 10 Retail Trip Percentage Trip Retail 0 N/A, 7.54 City Centre Super Regional Major Sub Regional Bulky Goods Others Regional & Reginal

N/A Food Groceries Alcohol drinks Clothing (NEC) Transport (NEC) Personal Goods (NEC) Household Goods (NEC) Other (NEC)

Fig 5-20: Percentage of retail trips to various types of shopping centres based on the reported purchased items - Weekends

100

90 80 70 60 50 40 30 20

Retail Trip Percentage Trip Retail 10 0 City Centre Super Major Sub Regional Bulky Goods Others Regional Regional & Reginal

N/A Food Groceries Alcohol drinks Clothing (NEC) Transport (NEC) Personal Goods (NEC) Household Goods (NEC) Other (NEC)

Fig 5-21: Percentage of retail trips to various types of shopping centres based on the reported purchased items - Weekdays

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 122

5.5 Conclusion

The results show the current trends in the travel behaviour of customers in Brisbane and suggest a number of key directions for making retail travel more sustainable. Major findings from the first part of analysis with the focus on the overall retail environment can be summarized as follows.

First, retail travel is the most unsustainable travel in terms of the proportion of trips made by car. While around more than 85 percent of these trips are undertaken by car, the role of active and public transport is up to 10 and 4 percent, respectively.

Second, no considerable difference is recognizable between the retail travel behaviour of men and women, however men have a higher rate of 1.5 compared to women.

Third, there is a large disparity between inner and outer Brisbane in terms of using sustainable transport (about 20 percent). The population density and urban pattern may appear to be the most plausible explanation for this disparity, but it may be related to the socio-economic characteristics of the people living in these areas. This makes it important for investigation as another possible factor affecting their mode share decision.

Fourth, according to the HTS data, more than 50 percent of the overall retail trips are made to supermarkets and shopping centres every week. This shows the significant role of the standalone shopping centres and big supermarkets in daily and weekly customers’ retail trips.

Fifth, the grocery and food shopping should be investigated more closely, given the importance/frequency they have in travel for retail purposes. Any modification in customers’ travel habits to these destinations could result in large differences in mode share.

Finally, a reduction in the number of these trip types (groceries, food, milk, bread, confectionary, non-alcoholic drinks, pharmaceutical) during the weekends shows that people prefer to make these trips mostly during the week on a regular basis rather at the weekends. This is an important point relating to

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 123 location and level of accessibility to neighbourhood centres if planners are looking for more sustainable alternatives in future.

Since the supermarkets and shopping centres showed a major role in overall trip activity, they have been analysed separately in more detail. Further analysis on customers’ travel behaviour for these destinations showed significant results.

For trips to supermarkets, more than 80 percent of trips are allocated to groceries, food, bread, milk, fish and meat and non-alcoholic products. This number reduces to almost 65 to 55 percent for shopping centres, since these large centres, from a super-regional down to sub-regional centres, contain at least one or even two big supermarkets. This is of great significance when looking at distance and people’s mode choice preferences to reach these destinations.

Retail trips experience considerable change in terms of means of transport regarding the distance between origin and destination. Walking and public transport could still comprise an important part of trips to shopping centres and supermarkets based on how far the trip is. The results show that an average of 3 percent of trips to shopping centres each week are currently being undertaken by walking, while almost 17 percent of trips to these centres are within a distance of less than 1 km, which is a walkable distance. While these numbers show some improvement when it comes to supermarket trips, the percentages do not go higher than 10 percent overall.

The role of public transport is becoming more important for shopping destinations as these destinations increase in distance by more than 5 km from the home base.

Many of these trips are made for types of products such as daily grocery shopping (milk or bread) or for clothes or shoes that can be easily carried on public transport. It should be also noticed that the HTS data do not provide the information for all the products bought in one trip and there might be other products being bought which make it harder to carry them all on PT. but the results aslo reveal a higher probability of the influences of people’s socio- demographic characteristics on their current travel behaviour.

Chapter 5 Analysing Customers’ Travel Behaviour in Brisbane 124

The large proportion of trips (>25 percent) that are only made to the large shopping malls, excluding the neighbourhood centres within the region, highlights the fact that one cannot solely focus on supermarkets and local shopping centres in making retail trips more sustainable. Instead, large scale plans for retail centres are essential in order to have an impact on people’s travel behaviour. The investigation into shopping malls is becoming more important when looking at the mode choice of trip makers to these centres. Apart from the CBD, all the trips to the other centres are being made by private cars (more than 80 percent), compared to only 10 percent of car trips to the CBD. Changing this behaviour will be a considerable challenge; however some malls in the region are changing their transport orientation to include public transport, especially busways. The Garden City shopping mall in Upper Mt. Gravatt is an example, being located on the southeast busway.

Almost half of the trips to shopping centres apart from the CBD centres are home-based trips. The time spent in the destinations is more in the larger centres. This time reduces to less than 1 hour when it comes to smaller centres such as the sub-regional, neighbourhood centres and also bulky goods outlets. For the CBD centres, 40 percent of the retail trips are work-based trips, and these trips follow a completely different trend and should be differentiated from trips to shopping centres for retail purposes.

The results show the great importance of research into shopping centres and their impact on the total trip distance travelled by different mode share in Brisbane.

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 127

6.1 Introduction

The analysis of the previous chapter on customers’ travel behaviour not only revealed a number of important facts on the travel behaviour of people, but also raised some significant questions regarding the influences of socio- demographic characteristics on their behaviour. Factors such as high dependence on private cars, even for close distance destinations of less than 1- kilometre, the important role of public transport in longer trips, and trip frequency and destination location for purchasing various products show another aspect of retail travel that has not been properly explored. These issues justify the need for more research into the socio-demographic characteristics of retail travellers.

People from different socio-demographic backgrounds are usually expected to behave differently towards transport opportunities, however traditionally, many transport planners ignore this fact. Stead and Marshall (2001) believe that socio-demographic characteristics are confounding factors that significantly impact on people’s travel behaviour and mode choice (Stead and Marshall, 2001). If retail travel is to become more sustainable, an accurate and comprehensive calculation of how different groups of people currently travel is required. Different methods could be applied in such a calculation, for example, predefined socio-demographic groups’ travel behaviour could be studied, illuminating the differences. Preferably, a method of inductive extraction of the dominant socio-demographic groups and an investigation of their behaviour would be more constructive.

In this study, the inductive method is applied using a cluster analysis technique, to identify major groups of retail trip makers in Brisbane based on SEQ-HTS data of 2009. Various aspects of the travel behaviour of these groups, such as retail trip rate, the mode share frequency and distance per capita by each mode , and the trip frequency to shopping destinations based on the “type of product” bought, are all investigated.

The chapter begins with a brief review of the literature on customers’ socio demographic characteristics and the impact on trip makers’ travel behaviour. As

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 128 the selected method for this analysis, the cluster analysis technique is then discussed in detail. This will be followed by a discussion on the data analysis and the results of the findings. The chapter concludes with suggestions for how this analysis might be used in the future to improve a more sustainable transport system.

6.2 Literature review

Much research has been conducted on how to improve the sustainability of travel through travel behaviour change. The focus of many of these studies has been on urban form elements: density, land use mix, street patterns, etc., in an effort to shorten the distance between origin and destination and at the same time encourage people to substitute car trips with public transport (PT), walking and cycling.

Although all these factors play a role in decreasing car trips, and increasing non-motorised modes, private cars are still the preferred mode of transport used by most people, even for short trips. In part, this can be explained by the practical benefits of private motor vehicles such as speed and convenience. However, it should be noted that there are other important reasons besides urban form issues that affect individual travel behaviour decisions.

A number of studies have attempted to include the socio-demographic factors in their investigations. These studies include factors such as age, income, gender, employment status, auto ownership and household size, employment, population density. See Cubukcu (2001) for more indepth review of the studies that investigated these factors (Bruton (1985); Koppelman and Pas (1984); Pas (1984); Boarnet and Crane (2001) Martin et al. (1961); Levinson (1976) ; HOBBS (1979)).

In his 1981 study, Hanson suggested that socio-demographic variables like age, income, gender and car availability are more important in influencing travel behaviour than spatial form or land use (Hanson and Hanson, 1981; Handy, 1996a). Pas (1984) studied the impacts of individuals’ and households’ life cycle and life style attributes on daily travel behaviour. He examined variables such as age, marital status, gender, employment status, education level, presence of

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 129 young children, car ownership, income and residential density, and concluded that the daily travel behaviour of individuals and households are totally subjective to their roles, life styles and life cycle characteristics (Pas, 1984). Carlsson-Kanyama and Linden (1999) found some differences between the travel patterns of different socio-demographic groups of Swedes, although private car use was always dominant. They claimed that the elderly, low income groups and women make fewer trips than those who are middle-aged, male and have high incomes. They also pointed out that public transport is used more by young people and women in contrast with the other groups.

The importance of personal attributes and status on mode choice and the distances travelled, along with urban form and design, were examined by Dieleman et al. (2002). Their study authenticated the fact that the probability of owning and use of a private motorcar is much higher in higher income families rather than in lower income families. Private car reliability for trips is much more common in families with children, in comparison with one-person household types. Among all groups of drivers, those who are of working age, male, have children and high incomes have the greatest tendency to drive (Ryley, 2006). Ryley’s study also determined that other studies have come to similar conclusions. A study of the impact of life-course events on car ownership and travel behaviour in Germany found that key events such as a change in the number of adults in a household and the birth of the first child can dramatically affect travel behaviour.

Other household variables such as age, number of cars owned and monthly income had a strong influence on car ownership growth (Prillwitz et al., 2006). Kattiyapornpong and Miller (2011) studied the relationship between the travel behaviour of people in Sydney using a binominal regression model to show that age, income and life-stage have significant differential and interactive effects on travel behaviour.

In the case of retail trips, a few studies have considered the connection between socio-demographic characteristics, retail structure and travel to these destinations. This is related to the fact that the factors that influence retail trips are also many of the same factors that affect the trips in general (Cubukcu,

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 130

2001). Murdie (1965) argued that people with different socio-economic and cultural backgrounds will make different travel decisions about their retail needs. He found that low income shoppers prefer to shop at local centres with infrequent visits to larger regional centres for more specialized goods. Conversely, high income consumers will travel further for both convenience and other shopping goods. Bromley and Thomas (1993) examined the travel behaviour of carless disadvantaged groups in the UK and argued that retail destinations were increasingly car-based, which was unfair to poor mobility groups while providing advantages to middle or upper income groups.

However there have been a number of studies including those by Robinson and Vickerman (1976), Vickerman and Barmby (1984) , Badoe and Steuart (1997) that focused on socio-demographic factors (i.e., income, car ownership, household size, number of licensed drivers, number of workers and number of vehicles) and found weak evidence to explain retail trip distance (Cubukcu, 2001).

While the focus of this study is on Brisbane, a few studies can be found on the socio-demographic or economic characteristics of trip makers in this area. One such study is that by Dodson et al., (2010) on transport disadvantage and social status, which focused on a Gold Coast case study. Brisbane has a range of socio-economic groups who make retail trips as a key part of their overall weekly trip-making (Dodson et al., 2010). While retail trips are mostly dominated by car based trips, it should be considered that each of these groups will have different preferences for how and when they make shopping trips. Therefore, when planners and policy makers are developing practical policies to achieve a more sustainable transportation system, it is necessary to have a clear understanding of how different socio-demographic groups travel to retail destinations, and whether any major differences in terms of mode share decisions can be found in the region.

6.3 Methods

To study the travel behaviour of different socio-economic groups in Brisbane, the same 7-day 2009 SEQ-HTS data is applied. As mentioned in the previous

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 131 chapter, the total number of retail trips reported in the dataset is 3,354 trips. Around 800 of these trips were made by the same people, meaning that some of the respondents made more than one shopping trip per week. Factoring this into the analysis reduces the sample to 2,525 individual socio-demographic cases. Study of retail travel behavioural inferences in such a large dataset is not practical unless the number of cases can be reduced to smaller sets. Based on the literature review and the current environment of the city, a number of factors are chosen to divide the dataset into manageable categories, based on the characteristics of individual retail trip makers. These factors include: age, gender, whether they hold a driving licence, main daily activity, household size, household structure and household income. Car ownership has not been considered as a separate factor, as only 4 percent of the retail trips were made by people from households that did not own a car.

Different methods such as factor analysis and discriminant analysis can be used for recognising and defining socio-demographic groups in a dataset. In discriminant analysis the number and the attributes of each cluster are known and the main task is to assign these cases in to the appropriate groups (Mooi and Sarstedt, 2011a). While these methods use deductive approaches to recognize existing taxonomies in a dataset, this research uses the inductive classification method. Cluster analysis assumes that no attributes have been specified. It does not apply any previous assumption on how to sort the objects into a predefined, specific number of groups (Norusis, 2008). This method divides the existing data into meaningful groups by looking at hidden (and useful) patterns inside data (Shih et al., 2010). It is a way of highlighting the relationships and the interactions that exist between observations in a dataset, which are not easily identifiable.

In this study, SPSS software was applied to implement the cluster analysis technique. Cluster analysis relies on measuring the distance between objects. When the distance between two objects gets larger, it means that they are becoming less similar. When it gets smaller, they are becoming more similar and can be combined to form a group. Observations are put together in one cluster based on their ratio of similarity in one group and the ratio of dissimilarity

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 132 between other groups. There are three methods of cluster analysis in SPSS based on this concept, with each relying on a different clustering algorithm (Mooi and Sarstedt, 2011b).

The first of these methods is hierarchical analysis, a practical method for clustering small amounts of data (Şchiopu, 2010). This methods creates a similarity matrix between all pairs of cases and for all but small datasets (less than 500 cases) would result in a huge and confusing matrix (Norusis, 2008). The second method is K-mean clustering which is restricted to continuous variables (Şchiopu, 2010). This method is best applied to interval or ratio scaled data, since the clustering calculation is based on Euclidean distances. Applying this method to categorical variables is not recommended and will result in important misinterpretation in the outcomes (Mooi and Sarstedt, 2011b). Finally, the third method is known as two-step clustering, and this is the most appropriate method for large datasets containing continuous and categorical variables. It uses an agglomerative hierarchical clustering method (Şchiopu, 2010), which starts with every case as an individual cluster, then merges these clusters based on their similarities (Mooi and Sarstedt, 2011b). In two-step clustering, there are two methods of measuring the distance “Euclidian’ and ‘log-likelihood’. The Euclidian distance is used for numerical (continuous) variables, and if or when a categorical variable is encountered, the method switches to log-likelihood distance (Şchiopu, 2010).

Many studies have been done using cluster analysis to identify groups in a dataset; for example identifying groups of customers or viewers of a particular TV show for the purposes of developing advertising, or in the case of deriving employers’ branding strategies (Mooi and Sarstedt, 2011b). In the case of travel behaviour, two relevant studies have been identified. The first research by Ryley (Ryley, 2006) examined the link between the life stage and travel behaviour of people in Edinburgh, Scotland. People were grouped into six socio-economic categorical variables: number of adults in the household; household income; dwelling type; individual life stage; gender; and number of children in the household. He applied the hierarchical cluster as the main method and tried four hierarchical techniques to identify groups. In the second

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 133 study Dodson et al. (Dodson et al., 2010) applied Ryley’s approach (hierarchical) to eight socio-demographic variables to identify disadvantaged groups in the Gold Coast, Australia. Six disadvantaged groups were identified and the travel behaviour of each was examined.

The two-step cluster method applied in this research was chosen, since it does not include the limitations of dataset size and is applicable to both categorical and continuous variables. Among the seven socio-economic characteristics, age and household income comprised the continuous variables and the five others, including gender, having a driving licence, main daily activity, household size and household structure formed the categorical variables. The following process was used to identify the socio-demographic groups of retail trip makers in the sample.

1- Perform the two-step clustering technique to identify the preliminary clusters: seven variables have been chosen to differentiate between retail trip makers: age; sex; household structure; household size; household income; licensed to drive; and main activity. The two-step method allows the user to choose how to determine the number of clusters - either manually or automatically. For this research, both methods were used in order to achieve the most comprehensive set of clusters. Comparing the various numbers of clusters to classify the sample, 20 clusters was chosen as the most efficient and practical number.

2- Merge the clusters with similar attributes to reduce the number of clusters: although the clustering technique tries to maximize the differences between the clusters, some shared very similar characteristics, while other characteristics did not play an important role in differentiating between the clusters. Examples were the slight differences between the number of persons per household (3, 4 or 5, which shows the number of children per household), or in some cases the negligible disparities in the household income level. These clusters can come together and make more sensible and bigger groups for further analysis. This step reduced the number of clusters to 10 groups.

3- Refine the clusters: when cases are allocated to a cluster it does not mean that they all have exactly the same attributes as each other. The two-step

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 134 cluster method gives us the chance to determine the exact number of cases sharing a specific attribute. For example, from the total number of cases that has been allocated to one cluster, if 20 percent belong to a couple with children, 45 percent to couples with no children and 35 percent are sole persons, the two-step method will name the group after the largest category, which in this case is couples with no children. Sole persons have to be considered as a major part of members in this cluster as well. Therefore, there is a need to review each cluster to see if the number of cases should include more than one group of characteristics. If the number of cases is close in different categories of one attribute, then they should all be considered as major attributes of that cluster.

4- Allocation of cases into appropriate groups: after all groups have been identified, the cases sharing the same attributes are put in the same group. The final number of groups made in this step is eleven, which include four main groups of people: working age, retired, housekeepers and students. The ten major groups are shown in Table 6-1.

The most common socio-demographic groups recognised by the cluster analysis include the following characteristics: age (=<18, 18< & >=64, 64< ); sex (male, female); household structure (couple with children, couple with no children, single parent, sole person) ; household size ; household income per week (less than 1000 AUD, between 1000 to 2000 AUD, between 1000 to 4000 AUD, more than 4000 AUD); licensed to drive (full license, no license); and main activity (full time/part time/casual work, pension, keeping house, primary or secondary student, university student).

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 135

Table 6-1: The most common socio-demographic groups in the 2009 SEQ-HTS

Working G1- female, full licence, having full time/part time or casual work, couple with children, 3 to 5 persons, 1000=

G2- male, full licence, having full time/part time or casual work, couple with children, 3 to 5 persons, 1000=

G3- female & male, full licence, having full time/part time or casual work, sole person, 1 person, income=<2000, 18

G4- female & male, full licence, having full time/part time or casual work, Couple no children, 2 persons, 1000=

Retired G5- female & male, full licence, pension, couple no children, 2 persons, income=<2000, 64 =

G6- female & male, full licence, pension, sole person, 1 persons, income=<2000, 50 =

Housekeepers G7- female, full licence, keeping house, couple with children, 3=

G8- female, full licence, keeping house/unemployed, couple no children, 2 persons, income=<2000, 64 =

Students G9- female & male, no licence, primary or secondary students, couple with children/one parent, 2=

G10- female & male, full licence, university students, 1000=

Amongst all the different groups, only school students do not have a full driving -licence.

6.4 Results and Discussion

Ten groups with different socio-demographic characteristics were identified from the two-step cluster analysis method. Out of the total number of 2,525 trips that have been considered in the study, 2,166 have been categorized into these 10 groups. These ten clusters can be further grouped into four major taxonomies: working group; education; retired and at home. Retail travel characteristics of each of these four groups were analysed: trip rate; mode share; kilometres travelled per capita by mode; and items purchased. The study’s working hypothesis was that each of these groups would have different travel

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 136 characteristics. If this hypothesis is supported it means that to alter people’s retail travel behaviour to a more sustainable pattern, a number ofpolicies may be required for different socio-economic groups.

Figure 6-1 shows the number of retail trips per person per week (retail trip ratio) and total trips per person per week (all trips ratio) taken by each group during one 7-day period. The retail trip rate ranges from a low of 1.11 for school students (G9) to high of 1.46 for working groups with medium to high incomes (G4). For all trips, females in the working group and the keeping house group with children (G1 & G7) make more than three trips per week. University students (G10) make the next highest group in terms of the total number of trips per week, while students (G9) make less than two. The figure shows that between all groups, for every person, more than half of the trips made per week are allocated to retail trips, except for male workers with children (G1), female housekeepers with children (G7) and university students (G10).

3.5

3.0

2.5

2.0

1.5 No.of Trips 1.0

0.5

0.0 G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 Socio-Demographic Groups (See Table 6-1 for definitions)

Retail Trips/Person/Week Total Trips/Person/Week

Fig 6-1: Trip Ratio for Retail and All-types of trips

Table 6-2 provides the number of retail trips for each group, the trip rate, the number of persons who have made these trips, the total number of trips and the number of persons making these trips. The comparison between the percent of retail trips to all trips (the far right hand column of Table 6-2) shows that the retired singles and couples (G5, G6) and married females with no children who are either unemployed or keeping house (G8) are the groups making the

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 137 highest percent of retail trips (ranging between 33 to 39 percent). In all the other groups with the exception of school students (G9) and married working males with children (G2), the retail trips to total trips percentage ranges between 17 to almost 23 percent. In the case of school students (G9) the percentage is less than 5, and for the married working male with children group it is slightly more than 14. According to this table, the only group who makes more retail trips on weekends rather than during the week are couples of working age with no children (G4).

Table 6-2: The number and ratio of retail-trips to all-trips made in one week

No. of RTs/ No. of TTs/ % of RTs1 RT TTs2 TT RTs to makers person/week makers person/week TTs Group1 431 333 1.29 2481 814 3.05 17.37 WD3 280 1962

WE4 151 519

Group2 355 275 1.29 2502 996 2.51 14.19 WD 173 1868

WE 182 634

Group3 140 106 1.32 756 329 2.3 18.52 WD 74 566

WE 66 190

Group4 460 314 1.46 2489 989 2.52 18.48 WD 216 1782

WE 244 707

Group5 282 196 1.44 870 369 2.36 32.41 WD 199 622

WE 83 248

Group6 121 98 1.23 367 172 2.13 32.97 WD 78 257

WE 43 110

Group7 126 94 1.34 553 180 3.07 22.78 WD 84 437

WE 42 116

Group8 63 47 1.34 161 79 2.04 39.13 WD 45 114

WE 18 47

Group9 113 102 1.11 2320 1183 1.96 4.87 WD 72 1872

WE 41 448

Group10 75 60 1.25 438 164 2.67 17.12 WD 52 328

WE 23 110

Total 2166 12937 1RTs Retail Trips / 2TTs Total Trips / 3WeekDay / 4WeekEnd /

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 138

6.4.1 Mode Share

For all ten groups, at least about 75 percent of trips were made by private motor vehicles (Fig 6-2). Working couples with children on medium or high salaries (G1 & G2) and females who are keeping house or are unemployed (G7 & G8) are the groups having the highest reliance on private motor vehicles – approximately 90 percent. School students and working couples with no children with medium to high incomes (G9 & G4) were the two groups with the lowest reliance on private motorised vehicle trips at 77.0 and 75.6 percent, respectively. As expected, these two groups had the highest level of non- motorised and public transport trips. For working couples with no children, non- motorised retail trips were 17.2 percent and public transport trips were 5.3 percent. Comparable values for school students were 15.0 and 7.1.

100%

95%

90%

85%

80%

sharePercentage -

75% Mode

70% G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 Socio-Demographic Groups (See Table 6-1 for definitions) Private Motor Vehicle Non Motorised Public Transport Other

Fig 6-2: Retail Trip Frequency and Mode-share

Comparing these couples without children (G3, G4) with their peers who have children and the same level of income (G1 & G2), the percentage of non- motorised trips and the trips with public transport is almost twice as high.

The table shows that public transport is mostly not applied by households with children, elderly people and females whose main activities are keeping house or who are unemployed (G1, G2, G6, G7, G8). For non-motorised trips, after

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 139 students (G9) and couples with no children, people who are living alone even in working age or retired (G3, G6) are making the most non-motorised trips.

6.4.2 Distance (km) travelled per capita by mode share

The only mode shares that have significant distances travelled are trips made by private motor vehicle and public transport. Male workers with children and a medium or high income (G2) travel 10.5 km by car to retail destinations per week, the highest of all the groups. The next highest is by working age couples with no children (G4) at 9.9 km per capita, followed by one person, working age and a low to medium income (G3) at 8.7 km. As was discussed above, these two groups (G3 & G4) are the ones making the fewest car trips relative to non- motorised and public transport trips. Thus, it can be concluded that when these groups do use a car to go shopping, they make longer trips.

7 Distance (km) 6

4 G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 Socio-Demographic Groups (See Table 6-1 for definitions) Private Motor Vehicle Non Motorised Public Transport Other

Fig 6-3: Distances Travelled for retail per capita by Mode-share

University students (G10) and females keeping house or unemployed with no children (G8) make around 5 km of retail trips by car each week. This is the lowest number amongst all the other groups. Females keeping house with children (G7) are the only group that did not make any trips to retail destinations by public transport.

Among all groups, low or medium income workers with no children (G3) are travelling longer distances by public transport at 1.6 km per capita. This number

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 140 drops to less than one km per capita for other groups. The analysis also shows that for those groups making the highest percentage of public transport retail trips (G4 and G9) (fig 6-2), most were for shorter distances than G3 which shows the highest number of kilometres.

6.4.3 Retail trip frequency by type of product

For retail trips, the type of goods purchased affects travel mode decisions and destinations. For example, when buying food, one travels to a supermarket, when buying household goods and furnishings one goes to a big box retail centre, and to buy clothes and shoes, one goes to a regional shopping centre. As Table 6-3 shows, the percentage of trips for groceries and food is similar across most groups; however, females still have an important role in these trips (G1, G7 and G8).

School students and university students made the highest percentage of trips for non-alcoholic drinks (about 6 percent), while they and females with children who are keeping house (G7) make fewer trips to purchase alcohol. Trips for purchasing tools and hardware were generally lower than most other trip types except for males (G2), sole people (G3) or couples with no children (G4, G8). For almost all the groups, trips to purchase personal goods or household goods were the second highest destinations after groceries and food. As would be expected, trips for clothes and shoes are more common for couples with children and students (G1, G7, G9 and G10).

Purchasing household furniture plays a small part in trips per week in all groups, at less than 2.5 percent. Couples of working age with no children (G4) are ranked as the first group with the highest percentage of retail trips at 3.2 percent, followed by retired couples (G5, 2.5), females of working age with children (G1, 2.4) and female housekeepers with children (G7, 2.3).

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 141

Table 6-3: Percentage of retail trips for purchasing diverse products

Groceries Non- Alcohol Clothes & Personal Household Household Tools & Other alcoholic goods furniture hardware & food shoes goods

G1 53.4 1.1 1.5 7.4 4.2 10.1 2.4 1.1 18.8

G2 39.7 1.9 3.2 3.5 11.1 8.6 1.9 6.8 23.2

G3 48.3 1.4 2.7 2.0 14.3 6.1 0.0 4.1 21.1

G4 43.8 1.2 2.0 7.3 7.5 10.5 3.2 6.0 18.5

G5 48.4 1.1 1.1 3.5 14.4 5.6 2.5 3.9 19.6

G6 50.0 0.0 0.8 3.2 16.1 4.0 0.8 0.8 24.2

G7 53.4 0.8 0.0 6.9 7.6 7.6 2.3 3.8 17.6

G8 54.5 1.5 3.0 3.0 9.1 7.6 0.0 4.5 16.7

G9 45.9 6.0 0.8 8.3 3.0 15.0 0.8 0.0 20.3

G10 36.9 4.8 0.0 8.3 10.7 13.1 1.2 1.2 23.8

6.5 Conclusion

The analysis of these socio-demographic characteristics reveals very important hidden aspects of travel behaviour. The results not only show the prevailing groups of retail trip makers in the city, but reveals how they behave in terms of their mode choices, the distances that they travel and also the type of places and products they choose.

More than half of the weekly trips are made for retail purposes, with the exception of those trips made by male workers, female housekeepers with children and university students. This is another factor showing the importance of retail trip type in the city. Even though the highest number of retail trips per week is mostly allocated to people of working age, the highest ratio of retail trips to all trips that a person makes per week is for single people and retired couples (G5, G6) and female housekeepers with no children (G8). A larger part of these people’s time is spent on retail trips, probably due to the fact that they have more free time.

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 142

Having workers as the group with the highest percentage of shopping trips supports the idea of transit oriented development and strengthens the spatial connection of retail and public transport for the workers to buy their essentials on the way home from work.

While it is believed that the weekends are the best time for shopping, the only group showing a higher percentage of trips for shopping on the weekends compared to the weekdays is couples with no children.

While private cars still comprise more than 75 percent of all retail trips for each group, working couples with children with medium to high incomes (G1, G2) and housekeepers or unemployed females (G6, G7) make the most trips by car; the next highest groups are solo or retired couples (G5, G5). Having children seems to be an important reason for the high usage of cars for retail trips. On the other hand, working groups with no children and students are shown to be less reliant on cars and more interested in use of other modes (non-motorised or PT) of transport compared to the other groups.

Working groups (G1 to G4) and retired couples (G5) make the longest retail trips based on the number of kilometres travelled by private car. In terms of non-motorised mode-shares, distances are not large - limited to 40 to 130 metres per capita per week, with the majority of these trips being made by lower income families (retired or workers) and students.

The analysis also shows little difference between lower income and the medium/higher income groups in terms of car usage. While the literature suggests that low income, disadvantaged or car-less are either unwilling or unable to drive or do not have access to a car (Giuliano, 2005), that was not the case in this analysis. Based on the 2009 SEQ-HTS, even low income families are making most of their retail trips by car. Of the 645 retail trips made by low income households, just 87 trips were made by people without a car. This may be related to two factors: the high price of PT; and/or the limited number of destinations served by PT.

The major difference for lower income persons in comparison to the other groups is in the distance travelled. In terms of trips by car, lower income families travel to nearby destinations while higher income families travel greater

Chapter 6 Socio Demographic Groupings and Revealed Retail Travel Behaviour 143 distances. For public transport trips, the opposite seems to be the case. Longer retail trips by PT are made by low income families and single workers. But the frequency of using PT for retail trips is higher for employed, single people or couples with no children and also by students (school or university), while retirees and housekeepers rarely use PT.

The analysis suggests that working couples with no children, working or retired persons living alone, and school students groups are the most sustainable retail travellers. On the other hand, females staying at home, retired couples and working couples with children and high/medium incomes represent the least sustainable retail travellers. These groups should be the target of efforts to change travel behaviour.

Finally, of all retail trips, those made to purchase groceries and food represent the largest percentage (more than 35 percent) between all the socio- demographic groups, followed by household and personal goods. This reflects another important reason to seriously consider these destinations in future land application planning policies to make the city’s retail structure more sustainable.

It should be noted that this study is the first research into retail structure and socio-demographic characteristics of customers in Brisbane. It reveals the major groups of socio-demographic customers and their retail travel patterns. However, more research needs to be done to understand how and why retail travel patterns vary among the socio-demographic groups identified in this analysis.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 147

7.1 Introduction

As discussed in the literature, a large number of factors including individual-specific, trip-specific, destination-specific or spatial-specific attributes might affect shoppers’ destination choice and travel behaviour. Chapter 5 and Chapter 6 analysed HTS data to clarify of various attributes of shopping trips and how differences in socio- demographic characteristics of shoppers affected their trip mode choices and distances to travel.

Results revealed that shopping centres are the destination for more than 40 percent of retail trips every week. The literature review revealed that shopping centres have increased in popularity since the 1960s, gradually replacing other types of retail destination by providing the one-stop, multi-purpose, easily accessible shopping opportunities together with recreational and social amenities in the same location. Consequently, shopping centre destinations are used as the focus destinations for the remainder of this research.

This chapter focuses on destination-specific attributes and the spatial characteristics of shopping centres as retail destinations. It investigates and measures the spatial characteristics of the settings around retail centres, tries to identify and understand possible similarities and relationships between these locations which might affect peoples’ behaviour and required to be estimated later in the study. Spatial attributes of shopping centres such as:  the residential population density within a centre’s catchment area  the land-use mix in a shopping centre’s locality  the retail employment density around a centre  the accessibility of a centres by different transport mode types have been mentioned elsewhere in the literature. These attributes will be analysed in this Chapter, subject to the limitations imposed by the availability and accuracy of the available datasets. The data used in these analyses are:  Demographic data (employment and retail employment density) extracted from the BSTM model for 2011

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 148

 Detailed information on the centres’ characteristics and their exact locations obtained from the shopping centre directory (SCD) dataset for Queensland (2011)  Land use data (2011) from the Queensland government  Information about the roads and transit network, cycling and walking pathways from the LUPTAI model, is also used to estimate the accessibility of shopping centres using different transport modes.

The chapter begins with a brief review of relevant spatial attributes and the ways in which they could influence shoppers’ retail travel behaviour. A methodology section follows, explaining the ArcGIS spatial analyses which will be used to quantify the different attributes for shopping centres in the Brisbane Statistical District. The results will then be discussed in detail.

7.2 Literature Review

Only a small portion of the numerous studies on urban form and travel behaviour interactions focus directly on shopping trips. These studies can be categorised as focusing on either regional or neighbourhood scales. Neighbourhood retail was considered an important component of urban structure in much of the early research in the urban planning field (Rapoport, 1987; Owens, 1993b; Whyte, 2012). Many studies were undertaken to investigate whether neighbourhood retail developments improved the quality of life in the residential areas by decreasing reliance on automobile trips and encouraging residents to engage in active transport (walking and cycling (Frank and Pivo, 1994b; Friedman et al., 1994; Maat et al., 2005; Krizek and Johnson, 2006)

Many of these studies focused on comparing the travel behaviour between traditional and modern pattern neighbourhoods. The influence of different aspects of urban form including mixed land-use, street pattern, accessibility of retail destinations by different transport mode, travel distance, travel time, travel cost, number of available options at the destination, retail outlet size etc. were investigated in these studies.

Handy studied the relationship between particular characteristics of urban form and the pattern of shopping trips (Handy, 1996b). She showed that a number of

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 149 arrangements at local level and neighbourhood scale could effectively encourage active transport. Examples included a greater range of destinations choices, reduced travel distances and a reduction in the number of major barriers (e.g. arterial roads). Handy argued that these arrangements were more effective for new developments rather than infill developments, and were more influential over the adoption of active transport behaviours than socio-demographic characteristics of resident households.

Krizek & Johnson (2006) showed that the travel distance to shops is a statistically significant predictor of people’s active transport adoption, but the relationship did not appear to be linear (Krizek and Johnson, 2006).

A mixed use core (MUC) in a neighbourhood has been identified as increasing the number of non-work active transport trips (Limanond et al., 2005). Land-use impacts on shopping trip frequency have been found to strongly affect 30-40% of home- based shopping trips (Agyemang-Duah et al., 1995; Lee and Goulias, 1997). Friedman et al. (1992) found that land-use mix and street pattern increase the average car trip rate by almost 60 percent for all different types of trips in modern suburban areas compared to traditional neighbourhoods in the San Francisco Bay area, followed by 30 percent shifts in the number of car trips for the home-based non-work trips (Friedman et al., 1994) .

Koenig (1980), Hanson and Schwab (1987) and Robinson and Vickerman (1976) all identified accessibility as a key influence over transport mode choice and trip frequency for non-work trips, finding strong relationships between trip frequencies and the accessibility of retail locations. Generally, they all believe that socio- demographic factors are often more influential on shoppers’ non-work trip decisions than spatial factors. Recker & Kostyniuk (1978) looked at urban grocery shopping trips to identify the influence of individual perceptions of the destination, the accessibility of the destination by different transport modes, and the relative number of shopping opportunities at the destination, on shoppers’ destination choice and travel behaviour. Recker & Kostyniuk’s study concluded that “accessibility is the primary aspect influencing destination choice and its effect is nonlinear” (Recker and Kostyniuk, 1978). (Ghosh and McLafferty, 1984) multipurpose shopping trip frequency model identified that "the rate of multipurpose shopping depends on the consumer's location (transport cost) in relation to shopping opportunities".

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 150

Subsequently, Hanson & Schwab (1987) point out that Ghosh and McLafferty's model hinged on the assumption that higher accessibility and lower cost will result in a higher number of trips amongst similar socio-economic groups (Hanson and Schwab, 1987).

Hanson and Schwab (1987) argued that accessibility to retail destinations can affect trip frequency, destination choice and trip complexity (Hanson and Schwab, 1987), whilst Iacono et al. suggested that improved retail accessibility could improve quality of life (Iacono, et al., 2010). Maat (2000) found that the retail trip mode share for cars in Houten, the Netherlands, was significantly lower (8 to 13 percent) than that of comparable Dutch towns, primarily due to improved pedestrian and cycling accessibility compared with car accessibility in the town’s transport network. Maat also mentioned that people using private cars make longer trips for their shopping compared to users of other transport modes cited in (Stead and Marshall, 2001). However, other research refutes the strength of these relationships. For example, Frank and Pivo found no significant connection between shopping trips and land- use mix or density as elements of urban form whilst these factors were still found to influence the number of work trips (Frank and Pivo, 1994b) (Crane and Crepeau, 1998). Holtzclaw (1994) claimed that changing the quantity of neighbourhood shopping did not significantly affect private car usage or the transportation costs of communities, whereas other factors such as transit accessibility and increased density could reduce the Vehicle Kilometre Travelled (VKT) per household (Holtzclaw, 1994). Crane and Crepeau’s study in San Diego showed only minor effects of land-use on the travel behavior, and found that the street network pattern did not influence non-work trip makers’ mode decision (Crane and Crepeau, 1998). Limanond & Niemeir (2004) developed an activity-based shopping model to analyse the impacts of land-use pattern on three different aspects of a shopping tour: shopping tour frequency, tour scheduling and mode choice. While land-use pattern appeared to exert no influence over overall shopping tour frequency, they found that land-use levels exerted considerable influence over the type of shopping tour including the one-stop or multi-stop shopping tours. Limanond & Niemeir’s (2004) results suggested that lower accessibility levels to various land-uses will result in less one-stop shopping, while high accessibility levels could increase the number of

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 151 one-stop shopping tours (Limanond and Niemeier, 2004). They finally concluded that the neighbourhood residents living closer to a mixed-use area and with a higher level of accessibility are more likely to make non-auto one-stop shopping tours.

A study of retail at a broader regional scale by Handy (1992) found that while higher accessibility of both local and regional shops can affect the travel distances of trip makers, it does not have any impact on total travel. Handy therefore claimed that better access to retail will not encourage people to make more trips, but it could improve the quality of residents’ lives (Handy, 1992),.

While the literature on the impacts of a retail environment on shoppers’ travel behaviour shows somewhat mixed findings, these issues are likely to be context specific. Overall, some spatial factors such as retail density and/or the level of mixed-use in the vicinity of a shopping centre, and a shopping centre’s accessibility by different transport modes have generally been found to affect the way people travel to retail destinations and how the choose their destinations.

Drawing on these findings from the literature, this chapter will quantify relevant spatial characteristics of shopping centres and their surroundings in Brisbane, using existing methods implemented using ArcGIS software. Potential relationships between these spatial characteristics and shoppers’ choice of retail destinations will be studied later in Chapter 8. Hence, the current chapter, focuses only on basic measurement and description of relevant spatial characteristics.

7.3 Methods

A variety of different analytical techniques can be implemented in ArcGIS software to measure spatial characteristics of shopping centres and retail settings. These characteristics include the retail density, population density and the level of mixed use in the vicinity of each shopping centre, together with a quantification of the accessibility of each centre via different transport modes. Accessibility in this context will be measured in terms of travel time; shorter travel time denoting increased accessibility.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 152

7.3.1 Measuring retail density and population density around shopping centres

Retail density and population density were calculated within a two-kilometre radius buffer around each shopping centre. The number of retail jobs in each traffic zone, extracted from the the BSTM (model developed for 2011), was used as a proxy for the number of jobs in each shopping centre. Population density was obtained from the ABS (2011 data).

7.3.2 Measuring the Land-Use Mix

Mixed use has often been mentioned in the literature as a factor which influences travel behaviour, especially at neighbourhood scale. Mixed-use development is defined by the Adelide City Council7 as:

Development which comprises a mixture of two or more land uses, either comprised within a single building (horizontally or vertically) or multiple buildings of different uses within a distinct development site.

It is generally believed that as the level of mixed-use increases, the number of people who prefer to walk to access retail destinations also increases. This is likely to be because the land use development pattern, specifically the land use mix, has been shown to have a significant effect on trip length, non-motorized and transit mode choice, and public transport choices (Tracy et al., 2011). A higher level of mixed-use increases the likelihood that the origin and destination of trips are closer together, making it more feasible for trip makers to switch from their cars to other modes of transport.

Different indices can be used to measure the level of land use mix within an area. These include the Entropy Index, the Dissimilarity Index (DI) and the Mixed type Index. Song et al. in their study of methods for measuring urban land-use mix identified two categories of mathematical approach for measuring land use mix: integral measures and divisional measures (Song et al., 2013). Integral measures

(7) From the Adelide City Council report “The guide to mixed use development” available at: http://www.adelaidecitycouncil.com/assets/documents/ACC-DIGS-mixed-use-development-guide.pdf, last accessed March 2015

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 153 are only sensitive to the overall distribution of land use types within the area being analysed, but are not sensitive to the arrangement or pattern of land use within that area. The Entropy index is an example of an integral measure of land use mix. Divisional measures, however, are sensitive to both the overall distribution of land use types and the internal arrangement of land use types within the area being analysed. The dissimilarity index is an example of a divisional measure of land use mix.

Cervero and Kockelman (1997) used the entropy and dissimilarity indices to measure the level of land use mix, and found that both indices were significant factors in explaining vehicle-miles travelled (VMT) and the number of non-motorized work trips (Cervero and Kockelman, 1997b; Bordoloi et al., 2013).

The Entropy Index will be used in this research to quantify the land use mix around shopping centres. This measure is well accepted and widely used in the literature (Frank and Pivo, 1994b; Cervero and Kockelman, 1997b; Mitchell Hess et al., 2001; Tsai et al., 2012; Bordoloi et al., 2013; Song et al., 2013) . Entropy index has been defined as “a measure of land use mix which takes into account the relative percentage of two or more land use types within an area” (Song et al., 2013). As the level of mixed use in an area increases, the entropy score rises. Mathematically:

푘

Entropy = − ∑ 푝푗 ∗ 퐿푛(푝푗)/퐿푛(푘) 푗=1

Where,

푘 ≥ 2 reports the number of land uses th 푝푗= the proportion of total land area of j land-use category in the area being analysed 푘 = total number of land uses in the study area The Entropy index is normalized using the natural logarithm of the number of land uses in the area. Consequently, the Entropy index can range between 0 and 1, where 0 indicates completely homogenous land use and 1 indicates the area being analysed is divided equally among all those land use types which are present (Bordoloi et al., 2013) (Song et al., 2013).

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 154

Fig 7-1: Measuring land-use mix in the vicinity of shopping centres

The Queensland Government land use data 2011 records 10 different categories of land use: Agricultural, Commercial, Education, Hospital/Medical, Industrial, Parkland, Residential, Transport, Water and Other. For this study, these land uses have been regrouped into six categories which are more relevant for the purposes of this study: Commercial, Education, Hospital/Medical, Parkland, Residential and Other.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 155

The Entropy index is calculated as a measure of land use mix for a one-kilometre radius buffer around each centre, as illustrated in Fig 7.1.

7.3.3 Accessibility of shopping centres via different modes of transport

“Accessibility refers to people’s ability to reach goods, services and activities, which is the ultimate goal of most transport activity” (Litman, 2015). Accessibility can be influenced by various factors such as mobility (physical movement), the quality and affordability of transport options, transport system connectivity, mobility substitutes, and land use patterns (Litman, 2015). Thus, once a city’s transport network has been established, the spatial distribution of different land use types across that city is likely to have a major influence on the accessibility of different types of destination via different modes of transport.

Different methods have been used in the literature to measure accessibility level, depending on study objectives. Geurs and Van Wee (2004) categorized the accessibility measures into four basic perspectives including infrastructure-based measures, location-based measures, person-based measures and utility-based measures. While the focus of infrastructure-based method is on performance of the transport infrastructure in terms of congestion and level of service, the location- based measures consider the accessibility to a set of spatially distributed activities. Person-based measures look into individual accessibility level for specific activity in which the individual take part and finally utility-based measures, investigate the utility achieved by accessing different types of activities. The infrastructure-based and location-based measures are mostly applied for transport planning and geographical studies (Geurs and Van Wee, 2004).

Accessibility aims to represent the overall ability of trip makers to reach their desired destinations (including services and activities). The time and money that they have to spend in reaching those destinations are therefore crucial in determining accessibility. Much previous research has evaluated transport system performance based only on factors which influence private vehicles such as roadway level-of- service, traffic speeds and vehicle operating costs; and aspects of accessibility have mostly been overlooked. These approaches have led to a focus on the issue of mobility within the city, rather than a focus on improving accessibility levels as a

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 156 general goal of sustainability studies.

New planning paradigms require a more comprehensive way of analysing accessibility via different transport modes, rather than focusing solely on ease of movement for private cars. Transportation and land use planners have therefore developed new tools which incorporate travel distances, travel time and travel cost incurred by various modes of transport (walking, cycling and PT) to quantify the accessibility of particular services and activities within the city.

Litman differentiated between three perspectives for measuring transportation - vehicle travel, mobility and accessibility - and claimed that accessibility was the only perspective capable of considering all transport modes and transport system features which are relevant for accessing particular types of destination (Litman, 2015).

As the literature considers accessibility to be an important aspect of travel behaviour, it is important for this study to quantify the level of accessibility of Brisbane’s shopping centres by different transport modes.

As discussed previously, similar types of centres located around the city usually provide similar types of products and services. It was therefore decided that it would be sufficient to measure the accessibility of a specific type of shopping centre (e.g. accessibility of small-sized centres, accessibility of medium-sized centres, accessibility of large-sized centres) from a particular location by a particular mode of transport, rather than calculating the accessibility of all individual shopping centres from that starting location.

The LUPTAI model developed by Department of Transport and Main Roads for SEQ region, has been used to measure the accessibility from different locations in the BSD to specific type of shopping centre (small, medium or large). The LUPTAI is a time-based measure of accessibility which reports the minimum travel time (in minutes) that it will take to get from a particular location in the city (specified as a transport nodes) to a specific location by a particular transport mode in the existing transport network. Further details of this accessibility calculation are provided in the next section.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 157

7.3.3.1 Land Use and Public Transport Accessibility Index (LUPTAI) 8

The LUPTAI model was developed by the Queensland Department of Transport and Main Roads (TMR) to quantify the accessibility of particular locations around SEQ LGAs by measuring the travel time (in minutes) from a specific origin to that location. The model was developed to illustrate how the level of accessibility and ease of movement by different modes of transport varied across a study area or the entire region. The transport modes considered were: walking, cycling, public transport and private cars. The model has been calibrated using Household Travel Survey (HTS) datasets and is updated every year. It uses a random utility framework to model how easy it is to get to different destination types by walking, cycling, public transport and private vehicle modes and combinations of modes.

LUPTAI accessibility scores are reported for trips to a number of major types of destinations: Education, Employment, Finance, Food, Health, Recreation, Retail and Services, or for other specific locations defined for particular research purposes. LUPTAI accessibility results can be compared and evaluated to determine the accessibility level of each type of location via different transport modes from different origins across the city.

The LUPTAI model can be used to produce either an accessibility analysis or a travel time analysis. LUPTAI’s travel time analysis uses the raw travel time while the accessibility analysis uses a cost-weighted travel time which includes both monetary and non-monetary costs of travel. The accessibility model considers the weighted travel time based on the generalised cost. The generalised cost measure tries to reflect all of the factors that go into people’s perception of the difficulty of travel including the generalised cost for walking to public transport station and waiting time in the station. The value of the time spent walking to the station and waiting for the train, the fare paid, the value of the time spent travelling and the value of the time walking from the station to their destination, will all be considered in the accessibility model. All these components are converted into a common utility unit such as

8 The information provided in this section is based on Consulting, P. D. (2008) Land Use And Public Transportation Accessibility Index (LUPTAI) model improvements, Queensland Transport Integrated Transport Planning Division, Transport and Main Roads (2014) Land Use and Public Transport Accessibility Index (LUPTAI).

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 158 dollars or minutes which are also convertible from one to another by multiplying by a constant factor.

This study uses the LUPTAI’s accessibility results rather than the raw travel time results since it can give us a more realistic view on the overall retail travel time (in minutes).

A large amount of input information is required for the LUPTAI model. This includes details of the road network, the pedestrian network, the bicycle network, the public transport network, car travel times, land use types, number of population, number of employment jobs, and household travel survey data. The Public Transport network is described using information on all weekday services from the public transport operator, Translink.

In the LUPTAI model, each node in the relevant network is regarded as an origin, and each particular destination point is a destination. The model then calculates an accessibility score from each node on the network to the selected destination by finding an average of the shortest path to the destination from a number of iterations.

The LUPTAI accessibility score (in minutes) gives an indication of the average time it would take to arrive at a chosen destination taking into account peoples’ preferences for transport mode, the PT services and frequencies available and modal interchanges.

7.3.3.1 Models assumptions

The LUPTAI accessibility model contains a number of assumptions which could affect interpretation of the results.  Accessibility is shown as the cost-weighted travel time to reach the closest destination  Public transport travel times are based on Translink timetables, interchange and waiting times  Car travel times are extracted from the South East Queensland Strategic Transport Model (SEQSTM) for 2013 as a base year and include terminal and congestion penalties  The active transport network (used for calculating accessibility by bicycle and walking) is based on the State Digital Road Network (SDRN)

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 159

 Data for population is obtained from the Estimated Residential Population 2013 from ABS  The walk speed modelled is 4 km/hr and average cycle speed is 20 km/hr9  While the model is not looking at the cost of the trip, it includes trip time delays such as the time needed to park the car in a ‘park and ride’ stop and then continue with the rest of trip  Every 10-minutes of waiting in the interchange stop is deemed to be equally ‘costly’ as 15 minutes of walking  Destination wait weight is equal to 0.8  Every 10 mins of walk is equal to 15 mins  The model assumes a maximum interchange walk time of 10 mins, a maximum max cycle time of 120 mins and a maximum max car time of 180 mins  The model assumes that trip makers will always travel to the closest destination for retail trips

7.4 Results and Discussion

7.4.1 Population density and the location of centres

Brisbane has a low population density in most of its suburbs. Based on the Australian Bureau of Statistics (ABS) website, in June 2013, the population density of Greater Brisbane was 140 people per km2. The most densely populated ABS Statistical Area Level 2 (SA2) divisions in Greater Brisbane were inner-city New Farm (6,300 people per km2), Kangaroo Point (6,000) and Highgate Hill (5,300). In the same year, the largest increase in population density in Greater Brisbane occurred in the Woolloongabba SA2, which added 230 people per km2.

While the population density is growing, the need for retail and services is also increasing and therefore new shopping centres are being constructed or existing centres are being extended throughout the city. Higher populations means more

(9) cycle speed on footpath and dedicated cycle paths differ from the average speed/bike and ride mode cyclists are required to cycle to a train station and bikes are not permitted on trains.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 160 customers and therefore it is important for developers to identify where these populations are going to be located and how they will find their preferred shopping areas. It also gives the opportunity for the planners to direct these new comers into designated areas aligned with along growth corridors and around designated activity centres.

Population density was studied using ArcGIS software. Overlaying the layer of shopping centre locations with the population density map, it is immediately apparent that no shopping centres, even in the smallest neighbourhood centres, are located in areas with a population density of less than 500 people per km2. This shows the importance of the supporting population, or what is called as the ‘critical mass’, shopping centre location (Fig 7-2).

As previously discussed the hierarchy and the distribution of these centres in the city, follows Christaller’s central place theory (Getis and Getis, 1966) based on the distances between the centres and the supporting population within their catchments. The largest category of centres, (the super-regional, major-regional and regional centres) have a distance of between 10 to 16 km in the outer suburb and when it gets closer to the boundaries of inner suburbs with denser populated areas, this distance reduces even into 2.5 to 6 km. Population sizes of between 150,000 and 300,000 are considered to be the required critical mass for super-regional centres (referred to within the profession and also extractable from the overall population of BSD and the number of super-regional centres in the area). Figure 7.2 shows that (in 2011) only 3 super-regional centres (Garden City, Carindale and Chermside) were sufficient to serve the whole population of Brisbane.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 161

Fig 7-2: Population density and shopping centre locations in BSD in 2011, based on ABS data

While the population within the inner suburbs (almost 446,698 people based on the ABS 2011) is much higher than that in the outer suburbs, the inner suburbs have not yet experienced a major expansion in big shopping centres, when compared to the

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 162 outer suburbs. This might be because of the difficulties in finding sufficiently large areas of vacant land, much higher land prices, resistance from smaller existing businesses, or possibly because of differences in customers’ preferences for shopping destinations. It might also be the case that existing large shopping centres in Brisbane CBD already adequately support the shopping needs of the inner city population.

When it comes to sub-regional and neighbourhood centres, a larger number of these shopping centres have been developed within the inner suburbs to service the population living in this area (Fig 7-2).

7.4.2 Retail density around the centres

As previously discussed in the literature, the statutory documents in Queensland such as city plans and the SEQ regional Plan strongly support the ‘centre policy’ that new retail land uses should be located close to existing centres, or close to locations which have been designated for future development of shopping centres. The Brisbane City Plan 2000 strongly discourages any retail development outside suitably zoned-land.

The distribution of retail in Brisbane shows that shopping centres are located predominantly in higher density retail zones (Figure 7-3). The zones with the highest density of retail jobs are generally adjacent to the different types of shopping centres and the shopping centres locations follow major transport corridors (Fig 3-1)10.

The entropy index was used to measure land use mix in a one-kilometre radius around existing shopping centres. The one-kilometre radius was selected as a walkable distance that is often used by researchers (O'Sullivan and Morrall, 1996;

Pikora et al., 2001; Burke and Brown, 2007).

(10) The retail density data used here are drawn from the 2016 projections of the BSTM model. Careful interpretation is required in some locations where high densities of retail jobs are reported, however. For example, areas along the Brisbane River or alongside the Beaudesert Road close to Archerfield are reported to have a high retail density, but these locations contain high densities of warehouses and storage areas and are therefore not actually a good representation of typical retail areas. The BSTM is typically less reliable in this regard than the Shopping Centre Directory.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 163

Fig 7-3: Retail density within Brisbane Statistical Division for 2016

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 164

7.4.3 Mixed use analysis in the direct vicinity of shopping centres

The entropy index ranges between 0 and 1. The closer it gets to 0 is the more homogeneous the land use within the 1km radius from the shopping centre. The closer the entropy index gets to 1 the higher the level of mixed use. To use these entropy indices in subsequent analysis, entropy scores were categorised into three groups: a low level of mixed-use (entropy index between 0 and 0.33), a medium level of mixed-use (entropy index between 0.33 and 0.66) and a high level of mixed-use (entropy index higher than 0.66).

The mixed use results are summarised in Table 7-1 for shopping centres of different hierarchy levels, from Sup-Regional and Major-Regional down to Sub-Regional and Neighbourhood. A full list of the entropy indices for all of Brisbane’s 157 shopping centres is provided in Table 7-2.

Table 7-1: The number and percentage of centres of different hierarchy levels with High, Medium or Low levels of Mixed-use within a 1km radius

High Mixed % Medium % Low % use Mixed use Mixed use Hierarchy Level

Super Regional 1 33 2 67 0 0

Major Regional 1 25 3 75 0 0

Regional 6 55 5 45 0 0

Sub Regional 2 7 19 70 6 22

Neighbourhood 17 15 71 63 24 21

Grand Total 27 --- 100 --- 30 ---

Tables 7-1 and 7-2 show that none of the Super-Regional, Major-Regional or Regional centres are located in low mixed use localities. Sub-Regional and neighbourhood centres, however, are mostly located in localities of medium or low mixed-use. These differences may indicate that smaller centres are often surrounded by residential lots whereas larger centres tend to have clusters of different types of land uses in their vicinity. This could potentially make these destination even more attractive for multi-purpose trips.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 165

Table 7-2: Entropy index measured within a 1-km radius of each shopping centre centre type entropy centre type entropy centre type entropy Super Regional 0.53 Neighbourhood 0.21 Neighbourhood 0.48 Super Regional 0.56 Neighbourhood 0.22 Neighbourhood 0.49 Super Regional 0.67 Neighbourhood 0.22 Neighbourhood 0.49 Major Regional 0.46 Neighbourhood 0.24 Neighbourhood 0.49 Major Regional 0.50 Neighbourhood 0.25 Neighbourhood 0.50 Major Regional 0.55 Neighbourhood 0.26 Neighbourhood 0.51 Major Regional 0.70 Neighbourhood 0.28 Neighbourhood 0.51 Regional 0.40 Neighbourhood 0.29 Neighbourhood 0.52 Regional 0.46 Neighbourhood 0.29 Neighbourhood 0.52 Regional 0.55 Neighbourhood 0.29 Neighbourhood 0.52 Regional 0.62 Neighbourhood 0.30 Neighbourhood 0.54 Regional 0.63 Neighbourhood 0.30 Neighbourhood 0.55 Regional 0.68 Neighbourhood 0.31 Neighbourhood 0.55 Regional 0.70 Neighbourhood 0.31 Neighbourhood 0.55 Regional 0.76 Neighbourhood 0.32 Neighbourhood 0.55 Regional 0.77 Neighbourhood 0.33 Neighbourhood 0.56 Regional 0.80 Neighbourhood 0.33 Neighbourhood 0.56 Regional 0.81 Neighbourhood 0.34 Neighbourhood 0.56 Sub Regional 0.10 Neighbourhood 0.35 Neighbourhood 0.57 Sub Regional 0.21 Neighbourhood 0.35 Neighbourhood 0.57 Sub Regional 0.28 Neighbourhood 0.36 Neighbourhood 0.57 Sub Regional 0.30 Neighbourhood 0.36 Neighbourhood 0.58 Sub Regional 0.31 Neighbourhood 0.37 Neighbourhood 0.58 Sub Regional 0.33 Neighbourhood 0.37 Neighbourhood 0.58 Sub Regional 0.36 Neighbourhood 0.37 Neighbourhood 0.59 Sub Regional 0.36 Neighbourhood 0.37 Neighbourhood 0.59 Sub Regional 0.38 Neighbourhood 0.37 Neighbourhood 0.59 Sub Regional 0.38 Neighbourhood 0.38 Neighbourhood 0.62 Sub Regional 0.39 Neighbourhood 0.38 Neighbourhood 0.62 Sub Regional 0.39 Neighbourhood 0.39 Neighbourhood 0.62 Sub Regional 0.40 Neighbourhood 0.40 Neighbourhood 0.63 Sub Regional 0.47 Neighbourhood 0.40 Neighbourhood 0.64 Sub Regional 0.49 Neighbourhood 0.40 Neighbourhood 0.64 Sub Regional 0.51 Neighbourhood 0.40 Neighbourhood 0.66 Sub Regional 0.52 Neighbourhood 0.41 Neighbourhood 0.67 Sub Regional 0.53 Neighbourhood 0.41 Neighbourhood 0.67 Sub Regional 0.53 Neighbourhood 0.41 Neighbourhood 0.67 Sub Regional 0.58 Neighbourhood 0.42 Neighbourhood 0.67 Sub Regional 0.59 Neighbourhood 0.42 Neighbourhood 0.67 Sub Regional 0.59 Neighbourhood 0.43 Neighbourhood 0.67 Sub Regional 0.59 Neighbourhood 0.43 Neighbourhood 0.67 Sub Regional 0.62 Neighbourhood 0.44 Neighbourhood 0.68 Sub Regional 0.64 Neighbourhood 0.44 Neighbourhood 0.68 Sub Regional 0.70 Neighbourhood 0.45 Neighbourhood 0.70 Sub Regional 0.70 Neighbourhood 0.45 Neighbourhood 0.70 Neighbourhood 0.09 Neighbourhood 0.46 Neighbourhood 0.70 Neighbourhood 0.10 Neighbourhood 0.47 Neighbourhood 0.72 Neighbourhood 0.15 Neighbourhood 0.47 Neighbourhood 0.72 Neighbourhood 0.15 Neighbourhood 0.47 Neighbourhood 0.72 Neighbourhood 0.17 Neighbourhood 0.47 Neighbourhood 0.73 Neighbourhood 0.18 Neighbourhood 0.47 Neighbourhood 0.74 Neighbourhood 0.19 Neighbourhood 0.48 Neighbourhood 0.20 Neighbourhood 0.48

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 166

7.4.4 Accessibility of shopping centres by different modes of transport

The level of accessibility for each category of shopping centre by private car, public transport, bicycle and walking was calculated using the weighted time cost approach described in Section 7.3.3. These calculations were performed by calculating the weighted travel cost from each of 135,530 defined nodes in BSD’s transit system (as specified in LUPTAI model) to the exact locations of Brisbane 157 shopping centres (as specified in the SCD).

While the model uses the 135,530 nodes as the origins of the trips to measure the accessibility of any identified destination, having the accessibility results specified from all the transport nodes will not be beneficial for planning purposes. Planning studies usually use regional or zonal information in their analysis rather than nodal information. Fortunately, the LUPTAI model has the capacity to transform the results into zonal information using any defined zonal system such as transport zones or different ABS statistical areas. The ABS collection districts (CDs) were chosen as the zonal resolution at which the accessibility of shopping centres would be reported in this study11. Accessibility scores for the different categories of shopping centre (large, medium and small)12 by the four different transport modes (private car, public transport, bicycle and walking) are calculated for each of Brisbane’s collection districts. The model calculates the median accessibility score for each collection district considering accessibilities from all the transport nodes within that district. An example of the accessibility map is shown in Fig 7-5.

In order to look into the accessibility level for each type of the centres within the study boundary, the percentage of the CDs with different accessibility levels (less than 15 minutes, between 15 to 30 minutes, between 30 to 45 minutes, between 45 to 60 minutes and more than 60 minutes) were measured and reported. Rather than considering all the CDs within the study boundary to measure the accessibility percentage, only 2,944 CDs out of a total 3,135 were selected for this analysis. By

(11) CDs are selected for this analysis because the results from these analysis will be used later in chapter 8 and having them in this format will make it possible to match them to other existing data and make it applicable to include in the discrete choice dataset (12) Different types of shopping centre are categorized into three major groups of large, medium and small centres. The reason for this selection is due to the similarity between the super-regional and major regional centres that makes them quite similar destinations for the customers to consider

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 167 looking at the population density map, it is obvious that, the population density in the other 191 CDs are very close to zero. Therefore, accessibilities from these sparsely populated districts were ignored to produce more meaningful results (Fig 7-4).

Fig 7-4: The populated areas within the study boundary used for calculating the accessibility levels of the different categories of shopping centres

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 168

Fig 7-5: LUPTAI public transport accessibiltiy map for regional centres

Table 7-3 shows the accessibility of different categories of shopping centre from Brisbane’s 2,944 CDs with population densities of more than 100 individuals per km2. Accessibilities are reported separately for different transport modes. Accessibilities are reported as percentages, i.e. referring to accessibility by car. The top row of results in Table 7-3 indicates that a large shopping centre is accessible within 15 minutes of car travel from 30.5% of Brisbane’s 2,994 populated collection

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 169 districts. It also reveals that a medium shopping centre is accessible in the same time by car from 75.8% of collection districts while a small centre is accessible in 15 minutes of car travel from 97.1% of collection districts.

Table 7-3: Percentage accessibility of different levels of shopping centre by various mode types from 2,944 Brisbane collection districts with population densities >= 100 individuals/km2

Car accessibility percentage for various type of shopping centres within BSD Large Centre % Medium Centre% Small Centre % < 15 min. 30.5 75.8 97.1 15 < & < 30 min. 51.9 23.9 2.8 30 < & < 45 min. 12.1 0.1 0.0 45 < & < 60 min. 4.3 0.0 0.1 60 < min. 1.1 0.1 0.0 Total 100 100 100 Public Transport accessibility percentage for various type of shopping centres within BSD Large Centre % Medium Centre% Small Centre % < 15 min. 0.6 2.2 11.0 15 < & < 30 min. 6.8 16.4 40.2 30 < & < 45 min. 19.3 35.1 31.2 45 < & < 60 min. 24.0 24.4 8.6 60 < min. 49.3 21.9 9.1 Total 100 100 100 Cycling accessibility to shopping centres within BSD Large Centre % Medium Centre% Small Centre % < 15 min. 8.8 24.1 62.6 15 < & < 30 min. 20.3 37.8 27.1 30 < & < 45 min. 17.4 21.3 6.3 45 < & < 60 min. 17.3 8.1 2.3 60 < min. 36.2 8.8 1.7 Total 100 100 100 Walking accessibility percentage for various type of shopping centres within BSD Large Centre % Medium Centre% Small Centre % < 15 min. 0.5 1.9 8.1 15 < & < 30 min. 1.7 6.2 20.0 30 < & < 45 min. 3.1 8.4 21.0 45 < & < 60 min. 3.6 9.3 15.8 60 < min. 91.1 74.3 35.1 Total 100 100 100

The results show that small centres are accessible within less than 15 minutes by private cars for almost all the studies’ CDs. When it comes to medium size centres,

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 170 accessibility time extends to 30 minutes to achieve approximately the same coverage of CDs. Large centres are accessible within 30 minutes by private cars from almost 83 percent of the CDs.

Turning to public transport, small centres are accessible within 15 minutes by public transport from only 11 percent of the CDs, however this increases to 51 percent when public transport travel is extended to 30 minutes of cost-weighted travel time. While it might have been expected that medium and large centres would be relatively accessible via public transport from a high percentage of CDs, due to the location of public transport stations at or near such centres, surprisingly, results do not support this. Large and medium centres are accessible via 30 minutes of public transport time from only 7 and 18 percent of CDs, for large and medium centres respectively.

Also somewhat surprisingly, the different categories of centre seem to be more accessible by cycling than by public transport, making cycling the second most accessible transport mode, after the private car. For example, small centres can be accessed by less than 30 minutes cycling from almost 90 percent of CDs around the study area. This reduces to 62 and 29 percent accessibility within 30 minutes cycling for medium and large centres respectively. However, not too many people in Brisbane are interested in cycling to shopping destinations as generally, cycling is not a preferred mode of transport (Chapter 5, Fig. 5-3, Fig. 5-4). This might be because of concerns relating to the quality and safety of cycling routes in Brisbane.

As expected, accessibility of all categories of shopping centres is considerably poorer by walking than the other modes of transport. The large and medium centres are not accessible within 45 minutes of walking from more than 85 percent of the CDs. Even for the small centres (neighbourhood centres) providing daily households requirements, walking trips of less than 30 minutes (1-way trip time) are possible from only 28 percent of CDs. Small, neighbourhood centres are accessible by 45 minutes walking from almost 50 percent of CDs. A 45 minute walk – with shopping purchases – is unlikely to be considered an attractive option for most of the population, particularly given Brisbane’s sub-tropical climate.

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 171

These accessibility results show that shopping centres are most accessible by private cars, followed by cycling, while levels of accessibility via public transport and walking are substantially lower for all categories of shopping centre.

7.5 Conclusions

This chapter has quantified various spatial characteristics of shopping centres that have been previously identified in the literature as important factors which affect individual travel behaviour. These spatial characteristics will be used as explanatory variables in the model of destination choice which will be developed in Chapter 8. That analysis will aim to determine whether any spatial characteristics influence the preferences and behavioural choices of retail trip makers.

The analyses in this chapter revealed that population density and retail density are closely associated with one another, and also with the location of shopping centres across different hierarchy levels. Results also showed a close relationship between the size of the centres and the level of mixed use around them. A larger number of different land uses and a higher level of mixed-use are typically found adjacent to larger centres. This may be beneficial from a sustainability perspective because these multi-purpose locations could satisfy a broad range of retail, recreational and residential requirements, thereby reducing travel distances.

The results from the LUPTAI accessibility model for different levels of the shopping centres showed that all classes of shopping centre are highly accessible by private car, moderately accessible by cycling, but only poorly accessible by public transport and walking from 2,944 Brisbane collection districts with population densities of more than 100 individuals/km2.

As expected, accessibility by private car is generally very high, with all categories of centre typically accessible within less than 30 minutes of driving. This makes private cars an easy and attractive mode of transport for retail trips. Accessibility by public transport is considerably lower, even within heavily populated areas. PT accessibility results indicate that it typically takes longer than 30 or 45 minutes to get to medium or large centres from most locations in metro Brisbane. While trips of longer than 30 minutes might be considered lengthy, this level of public transport accessibility for

Chapter 7 Spatial analysis of the shopping centres’ characteristics and accessibility 172 larger centres might be still respectable. Small centres could be accessed in less than 30 minutes from more than half of Brisbane’s collection districts, and in less than 45 minutes from almost 80% of collection districts by public transport. However, considering the type of products that are going to be bought (daily products such as milk, bread, etc.); half an hour seems to be a long trip to make. While less than 15 minutes seems to be more acceptable accessibility time on public transport for doing the daily shopping, small centres are only accessible within this time from 11 percent of CDs. Despite these lower levels of accessibility, some shoppers might still prefer to use public transport for their retail trips because they might find it more relaxing and enjoyable.

While small centres are highly accessible by private cars, and still relatively accessible by cycling and public transport, only 10% of collection districts are within 15 minutes walking distance of a small centre. The accessibility of medium size centres by walking reduces even further; a medium centre is accessible by 15 minutes walk from less than 2 percent of collection districts. The accessibility situation is even worse for larger centres. The findings show that retail trips are very easy to make in private cars followed to some extent by cycling, but due to the low level of accessibility by walking and public transport these modes cannot be considered as desirable or even feasible options for such trips. However, while the low accessibility level of shopping centres via active transport modes is a problem that requires serious consideration by urban and transport planners, other factors such as the cultural and social characteristics and preferences may also have a significant impact on shoppers’ destination choices and transport modes.

Chapter 8’s destination choice model will aim to determine whether the spatial characteristics of shopping centres exert significant influence over the retail choice behaviour of Brisbane’s shoppers.

Chapter 8 Shopping Centre Destination Preferences Modelling

Chapter 8 Shopping Centre Destination Preferences Modelling 175

8.1 Introduction

Retail trip destinations and the factors that drive customers’ preferences for travelling to particular destinations have long been a focus of attention for shopping centre owners. However, the last six to seven decades of substantial growth in car ownership, together with the introduction of new types of large shopping centres in many big cities, have forced planners to start rethinking about how these destinations should be distributed within the city. These new large centres are located close to major roads, have a large number of car parking spaces and are highly dependent on access by private motor vehicles. Alternative access possibilities for these destinations utilising more sustainable modes of transport could be investigated as a way of achieving a more sustainable transport mix, a way less dominated by private cars.

Identifying those attributes that attract customers to shopping destinations may initially appear impossible given the large number of factors that can potentially affect shoppers’ destination choices. Many researchers regard distance or other trip attributes (time, cost, etc.) as the most important factors affecting customer preferences (Richards and Ben-Akiva, 1974; Kitamura et al., 1998) (Spiggle and Sewall, 1987). However, the socio-demographic characteristics of trip makers (age, income, and employment) and the spatial characteristics of destinations (accessibility, mixed-use or retail and population density in the vicinity of the centre) might also influence the transport mode choice of customers directly or indirectly through their impacts on final destination choices. Furthermore, preferences may also be influenced by the qualitative or quantitative characteristics of the destination, such as the site area, the number of available stores and/or the level of comfort and enjoyment achievable therein.

The focus of this chapter is the Brisbane Statistical Division (BSD) and its expanding network of shopping centres. Drawing on the findings of previous chapters which showed that the attributes of individual travellers, destinations and trips may influence travel behaviour, this chapter aims to apply those findings by (i) modelling shoppers’ destination choices, and (ii) investigating how significantly these various

Chapter 8 Shopping Centre Destination Preferences Modelling 176 factors influence destination decision making, as the second major question of this research (Fig 8-1).

This chapter focuses on shopping centres ranging in size from those found in a neighbourhood to those of super-regional scale servicing a very large catchment area. The intention is to see which size of shopping centre is being selected by travellers for different types of shopping trips, and to identify major drivers that affect shoppers’ destination choices among the different sizes of centre. The focus is on the size of the centres rather than all their various types and hierarchies. This focus is due to the great similarities found between different types of centres, including super-regional, major-regional and regional centres, and therefore the study considers them as one group, followed by sub-regional and neighbourhood centres. The neighbourhood centres in this study include those which already contain at least one supermarket – i.e. they are definitely a shopping centre, not just a location with a single convenience store.

The chapter starts with a brief review of the existing approaches used to identify the retail destinations trip makers, focusing on discrete choice models. The methodology

Chapter 8 Shopping Centre Destination Preferences Modelling 177 adopted in this study and the datasets applied are then defined. Data analysis is explained and outputs are described. The chapter is closed with a discussion and conclusion addressing limitations and avenues for future research.

Fig 8-2: Location of shopping centres in the Brisbane Statistical Division (BSD)

8.2 Literature Review

8.2.1 Discrete choice models and the study of retail trips

As discussed in the literature review, understanding shoppers’ destination choices and the factors which influence these choices is a complicated task. A range of different factors is likely to influence the destination choice and mode choice decisions, and, in many instances in a shopping context, these two decisions are unlikely to be independent – i.e. in many situations, destination choice will dictate mode choice. Qualitative and quantitative studies have explored the motivations for retail trip choice based on attributes of the destination, attributes of the trip, or of destination and trip attributes in combination (Keefer (1966), Robinson and

Chapter 8 Shopping Centre Destination Preferences Modelling 178

Vickerman 1976, Lord (1988), Ploeger and Baanders (1995), Marshall and McLellan (1998). Initially the focus of these studies was simple empirical observation of the catchment area (trade area) of the centres, while later studies considered the attractiveness of the destination using approaches such as Reilly’s Retail Gravitation Law (1931) and that of Huff (1963). Ibrahim (2002) suggested that these studies focussed mainly on ‘discouraging’ factors such as travel time, travel cost and distance of trip, but ignored other ‘encouraging’ qualitative factors such as enjoyment of the trip to the centre or the cleanliness of transport mode used (Ibrahim, 2002).

Since the 1970s discrete choice analysis based on disaggregated datasets has been used to study retail destination choice. Discrete choice models aimed to predict the probability of individuals choosing one alternative from a choice set of discrete alternatives, based on the utility that they derived from their visit. As Hensher et al.(2005) argue, individuals make decisions consciously or sub-consciously by comparing alternatives and then select a particular option from those alternatives by making a specific choice. This approach, based on random utility maximization (RUM), assumes that decision makers are utility maximisers so they select the alternative that delivers them the highest utility. The randomness arises because the analyst cannot observe all the influences which individuals take into account when maximizing their utility. This is a challenge for the analyst who wants to determine how people make their choices because many factors that affect an individual’s decision are not easily observable (Hensher et al., 2005). In comparison to previous models, discrete choice models are able to include additional factors such as socio- demographic characteristics, which could influence destination choice. Hence, discrete choice models are usually used when the dependent choice variable is discrete, such as travel destination, mode choice, activity participation location choice, residential location choice or route choice. McFadden initially used logit- based discrete choice models in the 1970s and 1980s to understand the transportation mode choices of trip-makers (Hausman and McFadden, 1984; Athey and Imbens, 2007).

In the planning context, discrete choice models have been used to study decision makers’ spatial preferences in a number of different settings such as the choice of residential location (Habib and Kockelman, 2008), or the choice of the recreational

Chapter 8 Shopping Centre Destination Preferences Modelling 179 destination (Termansen et al., 2004). In terms of retail trips, consideration has been given to different factors which influence the level of utility that a customer derives from their trip. Factors such as trip distance, retail floor space, retail employment and population around the destination have been considered as being influential in different studies of retail destination choice (Adler and Ben-Akiva, 1976; Recker and Kostyniuk, 1978; Ghosh, 1984; Kitamura and Kermanshah, 1984; Bernard, 1987; Innes et al., 1990; Limanond et al., 2005; Carrasco, 2008). Various types of logit models have been applied in these studies including the simple Multi-nomial logit (MNL) model, the Nested Logit model and Mixed Logit models, etc. to identify important factors which influence retail trip makers’ decisions.

In 1974 Richards and Ben-Akiva developed their disaggregated simultaneous destination and mode choice model for shopping trips based on the MNL model. Their results indicated that disaggregated models can be statistically satisfactory, even when calibrated from a limited number of observations (Richards and Ben- Akiva, 1974). Recker and Kostiniuk (1978) limited their study to destination choice for grocery shopping trips. They hypothesised that grocery shopping destination decisions were strongly influenced by three factors: the individual’s perception of the destination, the individual’s accessibility to the destination and the number of opportunities available within the choice set. The results of their MNL model confirmed that accessibility was the main factor influencing destination choice for grocery shopping customers (Recker and Kostyniuk, 1978).

Multinominal logit analysis was also applied by McCarthy (1980) in order to study the qualitative characteristics of travel behaviour, the alternative destinations associated with a traveller’s shopping activity and how these factors combined to influence shoppers’ behaviour. McCarthy’s MNL model was developed separately for a city centre and for a suburban area. The dimensions McCarthy considered in his study included generalized trip convenience (trip time, trip cost, trip arrival time, etc.), generalized trip comfort (cleanliness of travel, protection from bad weather, ride comfort, opportunity to stop at other places on the way to the shopping area, etc.), generalized trip safety (safety from accidents during the trip, etc.), generalized shopping area attractiveness (a good variety of merchandise, low price of merchandise, etc.) and generalized shopping area mobility (ease of parking at the

Chapter 8 Shopping Centre Destination Preferences Modelling 180 shopping area, etc.). While these generalized attributes were seen as being significant, the distance was negatively affecting the trip makers’ decisions. A later destination-mode choice model was developed for the central city environment in order to better understand the influence of generalized attributes on choice behaviour as well as for identifying the factors affecting non-work travel behaviour. McCarthy found that consumers’ attitudes in addition to trip time and cost, safety and the attractions of the shopping area, significantly affected destination choice (McCarthy, 1980).

Gautschi (1981) and Bucklin and Gautschi (1983) found that both retail centre attributes and transportation mode attributes were major factors affecting trip makers’ destination decisions among a set of planned suburban shopping centres and traditional, unplanned downtown centres. Safety, from accidents and crime, convenience, reliability, flexibility, travel atmosphere, comfort of ride, protection from weather, transportation cost and parking cost were all considered in their study (Gautschi, 1981). Gautschi claimed that consumers consider a combination of retail centre and transportation mode attributes in their decisions to patronize alternative retail centres (Gautschi, 1981). Innes et al. (1990) used a binary logit disaggregated behavioural model to identify the major attributes affecting automobile users’ destination choice for shopping trips. The most influential attributes were the stores’ hours of operation, quality of goods offered, availability of parking, price of goods, accessibility of the shopping area, selection of goods offered and protection from environmental influences (Innes et al., 1990). Two other studies by Agyemang-Duah et al. (1995) and Lee and Goulias (1997) also applied discrete choice models and found the land use pattern to be a directly observable factor which affects home- based shopping trip frequency (Agyemang-Duah et al., 1995) (Lee and Goulias, 1997). Limanond et al. (2005) applied a tour-based nested-logit model to study neighbourhood shopping travel decisions. Household socio-demographics, the day of week, the level of service and the attractions available at the destination were all found to significantly affect destination choice and mode choice. It was also concluded that, as the number of retail jobs at the destination increased, the probability of a shopping trip to that destination increased whereas, as travel time, cost and distance increased, trip probability decreased.

Chapter 8 Shopping Centre Destination Preferences Modelling 181

In a comprehensive review of the conceptual and econometric framework for non- work activity location choice, Sivakumar and Bhat (2007) found that zonal size attributes, zonal non-size attributes (e.g. population density, geographic dummy indicators), zonal impedance measures, demographic variables (generally interacted with other variables), attributes of choice occasion (e.g. time of the day, day of the week, accompanying person) and feedback effects (variables that account for past experience) had all been considered in shopping destination choice models (Sivakumar and Bhat, 2007; Bekhor and Prashker, 2008). Carrasco (2008) investigated individuals’ selection of destination for grocery shopping by applying a discrete choice modelling technique and generating a choice set based on the travel time budget of the individuals, and a choice set constructed as a random sample from a fixed number of alternative potential destinations. MNL models were constructed separately for walking and car trips. For car-based trips, store size and travel distance were found to be the most relevant attributes in the destination choice process. Among the socio-demographic characteristics of the travellers, household size was found to be more important than age or gender (Carrasco, 2008). Schneider’s research in 2011 used mixed logit models to look at the mode choice of customers in different shopping districts. The results show that “keeping the travel factors such as time, cost, socio-demographic characteristics and individual attitudes unchanged, there is a negative association between the higher employment density, smaller parking lots and metered on-street parking in the shopping district and the usage of private cars. On the other hand, higher employment density, larger number of street tree canopy coverage, lower speed limits and less driveway crossing in the shopping district will be assspciated with more walking trips” (Schneider, 2011).

Table 8-1 (below) summarises the literature on modelling shopping destination choice, identifying the type of discrete choice model used, the choice behaviour being modelled, hypothesised drivers of choice behaviour and key results.

8.2.1.1 Summary

An examination of the literature shows that studies mostly focus on specific types of shopping trips, such as trips for buying groceries or trips that are undertaken by a specific group of trip makers such as automobile users, etc. Some studies have tried to look at the types of destinations being chosen by customers, such as

Chapter 8 Shopping Centre Destination Preferences Modelling 182 traditional/planned or city centre/suburban shopping destinations. However, from what was found in this research, no studies have investigated customers’ destination choices among various types or sizes of shopping centres. This is an important issue, considering the rapid growth in the number of shopping centres in many big cities around the world, including Australia, and the expanding portfolio of types or sizes of shopping centres available, from large super-regional centres to smaller local centres.

Even though it has not been reviewed in detail here, the body of literature that examines shoppers’ mode choice decisions – and the attributes and characteristics which affect mode choice - is considerably larger than that which looks at destination choice and the attributes and characteristics which affect destination decisions. When considering sustainability from an urban planning perspective, a focus on mode choice rather than destination choice is problematic. This is because planning policies such as development zoning, population and employment distribution and density, etc. are capable of influencing the size, nature and location of shopping centres as destinations for shopping trips. Planning policies are not, however, generally able to influence investment in transport infrastructure, which would be expected to influence choice of travel mode.

Besides, if a destination is attractive enough to catch the attention of travellers and provide the utility they are looking for, trip makers will probably find a mode that gives them easy access to that destination.

In relatively low density cities such as Brisbane, public transport investments have focused on key arterial corridors to the Central Business District (via rail, busway and City Cat ferries). Since shopping centres of various sizes are distributed throughout the Brisbane Statistical District (Fig 8- 1), this focus on arterial public transport corridors to the CBD has left private cars as the primary mode of transport for shopping trips (see the findings in Chapters 5 and 6). Given the distributed spatial layout of shopping centres, a completely unrealistic level of investment in citywide public transport networks would be required to overturn the dominance of the use of private vehicles for shopping trips.

Hence, this chapter will use a discrete choice model running on shopping trip data from Brisbane’s travel survey to investigate how shopping centre attributes, trip

Chapter 8 Shopping Centre Destination Preferences Modelling 183 distance, accessibility and shoppers’ socio-demographic characteristics influence the type of shopping centre which is chosen as the destination for particular categories of shopping trip (grocery shopping, clothes shopping, white goods shopping, etc.). An improved understanding of how these factors affect the type of shopping centre chosen for a shopping trip will help inform planning policies to facilitate reductions in the distance travelled to popular types of locations.

Chapter 8 Shopping Centre Destination Preferences Modelling 184

Table 8-1: Summary of the literature on discrete choice models for shopping trips

Researcher Year Type of choice Choice behaviour Hypothesised influential factors Results model modelled

Recker & 1978 Multinominal logit destination choice for Individual’s perception of the destination, Accessibility found to be an important factor

Kostiniuk (MNL) model grocery shopping Individual’s accessibility to the destination trips Number of destinations available to individuals

McCarthy 1980 Multinominal logit destination choice Generalized trip convenience (trip time, trip cost, Consumers’ attitudes and trip time, safety and (MNL) model and trip arrival time, etc.) availability of parking found to be important

mode-choice for city Generalized trip comfort (cleanliness of travel, factors centre and suburban protection from bad weather, ride comfort, areas opportunity to stop at other places on the way to the shopping area, etc.)

Generalized trip safety (safety from accidents during trip, etc.)

Generalized shopping area attraction (good variety of merchandise, low price of merchandise, etc.)

Generalized shopping area mobility (ease of parking, cleanliness, etc.)

Gautschi 1981 A commonly used logit destination choice for retail centre attributes & transportation mode A combination of destination attributes and

Bucklin & 1983 model both a set of planned attributes including safety from accidents and transport mode attributes found to be very Gautschi suburban shopping crime, convenience, reliability, flexibility, travel important centres and atmosphere, comfort of ride, protection from traditional unplanned weather, transportation and parking cost

Chapter 8 Shopping Centre Destination Preferences Modelling 185

downtown centres

Innes et al. 1990 binary logit destination choice of Distance, availability of parking, loyalty (habit); store hours of operation, quality of goods offered, disaggregated automobile users for hours of operation (in terms of shopping availability of parking, price of goods, behavioural model shopping trips convenience); price (differences between accessibility of the shopping area, selection of alternative locations); quality of goods; selection in goods offered and protection from the each location; and environment (outdoors in the environment are the major factors affecting CBD in contrast to the "controlled climate" at shoppers’ destination choice shopping centres)

Agyemang- 1996 Multinominal logit Destination choice Land use pattern directly affects home-based Duah et al. & and (MNL) model for home based shopping trip frequency Lee et al. 1997 shopping trip

Limanond et 2005 a tour-based nested- Residents’ decisions, Considered factors indicated in past literature, household socio-demographics, day of week, al. logit model living in a residential location within the neighbourhood, level of service, attractions, and number of retail neighbourhood regional setting and household structure jobs at the destination are all significant in the across five model dimensions of shopping travel: household tour frequency, participating party, shopping tour type, mode and destination choices

Chapter 8 Shopping Centre Destination Preferences Modelling 186

Sivakumar & 2007 A framework for non- Review of existing zonal size attributes, zonal non-size attributes Bhat work activity for a destination choice (e.g. population density, geographic dummy discrete choice model literature indicators), zonal impedance measures, demographic variables (generally interacted with other variables), attributes of choice occasion (e.g. time of the day, day of the week, accompanying person) and feedback effects (variables that account for past experience) have been studied in various research

Carrasco 2008 MNL models using two destination choice for Variables used in the modelling process included For car-based trips, store size and travel distance choice sets (one based grocery shopping retailers ID (categorical), size categories, opening in addition to household size were found to be on the travel time hours, dist. home-shop (car), detour distance the most influential factors budget of the (car), dist. home-shop (walk), detour distance individuals and the (walk), accessibility variable, age, monthly other constructed as a household inc., household size, gender variable random sample from a fixed number of alternative potential destinations.)

Schneider 2011 Mixed logit models mode choice of Higher employment density, larger number of customers in different street tree canopy coverage, lower speed limits shopping districts and less driveway crossing in the shopping district will be assspciated with more walking trips

Chapter 8 Shopping Centre Destination Preferences Modelling 187

8.3 Methodology

The literature review shows that different forms of Random Utility Model(RUM)- based logit-form discrete choice models have been used to predict the probability of individuals choosing a particular alternative from a choice set of discrete alternatives, such as the Multinomial Logit Model (MNL), Nested Logit Models, Random Parameter Logit (RPL) models, etc.

8.3.1 The Multinominal Logit Model

The Multinomial Logit Model (MNL) is the most widely used discrete choice model when more than two choice alternatives are available. The MNL assumes independent and identically distributed (IID) errors. This leads to a behavioural consequence of independence from irrelevant alternatives (IIA). IIA requires that the likelihood of selecting one alternative over another alternative is independent of the presence (or absence) of any additional alternative(s) in the choice set. This IIA property renders MNL models inappropriate for some choice applications. In practice, however, a simple MNL model is almost always developed to provide a ‘starting point’ description of choice behaviour. More sophisticated models such as latent class models (LCM) and random parameter logit (RPL) models can accommodate preference heterogeneity among decision makers by relaxing the MNL’s IID assumption. The application of more complicated models are due to the result of the Hausman test which will be run later to test the IIA assumption associated with the MNL models. The Hausman test of IIA is used to determine whether these more sophisticated choice models are required. If the results for the Hausman test do not support the IIA assumption, the study will be followed by the development of more complicated models1. However, as Hensher explains: “Only once one is confident that the MNL model is deficient should one progress to more advanced models” (Hensher et al., 2005).

(1) The Latent Class Model (LTC) which looks at the individual heterogeneity has also been used later to compare the results. These results will be explained in the appendices due to the results of the Hausman test, which will be discussed later in the chapter

Chapter 8 Shopping Centre Destination Preferences Modelling 188

In an RUM framework, Equation [1] represents the utility U of a particular alternative i in the choice set 퐶푛 of decision-maker n (푈푛푖) as consisting of an observed systematic or deterministic component (푉푛푖) and a randomly distributed, unobserved component (휖푛푖) capturing the uncertainty.

[1] 푈푛푖 = 푉푛푖 + 휖푛푖

Systematic utility 푉푛푖 is expressed as a function of 푋푛푖 attributes of alternative i and decision-maker n and corresponding estimated coefficients, 훽푖. The general form of systematic utility is:

[2] 푉푛푖 = ∑ 훽푖푋푛푖

On the other hand, “The analyst does not have any information about the error term” (Koppelman and Bhat, 2006a). This error term encompasses unexplained influences on utility that can arise from different sources, such as imperfect information, measurement errors, omission of model attributes, omission of the characteristics of the individual that influence his/her choice decision, etc. The distribution assumed for these error terms shapes the mathematical form of the discrete choice model.

Fig 8-3: Probability density function for Gumbel and normal distributions (same mean and variance) (Koppelman and Bhat, 2006b)

If error terms are assumed to be normally distributed, a Multinominal Probit (MNP) probabilistic choice model results. On the other hand, if the errors are assumed to be independently and identically distributed following the Gumbel distribution (extreme-value), a Multinominal Logit Model (MNL) will result. The

Chapter 8 Shopping Centre Destination Preferences Modelling 189

MNL model is widely applied in the literature (Hensher et al., 2005; Koppelman and Bhat, 2006a; Train, 2009).

Assuming that the alternative with highest utility is chosen, the probability 푃푛푖 of decision maker n choosing alternative i from choice set 퐶푛 is:

[3] 푃푛푖 = 푃푟표푏 (푉푛푖 + 휖푛푖 > 푉푛푗 + 휖푛푗) 푤ℎ푒푟푒 푗 ≠ 푖 푎푛푑 푖, 푗 ∈ 퐶푛

푃푛푖 = 푃푟표푏 (휖푛푗 − 휖푛푖 > 푉푛푖 − 푉푛푗)

The logit model’s IID assumption dictates that error components (휖푛푗 − 휖푛푖), are independently and identically Gumbel distributed across alternatives, which means there is no covariance between errors for alternatives i and j, i.e.

퐶푂푉 휖푛푗 − 휖푛푖 = 0 and error structure is identical for decision maker n across both alternatives i and j. The logit model for two alternatives (1 and 2) therefore produces the following choice probabilities:

[4] 푃푛1 = exp(푉푛1) / 푒푥푝 (푉푛1) + exp(푉푛2)

푃푛2 = 1 − 푃푛1 = exp(푉푛2) / 푒푥푝 (푉푛1) + exp(푉푛2)

With more than two alternatives, the model expands to the Multinominal Logit (MNL) model, with the choice probability for alternative i and decision maker n given by (Koppelman and Bhat, 2006a; Train, 2009):

[5] 푃푛푖 = exp(푉푛푖) / ∑ 푗 ≠ 푖 퐶푛 exp(푉푛푗)

8.3.2 Model estimations

Running an MNL model, it is important to measure the overall model fit and its significance, assess whether the inclusion or exclusion of different variables affects the overall model fit and to check the appropriateness of using the MNL model.

8.3.2.1 Maximum Likelihood Estimation as a measure of overall model fit

Rather than an ordinary least squares (OLS) model, discrete choice models use Maximum Likelihood Estimation (MLE) to find the parameter estimates that best explain the existing choice data. Therefore the statistical tests of model fit

Chapter 8 Shopping Centre Destination Preferences Modelling 190

commonly associated with OLS regression such as F-statistic cannot be applied to discrete choice models (Hensher et al., 2005).

Instead, the log likelihood statistic is used in a similar way to the residual sum of squares in a multiple regressions model (R2). It shows how much of the observed choice behaviour remains unexplained after the model has been fitted. As the log likelihood values get smaller (i.e. more negative), more of the observed choice behaviour remains unexplained, indicating that model fit is deteriorating (Field, 2013).

8.3.2.2 Determining model fit: the pseudo- R2

As well as the log likelihood measure, Pseudo- R2 and pseudo- R2 adjusted, statistics can be used to assess goodness of fit for discrete choice models.

To calculate a pseudo- R2 for a choice model the following equation is used:

2 푃푠푒푢푑표 푅 = 1 − (퐿퐿퐸푠푡푖푚푎푡푒푑 푚표푑푒푙/ 퐿퐿퐵푎푠푒 푚표푑푒푙)

Where:

LL denotes log likelihood (i.e. the log of the likelihood measure).

The estimated model is the model for which the Pseudo R2 is being calculated.

The base model is a simple model in which the deterministic component of utility is represented by a single constant, without any attribute-specific parameters.

2 2 The pseudo-R statistic generated by the model is different from the R statistic for a linear regression model. This goes back to the differences between the non-linear MNL models and the linear regression models. However, there is a direct empirical relationship between the two (Fig 8-4) (Hensher et al., 2005).

Chapter 8 Shopping Centre Destination Preferences Modelling 191

Fig 8-4: Mapping the pseudo- R2 to the linear R2 (Hensher et al., 2005)

8.3.2.3 Determining overall model significance

In order to measure the overall model fit, the log likelihood function estimate (LL) of the developed model should be compared with the LL function of a simple base model. If the LL function of the estimated model is statistically closer to zero than the LL function of the simple base model, this means that the estimated model provides a statistically significant improvement in fit, compared with the base model (Hensher et al., 2005). A model featuring only a single constant term is generally used as the ‘base model’ for this comparison.

If the LL function for a developed model is not statistically significantly closer to zero than that of the base model, this means that the inclusion of attributes (and attribute parameters) has not added significantly to the model’s capacity to explain the utility of the various choice alternatives, compared to the average utility calculated for each alternative in the simple base model. The statistical significance of the change in LL between the base model and the developed model is tested against a chi-squared distribution, with degrees of freedom equal to the difference in the number of estimated parameters between the base model and the developed model (Hensher et al., 2005).

A similar LL ratio testing approach enables assessment of the statistical significance of adding extra terms to the developed model. If the LL of the developed model improves significantly when extra terms are added, then those

Chapter 8 Shopping Centre Destination Preferences Modelling 192 extra terms should be retained in the model. Here again the change in LL is assessed for significance against a chi-squared critical value, with degrees of freedom equal to the number of additional parameters added to the developed model.

8.3.2.4 The Hausman test of the IIA assumption

The MNL model assumes independent and identically distributed (IID) error components, with a consequent assumption of independence of irrelevant alternatives (IIA) in the set of choice options. If the IIA assumption holds, the ratio of the probabilities of any two alternatives will remain the same regardless of the inclusion or exclusion of any other alternatives in the model. Conformity with the IIA assumption - and therefore also with the central IID assumption of the MNL – is tested by the Hausman test (Hensher et al., 2005).

In the Hausman test, one or more alternatives are removed from the set of alternatives available in the choice set and the change in predicted choice probabilities for the remaining alternatives (before and after removal of the excluded alternatives) is evaluated. For simple models with generic constants and attribute parameters, the Hausman test can be implemented via a single command in the modelling software. For more complicated cases, the comparison of probabilities is more involved. A T-statistic (q) for the difference in choice probabilities for the models from the unrestricted and restricted choice sets can be developed from the following equation:

′ ′ 푞 = [푏푢 − 푏푟] [푉푟 − 푉푢] [푏푢 − 푏푟]

Where - 푏푢 is a column vector of parameter estimates for the unrestricted model and 푏푟 is a column vector of parameter estimates for the restricted model; and - 푉푟 is the variance–covariance matrix for the restricted model and 푉푢 is the variance–covariance matrix for the unrestricted model.

If the test-statistic (q) is smaller than the 0.05 critical value, the null hypothesis of IIA would be rejected, indicating that the MNL model is inappropriate and that a more complex model which relaxes the IIA assumption would have to be developed (Hensher et al., 2005).

Chapter 8 Shopping Centre Destination Preferences Modelling 193

8.3.3 Data preparation and choice set construction

A number of choice models are developed in this chapter, aiming to explain shoppers’ decisions regarding retail destinations, and to identify influential factors which affect these decisions. However, before specifying particular choice models, the datasets available for model construction will be discussed. This is approached in two sections. Firstly, the types of the data available are described and the processes used to construct linkages and connections between those data are explained. Secondly, the construction of a specific choice set for use in discrete choice modelling is described.

8.3.3.1 Datasets and data preparation

Two types of datasets can be used for discrete choice models. Revealed preference (RP) data are derived from situations where the observed choice is made in real market situations, whereas stated preference (SP) data are derived from situations where a choice is made by considering hypothetical situations, for example by using specially constructed choice cards (Hensher et al., 2005). This study uses an RP dataset derived from Household Travel Survey (HTS) data which reports actual retail trips taken by travellers within the Brisbane Statistical Division (BSD) (The Urban Transport Institute, 2010).

The studies described in Chapters 5 to 7 have found that, in a broad sense, particular factors appear to influence the retail travel behaviour of customers. Data on these factors are combined to form a data set from which a set of specific choice alternatives, a ‘choice set’, can be constructed for the destination choice models which will be developed in this chapter. The factors from which the choice set could be constructed include information on:  attributes of the shopping trip [trip distance, type of product to be purchased (e.g. groceries, clothing, white goods), day of the week]  attributes of the location around a specific shopping centre which either was, or could have been, chosen as the trip destination [retail density, accessibility level, mixed used density, population density, all within a 1 km radius buffer around a centre]

Chapter 8 Shopping Centre Destination Preferences Modelling 194

 attributes of the specific shopping centre which was, or could have been, the chosen destination ground lettable area for retail (GLAR), number of parking spaces, number and floor area of major tenants and speciality stores within a centre, etc. – all obtained from the Shopping Centre Directory for Queensland  characteristics of the individual trip maker [age, occupation, car ownership and whether or not there are children in the trip maker’s household].

A number of assumptions have been made to enable connections to be constructed between the datasets in a meaningful way. While these various datasets have been described in previous chapters, they will be named and defined again here to assist understanding of how the connections between them have been established.

1 - South East Queensland Household Travel Survey data (SEQ-HTS): As previously explained in Chapter 5, the 7-day 2009 SEQ-HTS provides records of shopping trips. While in the analysis in Chapter 5 all 3,354 shopping trips with the purpose of buying something were considered, this chapter will only focus on 1,727 shopping trips for which the destination was reported to be shopping centres and supermarkets. Even though the shopping centres are defined as ‘regional under-cover shopping centres’ in the HTS, it is assumed that the trips to ‘supermarkets’ (the second most frequent destination for HTS retail trips after ‘shopping centre’) refer to neighbourhood centres comprising of at least one large supermarket, such as Coles or Woolworths, together with a number of speciality shops.

2 - Directory of Shopping Centres/Queensland 2011 (SCD): This dataset is produced by the Property Council of Australia. It details spatial characteristics of shopping destinations in Queensland, including information about the 195 shopping centres in the BSD. The SCD follows the classification of shopping centres employed in all of the Property Council’s shopping centre directories. In descending order of size these categories are:

 City centre - total gross lettable area retail exceeds 1,000 square metres.

Chapter 8 Shopping Centre Destination Preferences Modelling 195

 Super-regional – total gross lettable area retail exceeds 85,000 square metres  Major regional - total gross lettable area retail generally ranges between 50,000 and 85,000 square metres  Regional - total gross lettable area retail typically ranges between 30,000 and 50,000 square metres  Sub-regional - total gross lettable area retail will typically range between 10,000 and 30,000 square metres  Neighbourhood - total gross lettable area retail will typically be less than 10,000 square metres.

There are separate categories for themed, market and bulky goods shopping destinations.

Although the 195 shopping centres in the BSD encompass all the abovementioned categories, trips to the city centre, bulky good centres, market and themed centres were not retained for the destination choice analysis. This was because trips to these types of destinations differed significantly from the remainder of shopping trips in the HTS in terms of the retail trip type, customers undertaking those trips and also the type of products provided at the destination. As an example, it was found in Chapter 5 that more than 40 percent of the retail trips to the city centre are work-based shopping trips which take place during working hours every week. There is a high reliance on travel modes other than cars in this area, due to the shortage of parking spaces and the high cost of parking. The majority of customers are younger, employed customers and a low proportion of visits are by retired, elderly groups. Besides, the urban form attributes are very much different in the CBD compared to the rest of the city.

After this exclusion, 150 shopping centres remained as potential destinations for inclusion in the choice set for destination choice analysis.

This leaves five centre categories which could potentially be included in the study. An inspection of the characteristics of the different sizes of centres reveals considerable similarities between the super-regional, major-regional and regional centres. The three categories of super-regional, major-regional and regional centres were therefore merged together to form a single large centres (LC) group. The remaining centres were then conveniently categorised as

Chapter 8 Shopping Centre Destination Preferences Modelling 196 medium centres (MC) (for the sub-regional centres) and small centres (SC) (for neighbourhood centres) – as summarised in Table 8-2 below.

Table 8-2: Number of shopping centres in each condensed size category

Large centre Medium centre Small centre

Number of cases 18 27 105

Size of the centre More than 30’000 m2 10’000 to 30’000 m2 Less than 10’000 m2

3 - LUPTAI Model: to measure the accessibility of various types of centres within the BSD, a Land Use and Public Transport Accessibility Index Model, developed by the Department of Transport and Mains Road (TMR) was applied, as previously explained in Chapter 7. The accessibility results – which detail the median accessibility of the different categories of shopping centre from ‘collection districts’ identified as origins for HTS trips - can be used as a ‘destination attribute’ in the destination choice model .

4 - Land-use data: land use data provided by the Queensland Government as part of a national catchment scale land use mapping project, coordinated by the Australian Bureau of Agricultural and Resource Economics and Sciences (ABARES), are available from the ABARES website for various years including 2011. Explained in detail in Chapter 6, the land use mix index around each centre was measured using the entropy technique, and the findings were used here as a destination location attribute in the destination choice model.

5 - Australian Bureau of Statistics (ABS) 2011 provided information on the population and household characteristics within its small zonal system (mesh blocks) around the shopping centres’ locations.

6 - Brisbane Strategic Transport/Land use Model (BSTM): The BSTM model is a strategic transport and land use model provided by the Department of Transport and Main Roads and, calibrated with household travel survey data, includes zonal data for almost 1,500 zones covering the BSD area. The BSTM provides detailed information on the socio demographic characteristics of the households living in each zone, as well as the number of jobs categorised into sectors: retail, agriculture, cultural-recreational, wholesale etc. The BSTM also

Chapter 8 Shopping Centre Destination Preferences Modelling 197 contains detailed information on the transport network within the city, including public transport routes and the street network.

The six datasets were linked together to form a composite dataset for use in the destination choice model. ArcGIS software was used to overlay the different zonal systems from the HTS (Census Collection Districts (CCDs)), the BSTM model and the geocoded location of the shopping destinations.

The 150 geocoded shopping centres were overlaid and matched with the BSTM zones and the HTS CDs, so that the zonal information could be transferred to the centres as well (Fig 8-5). Besides matching the locations of the shopping centres with the destination collection districts from the HTS, this trip and destination matching process revealed that 694 of the shopping trips reported in the HTS ended at a destination zone which did not contain a shopping centre, as designated by the SCD. These 694 trips were therefore removed from the dataset, leaving a total of 1,033 trips to go forward to the destination choice analysis.

Chapter 8 Shopping Centre Destination Preferences Modelling 198

Fig 8-5: The zonal boundaries for the Census Collection District and the BSTM model

Chapter 8 Shopping Centre Destination Preferences Modelling 199

The dataset for destination choice analysis thus comprised 1,033 retail trips, each described in terms of the following attributes:  Trip distance  Accessibility level  Type of product purchased  Day of the week on which the trip was undertaken  Retail density within a 1 km radius buffer around the destination  Mixed used density within a 1 km radius buffer around the destination  Population density within a 1 km radius buffer around the destination  Ground lettable area for retail at the destination shopping centre  Number of parking spaces at the destination shopping centre  Number and floor area of major tenants and speciality stores within the destination shopping centre, and additional information, such as the presence of a food court or cinema, etc.

In addition, the following characteristics are known for the trip makers who undertook each shopping trip:  Age  Occupation  Whether or not the trip maker’s household contained children  Car ownership

Fig 8-6: Available attributes of the trip, destination centre and destination location, and characteristics of trip makers – for potential inclusion in the destination choice model

Chapter 8 Shopping Centre Destination Preferences Modelling 200

8.3.3.2 Selecting attributes for inclusion in the destination choice model

8.3.3.2.1 Correlation between the attributes

The level of correlation between the various attributes was assessed to help decide which attributes to include in the destination choice model. Various methods have been used to check the correlation between factors, including scatter plots, R2 comparisons and Spearman correlation methods, followed by simple tables to check the pattern of categorical factors (Field, 2013).

Many of the shopping centre destination attributes from the SCD are highly correlated. For example, the number of parking spaces, Gross Lettable Area- Retail (GLAR), a centre’s site area, the number of major tenants, the number of specialty stores, Moving Annual Turnover (MAT), and the presence of a food court or cinema could all potentially be correlated. Closer inspection of the characteristics of the centres and the relevant planning regulations indicated underlying linkages between the different attributes. For example, the number of parking spaces per square metre of retail area is directly correlated with the size of the centre and how many stores it contains. It was therefore unsurprising that a scatter plot and resulting R2 analysis revealed strong correlations between many of the destination attributes. Strongly correlated attributes cannot be applied in the same discrete choice model. The resulting R2 correlations from a simple scatter plot (excluding rows with missing information) between some of these variables are reported in Table 8-3.

Table 8-3: Correlations among attributes of destination centres

Variables R2

GLAR1 and MAT2 0.92

GLAR and site area 0.47

GLAR and no. of parking spaces 0.86

Site area and no. of parking spaces 0.54

1Ground Lettable Area for Retail / 2 Moving Annual Turnover

Other critical information about the function and character of the centres which could be representative of the centres’ performances when alternatives are

Chapter 8 Shopping Centre Destination Preferences Modelling 201 assessed, such as the centre’s turnover, the number of visitors or other qualitative information on the centre’s environment, entertainment and level of comfort, are not captured in the dataset or are only partly reported (with up to one quarter of missing data), making it an unreliable factor for any further use in the modelling step.

Table 8-4: Important attributes of Super Regional Centres (data from 2011)

Major tenants GLAR No. of parking spaces Site area (ha)

Westfield Carindale 67,880 5,400 15.8

Westfield Chermside 67,826 6,500 14

Westfield Garden City 55,373 4,863 25

The Spearman correlation test and simple tables were used to test the collinearity between other attributes. These were reported as continuous or categorical variables, e.g. retail and population density, or mixed use density of location around the destination centre. The Spearman’s correlation coefficient was chosen since it can identify correlations among data which violate parametric assumptions (Field, 2013).

Multicollinearity between attributes will make it difficult for the model to assess the individual importance of each attribute in the destination choice decision. Therefore, given the correlations among the destination attributes shown in Table 8-2, it was decided to include only the GLAR attribute in the decision choice model as being representative of all the other destination location attributes.

Tests indicated considerably lower levels of correlation among the trip attributes and destination vicinity attributes (Table 8-4). Therefore, the following attributes were included in the initial MNL model of destination choice:  trip distance  accessibility  type of product being purchased  day of the week  retail density within a 1 km radius buffer around the centre  mixed use density within a 1 km radius buffer around the centre

Chapter 8 Shopping Centre Destination Preferences Modelling 202

 population density within a 1 km radius buffer around the centre  Ground lettable area for retail in the centre (GLAR).

In addition, the following characteristics of trip makers were included as potential interaction terms in the initial MNL model of destination choice:  age  occupation  whether there are children in the trip-maker’s household  car ownership.

Table 8-5: Correlations among attributes describing the spatial vicinity around shopping centres

DIST- Entropy Pop Retail Correlations GLAR KM (log) density density

Correlation ** ** ** ** 1.000 .206 .069 -.137 .074 Coefficient distance-KM Sig. (2-tailed) . .000 .007 .000 .004 N 1544 1544 1544 1544 1544

Correlation ** ** ** ** .206 1.000 .637 -.501 .498 Coefficient GLAR Sig. (2-tailed) .000 . .000 .000 .000 N 1544 1544 1544 1544 1544

Correlation ** ** ** ** .069 .637 1.000 -.253 .090 Coefficient Spearman's Entropy(log) rho Sig. (2-tailed) .007 .000 . .000 .000 N 1544 1544 1544 1544 1544

Correlation ** ** ** ** -.137 -.501 -.253 1.000 -.579 Coefficient Pop density Sig. (2-tailed) .000 .000 .000 . .000 N 1544 1544 1544 1544 1544

Correlation ** ** ** ** .074 .498 .090 -.579 1.000 Coefficient Retail density Sig. (2-tailed) .004 .000 .000 .000 . N 1544 1544 1544 1544 1544 **. Correlation is significant at the 0.01 level (2-tailed).

8.3.3.3 Formation of the choice set for the destination choice model

A brief summary of the literature, provided below, indicates that preparation of the choice set of available alternatives for RP data is an important issue in

Chapter 8 Shopping Centre Destination Preferences Modelling 203

RUM-based destination choice modelling. Choice set construction can affect the outcome of RP-based discrete choice modelling considerably. The first of the sub-sections below presents a brief summary of the literature on choice set construction for RP-based discrete choice modelling. The second sub-section explains why a particular approach for choice set construction was adopted in this research.

8.3.3.3.1 Literature relating to choice set construction for RP-based discrete choice modelling

Discrete choice models generally assume that a decision maker selects the alternative which delivers them the highest utility among the options available at the time that the choice is made. The choice set from which an individual makes their decision is considered to be mutually exclusive, collectively exhaustive and the number of alternatives is considered finite (Nerella and Bhat, 2004).

A major, practical issue which arises in RUM-based discrete choice modelling is the preparation of the choice set of available alternatives, since in destination choice settings the number of alternatives in an individual’s choice set might be very large indeed. For example, for residential choice situations or shopping trip destinations, a decision maker may potentially have up to a few hundred choice alternatives available to them (Bernard, 1987; Nerella and Bhat, 2004). Swait (2001) suggests that in RP settings the actual number of alternatives available to a trip maker is not known to the analyst, since the analyst is only aware of the actual decision made by the traveller. Even though the individual’s choice set might be deterministic (which means that the trip maker has a specific location in mind in advance), in RUM-based modelling the choice set must be identified in probabilistic terms because the analyst has incomplete information about the factors that generated the trip decision, such as unobserved attributes or tastes. The analyst must therefore devise a choice set which includes meaningful alternatives to the destination which was actually chosen (Swait, 2001; Yang et al., 2009).

Early models assumed that all possible alternatives were evaluated by an individual before selecting a preferred alternative (Fotheringham, 1988). Later this assumption was challenged because of its infeasible data processing

Chapter 8 Shopping Centre Destination Preferences Modelling 204 requirements, and also because of the practical limitations that inevitably restrict the number of alternatives available to any individual decision maker (Carrasco, 2008). Subsequently, the set of destination choices available to an individual was generally limited to only a portion of all available options.

Akiva and Lerman proposed using a “simple random sample of alternatives” drawn as a subset from the full set of possible alternatives (Akiva and Lerman, 1985). This method attempts to confine the number of available alternatives included in the model; however, it will be very unlikely to provide an appropriate basis for choice set construction in some circumstances. This is because not all of the randomly chosen alternatives in the artificially constructed choice set are likely to be meaningful alternatives to the chosen destination for a particular trip maker.

Other approaches to choice set construction select alternatives for inclusion in the choice set based on an alternative’s attractiveness or relevance to the trip maker. For example, selection of appropriate alternatives for inclusion in the choice set could be based on levels of attributes such as retail floor area, distance from home or the absence of major barriers (e.g. a highway, river, etc.) between the origin and the destination. In this type of approach, choice set formation – in addition to destination choice – is regarded as probabilistic within the RUM framework (Ying He, 2006).

Examples of such methods include the time-geography perspective of Hägerstraand (1970) and the space-time prism approach of Thill (1992). The space-time prism embodies spatial and temporal constraints on the movement of individuals and tries to measure the accessibility of the destinations available to any individual. This focus on the constraining environments makes the space-time prism a suitable framework for constructing a choice set for shopping destination choice (Timmermans and Golledge, 1990). Many different techniques have evolved around this concept, considering factors such as location and operating hours, and the number of routes to the activity (as a surrogate for accessibility). Some studies have focused on the impacts which imposition of these constraints on the choice set may have on the precision of the discrete choice models, which will subsequently use the constructed choice

Chapter 8 Shopping Centre Destination Preferences Modelling 205 sets. Examples include Kitamura and Kermanshah’s study into four types of activities using the geography framework and temporal factors, which considered time-of-day dependencies of activities and trip makers’ time budget for going on the trip and returning home (Kitamura and Kermanshah, 1984; Arentze et al., 2006).

Miller (1991) developed an algorithm for implementing a space-time prism in a GIS environment. Miller defined basic space-time prism concepts based on operational definitions such as travel times through network nodes, turn times and stop times. Miller claimed that since the travel time varies across the network and fluctuates through time, the shortest path through the network should be selected to limit the number of alternatives. In 2006, Scott also suggested a shortest path approach for estimating the potential path area. In Scott’s method the shortest path travel time (SPT) was calculated using a visual basic for application (VBA) for programming in ArcGIS and this was then compared with the time budget (TB) of the trip maker. This time budget, which included the total time of the trip plus the participation time, was extracted from the travel survey data. Scott found a significant difference between the results of a discrete choice model which used a space-time constrained choice set, compared to the results from models which used unconstrained choice sets (Scott, 2006).

When decision makers can choose from a wide set of options, any of the abovementioned choice set selection methods, “simple random sample of alternatives”, the “time-geography perspective” or the “space-time prism approach”, could all be useful. The particular selection method used would depend on the type of data available to the researcher, the design of the study and the choices which would be investigated in the subsequent discrete choice model.

Chapter 8 Shopping Centre Destination Preferences Modelling 206

For this study, the presence of almost 150 shopping centres within the BSD as potential destinations for over 1000 shopping trips provides a sizeable pool of locations from which trip makers could potentially choose their destinations. A choice set selection method that can handle a data set of this (large) size is therefore required; a ‘labelled’ choice set in which the three alternatives facing each decision maker are described by the attributes shown in Section 1.1.1.1, together with a ‘label’ denoting the size category of the alternative: small, medium and large.

As explained in Section 1.3.3.1, the five original size categories of centres are condensed into three more general categories: small, medium and large. Centres within the same size category probably offer very similar shopping experiences. Hence, it is likely that the decision facing the trip maker is essentially which size of centre to visit: small, medium or large. It was therefore decided that the choice set for each trip maker in the destination choice model would feature the chosen centre (of a particular size category: small, medium or large), plus one example of each centre from the other, non-chosen size categories. For example, the choice set used to model the decision choice of a trip maker who chose to visit a small centre would feature one medium centre and one large centre, in addition to the chosen small centre. Whereas the choice set used to model the decision choice of a trip maker who visited a medium centre would feature one small centre and one large centre.

Both a simple random sample and a time-geography approach were considered for choice set construction.

Two different approaches were tested for this study to make the destination choice set of customers based on the concept of Hagerstrand’s time-geographic model and Akiva and Lerman’s simple random selection. Due to some limitations in data availability including the exact origins of the trips and the assumption considered to overcome this issue, the results from the first approach (based on time-geographic model) were not reliable and came up with

Chapter 8 Shopping Centre Destination Preferences Modelling 207 some complications. Therefore, the second method of simple random selection was selected to make the choice set 1.

In the second approach for constructing a labelled choice set, based on Akiva and Lerman’s method (1985), the choice set was constructed from a simple random sample of all available alternatives in each type (large, medium and small) excluding the chosen one (Akiva and Lerman, 1985). Therefore, it was decided to use the random sampling approach for including one destination from each shopping centre size category in the set of non-chosen alternatives, to add to the (known) chosen destination of a particular size category. The choice set alternatives for the two non-chosen types of centre were selected randomly from the full set of centres of that size category in the BSD.

A final complication arises in specifying the actual destination for the chosen trip if there is more than one shopping centre in the collection district which was identified as the destination for the trip. In these circumstances, the particular shopping centre destination was drawn by random selection from the multiple centres within the known destination collection district.

1 The first approach for construction of a small, medium or large centre size labelled choice set uses a deterministic approach similar to the time- geographic model of Hagerstrand. Knowing the origin of each trip, the closest shopping centre (in terms of Euclidian distance) of each of the non-chosen size categories is identified and included in the choice set for that trip, together with the chosen destination (of a known size category). This is an attempt to put some rational limitation on the way in which particular destinations are selected. Since all the centres within a specific level of hierarchy are sharing almost exact characteristics in terms of the type of products and services they provide, it is assumed that people will consider the closest centre of a particular size category as a potential destination for their trip. It is assumed that if there is very little difference in the shopping experience and facilities between different centres within a given size category, then a trip maker will choose between the closest destinations of each size when planning their trip. Reverting for moment to the original five size categories of centres – for example, for a trip maker who intends to shop at a super-regional centre, Westfield Garden City and Westfield Chermside probably appear very much alike – at least in terms of quantifiable attributes available to the analyst. Hence it is reasonable to expect that a trip maker who intends to visit a super-regional centre would choose whichever of these destinations is closer to the origin of their trip. A VBA algorithm was used to calculate the shortest Euclidean distance from the trip origin to centres of each of the three size categories. The closest centre of each of the non-chosen categories was then included in the choice set for a particular trip. Although the application of real network travel distance/time might provide a better understanding of the distances between an origin and all available destinations, this was not possible in this study because of the lack of information on the exact location of the trip origin. The HTS only provides the origin collection district within which the trip started. In the VBA algorithm for Euclidean distance, trip origin was therefore approximated by the origin collection district’s (X, Y) centroid.

Chapter 8 Shopping Centre Destination Preferences Modelling 208

8.4 Data analysis and results

8.4.1 The MNL model developed for this study

As explained above, eight independent attributes of the trip, the destination shopping centre and the vicinity around the destination, were retained for inclusion in the MNL model of destination choice. In addition, four characteristics of the trip maker were also available for inclusion in the MNL model as interaction terms. The trip distance, destination and destination location and accessibility from the origin attributes were alternative specific (i.e. related to the alternative destinations included in the choice set). In contrast, the trip makers’ characteristics were trip-maker specific (i.e. relating to the trip decision makers) and the shopping purpose attribute (type of goods purchased) and day of the week did not vary across the alternatives in a particular trip maker’s choice set.

Different software can be used to run discrete choice models including LIMDEP, Nlogit, R-statistics, etc. Nlogit software was used in this study.

In order to determine which one of these attributes should be included in a parsimonious model, a number of models were developed starting from a simple model which featured only two alternative specific constants (ASCs). When included alone in a choice model, the ASCs report the relative attractiveness of the different shopping centre types, devoid of any additional informational attributes such as trip distance, GLAR, mixed use density in the vicinity of the destination, etc. This ASC-only model formed the ‘baseline model’ for subsequent calculation of model performance (via Pseudo R2) and comparisons of overall model fit (using the log likelihood test). Attributes were then added to the baseline model as extra explanatory variables, e.g. trip distance, GLAR at destination centre, etc. The log likelihood and pseudo- R2 results were tested after each extra explanatory variable was added to see whether the performance of the each more complicated model improved, step by step, on that of its predecessor. The largest model tested included two alternative specific constants together with all eight attributes of the trip, destination and destination vicinity. Characteristics of the trip makers, day of the

Chapter 8 Shopping Centre Destination Preferences Modelling 209 week and type of goods purchased were also included in the larger models as interactions with the trip distance attribute. For example, an interaction term between trip distance and ‘children in the trip maker’s household’ would aim to detect whether trip makers with children were more averse to travel for some distance than trip makers without children.

Attributes were expressed as either categorical or continuous. Categorical attributes were:  category of good purchased  day of the week on which the trip took place (weekend vs. weekday)  socio-demographic characteristics of individual trip makers (car ownership [yes/no], children in the household [yes/no], occupation of the trip maker [employed, unemployed, student, retired]

Continuous attributes included:  Trip distance in km  Retail density in a 1km radius around the destination  Population density in a 1km radius around the destination  Level of mixed use in a 1km radius around the destination  Accessibility by car, public transport, bicycle or walking  Gross lettable area (GLAR) at the destination centre.

All the categorical attributes in the choice set were represented by dummy 푘 variables 퐷푛 and interacted with a trip specific attribute 푋푛푖 which varies across alternatives in the choice set. In the models developed here, these interactions were formed with the trip distance attribute. Hence the resulting interaction parameters would aim to identify whether distance aversion changed as the characteristics of the trip maker changed, the day of the week changed or as the purchasing objective of the trip changed.

푘=퐾−1 푘 [6] 푈푛푖 = 퐴푆퐶푖−1 + ∑ 훽푖푋푛푖 + ∑ 훽푖푋푛푖퐷푛 푘=1

Considering all the constants, the interaction terms for the categorical attributes and the direct influences of the continuous attributes, the total number of parameters included in the largest model is 22. The small centre (SC) is

Chapter 8 Shopping Centre Destination Preferences Modelling 210 selected as the baseline destination among the three size categories of alternatives, and all parameter estimates for impacts on preferences for the other categories of destinations will be interpreted relative to preferences for small centres as a baseline destination.

For each interacted categorical attribute (which enters the model via a dummy variable), a base level is chosen, against which the estimated parameters for the other levels of that variable will be compared for interpretation. Table 8-6 provides all relevant details for the different attributes.

Apart from the two alternative-specific constants, all the parameters in the model are considered as generic (rather than alternative-specific) and therefore the final model is:

[7] 푈 (퐿퐶, 푀퐶, 푆퐶 1) = < 퐿퐶, 푀퐶, 0

> + 푃푁푉퐸퐻푆퐷 ∗ 푁푉퐸퐻푆퐷 + 푃퐷퐼푆퐾푀 ∗ 퐷퐼푆퐾푀 + 푃푊퐸푆퐷 ∗ 푊퐸푆퐷

+ 푃푅퐸푇퐷퐸푁 ∗ 푅퐸푇퐷퐸푁 + 푃퐸푁푇푅푂푃푌 ∗ 퐸푁푇푅푂푃푌 + 푃푃푂푃퐷퐸푁 ∗ 푃푂푃퐷퐸푁

+ 푃퐶퐴푅 ∗ 퐶퐴푅 + 푃푃푇 ∗ 푃푇 + 푃퐵퐼퐾퐸 ∗ 퐵퐼퐾퐸 + 푃푊퐴퐿퐾 ∗ 푊퐴퐿퐾 + 푃퐾퐼퐷푆

∗ KIDSD + 푃퐴퐺퐸1퐷 ∗ AGE1D + 푃퐴퐺퐸3퐷 ∗ AGE3D + 푃푀퐴퐼푁퐴2퐷

∗ MAINA2D + 푃푀퐴퐼푁퐴3퐷 ∗ MAINA3D + 푃푀퐴퐼푁퐴4퐷 ∗ MAINA4D

+ 푃푀퐴퐼푁퐴5퐷 ∗ MAINA5D + 푃퐸푋푃2퐷 ∗ EXP2D + 푃퐸푋푃3퐷 ∗ EXP3D $

Developing the MNL model, the results are expected to help us recognise the elements that make each of the destinations more or less attractive (a large, medium or small type of centre) than another when a decision maker is starting his/her shopping trip. The results also provide an initial indication of which destination-specific, spatial-specific, trip-specific and individual-specific attributes appear to influence retail destination choice among Brisbane shoppers, based on the availability of datasets for our study. These results will

1 The small centre (SC) is selected as the baseline destination among the three size categories of alternatives

Chapter 8 Shopping Centre Destination Preferences Modelling 211 be very valuable in informing planners of customers’ preferences and in suggesting ways in which it might be possible to modify retail structure and reduce travel distances to popular types of shopping destination.

Table 8-6: Attributes included in the MNL models, their defined categories and the selected baseline level

ATTRIBUTE’ ATTRIBUTE BASE CATEGORIES ABBREVIATIONS DESCRIPTION CATEGORY

Distance between DISTKM (continuous)- origin and destination --- In kilometres Retail density in 1 km RETDEN (continuous)- --- radius around centre Mixed-use index in 1 ENTROPY (continuous)- km radius around --- centre Population density in POPDEN (continuous)- 1 km radius around --- centre Car travel time CARACC (continuous)- --- accessibility (mins) PT travel time PTACC (continuous)- --- accessibility (mins) Bike travel time BIKEACC (continuous)- --- accessibility (mins) Walk travel time WALKACC (continuous)- --- accessibility (mins) Day of the week on WESD [weekend] WESD which the trip was WDSD WDSD [weekday] undertaken NVEHSD [no vehicle in Household car NVEHSD household] VEHSD ownership *distance VEHSD [vehicle in household] NKIDSD [no children in Children in the KIDSD household] KIDSD household KIDSD [children in household] AGE1D (<18) Age category of trip AGE#D AGE2D (18< & <65) AGE2D maker AGE3D (>65) MAINA1D (full time and part time worker) MAINA2D (attending higher education) Main activity of trip MAINA#D MAINA1D MAINA3D (unemployed and maker keeping house) MAINA4D (pensioner) MAINA5D (other) EXP1D (groceries and food) EXP2D (clothes and Category of good EXP#D EXP1D household goods) purchased EXP3D (other)

Chapter 8 Shopping Centre Destination Preferences Modelling 212

8.4.1.1 Estimating the model fit and model significance for the developed model

As previously discussed in the methodology section, as additional terms were added, the performance of successive models was compared using the log likelihood function and pseudo-R2. The results for the 12 MNL models which were developed successively are reported in Table 8-7 citing the individual LL functions, the degrees of freedom (dfs), chi-squares and the measured -2LL improvements for comparison of successive models.

The results show a continuing improvement in the overall model fit as additional variables were added to the model, except for the addition of the population density, day of the week, trip maker’s age and children in the household attributes.

Terms representing trip distance, retail density, accessibility level by car, public transport, bicycle and walking, mixed use level and GLAR were all found to significantly improve the performance of the destination choice model. Interaction terms between distance aversion, the type of goods purchased and trip maker’s occupation were also found to significantly improve model performance. The pseudo-R2 of 0.7 to 0.8 indicates a very high level of model fit (Hensher et al., 2005).

Based on these results, the destination choice model, including all the additional attributes and characteristics and two alternative specific constants, was chosen as the best fitting MNL model to assess the impact of attributes on destination choice and also to identify the utility which trip makers obtain from visiting the three different size categories of shopping centres. Results will be discussed in detail in the parameter interpretation section. A full print out of model results from NLOGIT is included in Appendix 1 table A1-1 and table A1-2.

Chapter 8 Shopping Centre Destination Preferences Modelling 213

Table 8-7: Results for model fit and significance of the 12 successive MNL models, as additional terms are added

LL function Degrees of chi-square* -2LL pseudo- 2 freedom ** indicates a R change statistically significant between improvement in model models performance

Base model -1126.24 ------+ DISTKM -267.00 1 3.841 1718.47** 0.76

+ RETDEN -262.36 1 3.841 9.3** 0.77

+ POPDEN -262.31 1 3.841 0.1 0.77

+ PT -256.28 4 9.488 12.1** 0.77 +CAR + WALK + BIKE

+ ENTROPY -252.50 1 3.841 7.6** 0.78

+ GLAR -247.91 1 3.841 9.2** 0.78

+ WESD -247.62 1 3.841 0.6 0.78

+ NVEHSD -245.64 1 3.841 4.0** 0.78

+ KIDSD -245.21 1 3.841 0.9 0.78

+ AGE1D -244.77 2 5.991 0.9 0.78 + AGE3D

+ EXP1D -240.94 2 5.991 7.6** 0.79 + EXP2D

+ MAINA2 -236.94 4 9.488 8.0 0.79 + MAINA3 + MAINA4 + MAINA5 Each added row indicates only the added variables to the previous model

*Chi-square distribution for the indicated degrees of freedom and the p-value of 0.05

Chapter 8 Shopping Centre Destination Preferences Modelling 214

8.4.1.2 The Hausman test for independence of irrelevant alternatives in the developed MNL model

As described in the Methodology section, the Hausman test is used to decide whether the developed MNL conforms with the independence of irrelevant alternatives (IIA) assumption. If the IIA assumption holds, then more sophisticated models such as the latent class model or the random parameters logit model, which relax the IIA assumption should be investigated.

To run the Hausmann test and measure the test statistic (q), parameter estimates and the variance-covariance matrix should be developed for both for unrestricted (full choice set) and restricted (choice set modified with an alternative removed) models. Any interacting variables in the model must be removed before the test to avoid complications (Hensher et al., 2005). Consequently the unrestricted model for the Hausmann test contained only the two alternative-specific constants, distance, retail density, population density, car, public transport, bicycle and walking accessibility, mixed use level and GLAR. Restricted models were run with either the large-sized centre or the medium-sized centre options removed.

The Hausman test produced a large test-statistic (q=16.38) compared to the 5% p-value (Appendix 1, Table A1-3), indicating that the IIA assumption cannot be rejected and therefore that there is no need to develop models which are more complex than the MNL.

As mentioned in the methodology section, different forms of choice set were constructed, and a more complicated latent class model (LCM) was also developed. However, results from these alternative approaches did not provide any additional insights beyond those obtained from the (methodologically sound) MNL mode.

8.4.2 The model parameter estimates, utility and probability measurement

8.4.2.1 Parameter interpretation

The model results produce parameter estimates for the attributes included in the destination choice model. These parameters indicate how the various

Chapter 8 Shopping Centre Destination Preferences Modelling 215 attributes contribute to the utility which trip makers derive from their destination choice. However,

The utility derived from a choice model is only meaningful when considered relative to that of the utility for a second alternative. While the utility functions derived from a discrete choice model are linear, the probability estimates are not. It is possible to provide a direct behavioural interpretation of the parameter estimates when discussing utilities (although only in a relative sense) but not when discussing probabilities (Hensher et al., 2005).

This section therefore begins with an interpretation of the statistically significant parameters in the estimated MNL destination choice model.

As previously discussed in the choice set section, for the simple random sample destination choice set, the alternatives were randomly selected in each case. In order to check the model and ensure that the results are consistent, 15 different choice sets were created by random draws. The model was run for each choice set separately and the results were compared. This approach showed that the log likelihood of the models remained very similar across the different randomly constructed choice sets. To compare parameters from the models derived from the different choice sets, parameter estimates were scaled relative to one particular parameter (in this case, the distance parameter). This is necessary because the scale of the errors surrounding the estimated models varies as the choice set varies. These comparisons did not identify any major differences in parameters across the models estimated from the different choice sets, even though some parameters were only estimated to be significant at the 10% level (denoted as * in the results table). The results from one representative model are reported in this chapter and are discussed below.

The results reveal that travel distance is a major determinant of the attractiveness of shopping centres as destinations for trip makers. The relationship is negative, indicating that all else being equal, for all types of shopping centres, a destination become less attractive to shoppers as their trip distance increases. The model also shows that, ceteris paribus, the attractiveness of a destination increases as its gross lettable area-retail (GLAR)

Chapter 8 Shopping Centre Destination Preferences Modelling 216 increases. This result holds within the shopping centre size categories, as well as between them. Thus, bigger centres in any one of the size categories (small, medium or large), are more attractive, ceteris paribus. This presumably reflects the broader range of shopping possibilities which become available as GLAR increases. The accessibility of shopping centres via walking from a particular trip origin affects trip makers’ destination decisions. The longer it takes to walk to a particular size of shopping centre, the less likely a shopping centre of this type will be chosen as the shopping destination (even after separately accounting for trip distance as a separate term in the model). Once factors such as trip distance, GLAR and walking accessibility have been accounted for, no significant preference is evident for visits to a particular size of centre, as evidenced by the statistically insignificant ASC parameters.

This means that the utility derived from visiting the three sizes of shopping centre can be well explained by the observed influential factors which have been included in the model.

Socio-demographic and trip-specific factors, such as the day of the week, the type of products being bought, car ownership, having children in the household, age and occupation of the trip makers, were interacted with distance in the model. Several of these factors significantly affect the strength of distance aversion. Again, all things being equal, distance aversion reduces by almost 50% for trips to purchase clothing and household goods, compared with trips to purchase groceries and food. Distance aversion reduces by around 33% for trips to purchase ‘other’ categories of goods, compared with trips to purchase groceries and food. Considering this from the opposite perspective, distance aversion is much stronger for grocery and food trips than for trips to purchase other types of goods. This is not unexpected, as grocery and food purchasing options are probably easily replicated across all shopping centres, as all centres, even the smallest neighbourhood centres in the dataset, contain at least one supermarket. Hence, there seems little to be gained by travelling further than necessary on a ‘grocery and food’ trip. Where clothing and household goods are the target, however, a wider variety is likely to be available at the larger centres, which will typically entail longer travel distances. These

Chapter 8 Shopping Centre Destination Preferences Modelling 217 purchasing patterns show up in the results as decreased distance aversion for non-food and non-grocery trips.

The model results also show that while an aversion to travel distance is common across most occupational groups, distance aversion approximately doubles for students compared with the baseline ‘employed’ occupational category.

Contrary to what was expected from the preliminary analysis of the HTS in Chapter 5, although there is some indication that having children in the household increases distance aversion, this is not a significant influence on trip- makers’ destinations. This might be related to the fact that most of the trips are undertaken by car, so parents are not affected by having one or more children sitting in the car.

It was expected that more distant, larger shopping centres might be more appealing destinations during weekends; the model results indicate an increase in distance aversion, ceteris paribus, for weekend trips, significant at the 10% level. In several of the other model runs (using the different randomly drawn choice sets) this weekend parameter did not have a statistically significant impact on distance aversion. However, the parameter did consistently show a negative sign, indicative of increasing distance aversion, so the effect appears to be consistent – although difficult to explain.

Young or old trip makers might be expected to be more averse to distance than 'standard age' trip makers. The model shows this same result, although the increase in distance aversion is not significant. Householders which do not own a vehicle also show a non-significant increase in distance aversion, compared with car-owning households. This lack of significance might be explained by the fact that only 49 out of 1033 return trips in the dataset were undertaken by individuals from households with no car. This emphasises the high level of car ownership and car dependency amongst the residents in Brisbane.

Regarding the spatial characteristics surrounding the shopping centres - retail density, population density and level of mixed use - these factors appear to have no significant impacts on destination choice. However, all things being equal, retail density and mixed use level (entropy) both appear to exert a

Chapter 8 Shopping Centre Destination Preferences Modelling 218 positive, but statistically insignificant, effect on destination choice, whilst population density in the locality of a centre appears to exert a negative effect on destination choice. However, this effect appears to be statistically insignificant. While it might be expected that centres in more populated areas with a busier and livelier environment would be more attractive destinations for trip makers, the model does not detect this effect. The reason could be that shopping trips in Brisbane are mostly undertaken by car, and therefore the fact that more people live closer to the centre may not necessarily make a destination more attractive.

8.5 Conclusion

While a wide range of attributes including the individual attributes, trip attributes, destination and spatial attributes have been considered and estimated in this study using the MNL model, only a limited number of these factors are significantly affecting the destination choices of customers between large, medium and small size of shopping centres.

Among all different factors, distance and centre’s GLAR have come up as the most significant ones that attract individuals to a specific centre. Individuals seems to have a strong aversion to more distanced destinations and are attracted to larger centres providing more shopping opportunities for them. Other factors such as population density around the centres and mixed use are not significant indicators on people’s destination choice. Accessibility by different mode types did not come up as a significant indicator of customers’ destination decision except for the walking accessibility. This might be due to the fact that other mode type except for the private cars are not generally considered by customers as their trip mode.

Grocery and food trips are the only trip types that are significantly averse to the longer distanced trips. This is of great importance due to the findings from chapter 5 that almost half of weekly shopping trips are allocated to grocery and food.

The results suggest that changing planning regulations to facilitating the construction of more supermarkets across the BSD, might reduce the total

Chapter 8 Shopping Centre Destination Preferences Modelling 219 distance travelled for grocery/ food trips, given the strong distance aversion which surrounds choice of destination for grocery shopping trips.

Chapter 9 Professionals’ Insights on Retail Accessibility

Chapter 9 Professionals’ Insights on Retail Accessibility 223

9.1 Introduction

The discussions with professionals were done to achieve two goals. Firstly to find out if the findings from previous range of analysis based on the revealed types of data are in line with the ideas of professional people working within this context; and secondly to reveal their views on the realities and issues which are not accessible to us through the analysis. While it might be disputed that ongoing discussions around retail environments are occurring in the profession, there is hardly any evidence of investigation into the Australian retail context in the academic literature.

A number of major issues can be derived from these interviews, and these can be summarized as the most important concerns about the current and future development of the retail environment in Brisbane. The interviews reveal various issues which cause reactions and performance in specific ways, specifically in the current retail environment in Brisbane, as well as in economic enforcement, planning restrictions or promoters (catalysts) for retailers within the system. These findings are expected to provide a better platform on which to base proposed scenarios for the possible future of retail spaces, scenarios which may significantly improve our appreciation of the retail environment.

While some of these issues might seem to be irrelevant to the travel behaviour of customers, as the major focus of this research, they seem to be important in directing the current and future form and geography of retail spaces in the city. As a result, they have significant influence on the way people travel to and from them. This chapter will discuss issues such as the current planning system and the way it affects retailers’ and developers’ business decisions, the role and place of the planners in the physical formation of these destinations, their policies and the importance of their approaches, the feasibility of their plans for the city and the acceptance of the market of their values and ideas, and the incentives that are pushing forward the current form of the retail space and its possible future form.

The chapter is organized in the following way. The introduction will be pursued with a description of the methodology and includes some information about the interviewees, the boundary of focus, the designed interview questions, and the method of analysing the interviews. This is followed by a section identifying key

Chapter 9 Professionals’ Insights on Retail Accessibility 224 points, how they have been labelled and put together and a detailed discussion around these points. The concluding section summarizes the important findings, places them in major categories of research questions and provides brief discussions about each of them.

9.2 Methods

Ten interviews with highly ranked decision makers such as city planners, transport planners, urban economists and retail owners and within the public and private sectors were conducted in April and May 2014. The aim of these interviews was to obtain these decision makers’ views on the structure and synergies of Brisbane’s retail environment. The questions were concentrated on Brisbane City Council and its planning system. Interviewees’ ideas on the influential factors affecting the form and geography of retail spaces, such as urban/transport planning regulations, possible impacts of various factors on peoples’ transport behaviour, existing perceptions and views on the planning of retail spaces in the city, possible future structure of retail spaces, and views about a more sustainable retail transport environment were questioned in the interviews.

Each interview lasted for approximately 1-1/2 hours and ethics approval was obtained from Griffith University prior to conducting the interviews. The five interviewees in the public sector involved urban and transport planners and economists who were, or are still, involved in the process of planning and design of Brisbane’s previous/new city plan. The other five private sector interviewees were retail developers; those who have been long engaged in the preparation of an economic impact assessment for developers and who provide them with advice to enable successful proposals to the council.

As it was deemed to be less intrusive to the interviewees, a semi-structured interview approach was selected which encourages two-way communication between the interviewer and the interviewees. Since the participants’ backgrounds comprises various areas of economics, business, geography, transport and planning, there was a need for a more open and flexible method of interview which could provide opportunities for an informal grouping of the topics and questions and also asking

Chapter 9 Professionals’ Insights on Retail Accessibility 225 questions in different ways. Therefore, the method selected was the semi-structured interview.

The semi structured interview gave the researcher the chance to modify and tailor the questions based on the interview context and the interviewees’ backgrounds, rather than using the structured interview which has rigorous sets of questions and lacks flexibility. The interview questions were set in four major groups, which will be discussed below.

Each interview was reviewed and major points were summarized to help set up the later interviews. This was done in order to improve the questions if any points had been missed or needed more clarification or emphasis. In the next stage, the transcript for all the interviews was prepared and was carefully read to find any possible patterns, themes, relationships or sequences. Relevant discussions were put together, labelled (coded) and discussed and the key findings were specified for each section of the interview questions. Finally, the conclusions summarized around the four major areas of the interviews.

9.2.1 Interview questions

Fifteen questions were designed for the interviews, which took place with participants from both the public and private sectors. The questions were grouped into four main categories (Table 9-1). Analysis of the responses related to each category was undertaken, and separate discussions are provided below. These four categories included the planning regulations and processes for retail development, views and perceptions about retail planning, future trends and sustainability recommendations.

Even though there was a slight difference between the questions for the interviewees in the public and private sectors, the overall theme of questions was very similar.

Chapter 9 Professionals’ Insights on Retail Accessibility 226

Table 9-1: Questions for public and private sectors interviewees

Public sector Private sector What is your position/role in the council? What is your background in the retail planning field? Planning Regulations and Processes for Retail Development Planning Regulations and Processes for Retail Development How does Brisbane City Council plan for retail development in the What are the critical regulations and expectations that developers have to city? How is this represented in the plan? meet when getting planning approval for new retail establishments?

Does the proposal and approval process for new retail development Same question to the Public Sector differ from other development proposals (e.g., residential, commercial, industrial)? If so, how? Apart from the City Plan, what other guidance are planners using for Same question to the Public Sector retail space decision-making? Does the draft City Plan 2014 embody any changes in retail planning Same question to the Public Sector policies?

Prompt: Mini-retail hubs How do you perceive the in-centre policies such as those that were included in Brisbane City Council’s City Plan 2000? Same question to the Public Sector Prompts: have these had any impact on the retail location decisions?

Are they the best way to achieve such outcomes? What are the impacts on small businesses? What are the major issues affecting decisions regarding commercial impact assessment (CIA) for out-of-centre developments? Same question to the Public Sector Prompts: cost-effectiveness of the centre, means of transport, etc. How are trade areas of the centres currently determined? Views and perceptions about retail planning Views and perceptions about retail planning What do you see as the key challenges in planning for retail Same question to the Public Sector development?

Chapter 9 Professionals’ Insights on Retail Accessibility 227

Prompts: cost-effectiveness of the centre, means of transport, etc. How are trade areas of the centres currently determined? What are the critical factors in making retail location decisions, How do the planning/transport planning retail considerations affect finalising particularly for major centres? How important is planning? the retail location and characteristics decisions? Transportation? What role does transport and customers’ travel behaviour play in How do the developers consider customers’ accessibility requirements retail planning decisions? when they plan for new developments? Future trends Future trends What trends in retail development should be considered by the planners? Prompts: decreasing role of regional shopping malls, increase in role Same question to the Public Sector of district centres, niche lifestyle strips (i.e. James Street), e- shopping, bulky goods, changes in travel behaviour/car ownership, big-box retailers, etc. What is your prediction of the future hierarchy and distribution of Same question to the Public Sector shopping centres in Australian cities? If you had the ability to designate additional retail centres, where Same question to the Public Sector would they be located? Sustainable transport for retail establishments Sustainable transport for retail establishments What do you think could be done to encourage more people to walk, Do you think the concept of TOD followed by the council would be effective cycle or use public transport to access retail establishments, while and operational in making a more sustainable transport retail environment providing good economic returns to retailers and centre managers? in Brisbane Statistical Division (BSD)? Prompts: return of small convenience shops in suburbs, revival of Prompt: increase in population density in TOD nodes and the main streets, location of new development in transit oriented following traffic congestion, topography of the surrounding areas, etc. development (TOD) sites, limitations on growth of big-box, etc. What can planners do to deliver retail development that supports In what way could developers get more involved and also more interested in sustainable transport outcomes? the transport sustainability discussions?

Chapter 9 Professionals’ Insights on Retail Accessibility 228

9.3 Results and Discussion

The key findings from the interviews are discussed in this section. Following the question of the interviewees’ position or field of expertise, the first section focuses on planning regulations and processes for retail development (interview questions 2-7), while the second section is on the views and perceptions about the planning for retail space (interview questions 8-10). The following sections, three (interview questions 11-13) and four (interview questions 14-15), focus on the future form and structure of retail space and the issue of improving retail transport sustainability.

9.3.1 Planning Regulation and Processes for Retail Developments (interview questions 2 to 7)

- How well does Brisbane’s planning system work for retail development? Many professional people – planners, retail consultants and developers - questioned the efficiency of the existing planning system in Brisbane. The application of the ‘centre policy’ within the existing city plan, which divides the retail development into an already zoned retail precinct and out of centre development, has been backed up or rejected by various groups. Based on this policy, any retail development proposal, which is already zoned and consistent with the city plan, will need no more permission, however anything out of the currently centre-zoned area should seek approval from the council. This is the point at which the council requires an economic impact assessment (EIA). This assessment looks at the economic impact of the proposed development on the existing/future community needs in order to give permission for or rejection of the development.

The EIA looks into the planning influences of a proposal in the long term to investigate how the new development might impact on the long term existing or future plan. Finally, the change of land use or change of material application will go to the council for approval. The council will weigh up whether or not the arguments override the planning scheme (the economic benefits, etc.). If it is proved that there will be problems, a different approach can be made to solve

Chapter 9 Professionals’ Insights on Retail Accessibility 229 the issue, such as payment of compensation by the developers. Planning authorities refer to this process as being:

One of the most efficient systems to prevent the developers to overload the market and reflect the reverse and negative impacts of proposed developments on the environment and surrounding community which can end up as vacant shops and empty shopping centres similar to what happened in the US (Planner 1).

While planners believe that this process can reassure everyone that the impacts of something that is beyond what has been planned for will be negligible or minimal to the surrounding community before making any further decisions, developers are worried about what they call the “detrimental impacts of the procedure on the market and retailers” (Developer 5).

Developers believe that this policy imposes many limitations on the existing opportunities in the market for activities which cannot be located in the centre, either because there is no spare land for them or because they have not proved to be necessary to the community. However, from the perspective of the developers, they are identified as being a market opportunity. The centre policy forces proposals to go through the rezoning process. The proposals may be successful in many cases, but will impose a great deal of pressure in terms of the amount of time and money spent by their instigators.

As stressed by the developers interviewed, the unzoned or non-commercial zoned land which is proposed to go through the rezoning process can also bring third parties and their submissions against the proposals. For example Developer 1 mentioned, “Westfield and its ambitions for not having any competition in the field”. It is believed that in the current environment, “Westfield objects to almost anything that moves” (Developer 1). Westfield has plentiful funding and tries to protect its shareholders and their investment. A third party’s submission against a development can end up in a court appeal, which increases the time frame of a development by more than nine to twelve months, resulting in a much higher cost for developers. Many developers believe that there should be statutory frameworks to limit third party complaints and cut the complications of the impact assessment process.

Chapter 9 Professionals’ Insights on Retail Accessibility 230

Developers also believe that while timing is essential for any developer to survive and be able to continue with their business,

Planners do not appreciate this concern and they push the proposal into long encumbering and delaying processes which can end up in withdrawing the proposal or too much waste of time and money for the developers (Developer 3).

This also results in strengthening the current problem of empowering omnipotent enterprises such as Westfield, allowing them to become even bigger and more powerful. Small businesses cannot afford to hold their property/money/assets, etc. and wait for the planning authority or court decision to enable them to go ahead with their plan.

Referring to a recent study by the Australian Institute of Urban Studies, one of the developers mentioned that:

The average development timeframe to get a major project off the ground, to identify the opportunity by the developer, to get it together and to get it through the planning process to the time that the project actually kicks off is almost 10 years (Developer 5).

This significantly affects the developers or the owners of the land by keeping the land unproductive during this period. In most cases, smaller businesses find it impossible to cope with these delays.

- Does the planning scheme have the capacity to guide the future distribution of retail space? The private consultant strongly believes that “no planning scheme” can predict the future distribution of centres with any accuracy within the life of the plan which is usually about twenty years, particularly for small centres such as convenience based centres or the ones that are situated around big supermarkets such as Coles, Woolworths or super IGA stores.

Planners cannot predict at the micro level, they might get the big things and that is why we (consultant companies) are part of the process of getting these centres where they are needed, which necessarily may not

Chapter 9 Professionals’ Insights on Retail Accessibility 231

be where the planning schemes intended them to be or just haven’t considered it (Developer 2).

While the role of the planners and planning schemes is identified as being to anticipate where future development should go, it is assumed that, at a strategic level, they might only be able to make predictions for big centres such as Northlakes, Springfield or next centre in Ripley Valley in Ipswich. After the council designates where the centre should go, even in these cases, developers believe that it is not possible to foresee whether the centre ends up there or not.

Centres’ locations are ultimately determined by who owns the land not the planners (Developer 3).

Therefore, it is the role of the market to eventually sort it out. The “entry sequence” of the centres plays a big part in what is happening in the market. The first centre to be developed will have a bigger market advantage than the later ones. If a centre is well-located, it is harder for a later retail space to come in, since there are usually minimum distance requirements that the retail spaces try to keep between themselves.

If there is only one owner and a master plan from the beginning, it might be easier to envisage the future, otherwise in many other cases the planning schemes come up with sets of town centres and a bunch of other types of centres in their plan and only a few of them might be built in the same locations. The example is Northlakes with almost 35 identified possible sites for retail but only about 6 were built in the end. So there are probably more centres either proposed or having some preliminary approval than the market can actually support. So it has been done sort of piecemeal (Developer 3).

What is strongly believed by the owners, the developers and the investors of future centres is that the council cannot get everyone to follow what they plan for in their city plan or any other planning document. The market is looking for the opportunities that can assure them of expanding their profits, and developers will go ahead with proposals that are in line with their objectives.

Chapter 9 Professionals’ Insights on Retail Accessibility 232

- Differences between City Plan 2000 and the City Plan 2014 for retail planning and development While for almost 14 years the City Plan 2000 has been planners’ major reference in how the retail environment should be directed, the new City Plan 2014 came up with a number of major differences in terms of the existing understanding of the retail environment and the way it works. These changes are believed by city planners to significantly affect the future of retail spaces in Brisbane.

What the new city plan recognises more than the previous one, and is probably one of the most important differences between the two plans, is that Brisbane’s future in terms of its development and its functions will be around corridors. These corridors are based on mass transit where there are railway lines, busways or major roads. The number of centres, of different scale from regional centres to small neighbourhood centres, will help these corridors to function in the future, as opposed to previously being separated dots on a map.

The planning intent now is to recognize that we do not develop any more around dots; we develop in corridors which are accessing multiple things and are linked on the transport routes (Planner 3).

A “revolutionary change” mentioned in the planning scheme is that the council will no longer ask for a commercial impact assessment. This means that planners will no longer ask a developer, an owner or a business to justify why a new shopping centre, a grocery store or an extension of the currently existing centre/store is required in a location. The shift is in the fact that there will be no more questions about the appropriateness of the type of retail space that the developer is proposing, instead the focus will be very much on planning outcome.

How we get to that planning outcome is by locational attributes as opposed to asking applicants to justify why there can be two supermarkets in that locality. As long as it can be proved that the centre is in close proximity to public transport, is not bigger than a specific scale and is following the overall council parameters, the developer can build their development wherever they like (Planner 1).

Chapter 9 Professionals’ Insights on Retail Accessibility 233

These changes are being seen as the planning responses to the market not being capable of identifying certain retail functions, particularly around the out of centre retail spaces. Therefore, if the developer cannot fit his development in the currently planned locality, he can much more easily than previously apply for a new location.

Beside the two major shifts that have been discussed above, there are major additions to the strategic planning framework. These additions relate to an understanding of the dynamics and the economics of the city, the way the city develops and the way things are located in certain areas, as a result, how the planners are trying to respond.

Other issues, such as the colocations of types of uses and the clustering issue have also been investigated such as a colocation of a hospital with its large workforce, with shops to cater for this workforce, or major industrial areas such as the airport, where visitors need access to food outlets, a post office, or a bank when they are on a lunch break. Where there is a specialized centre with high numbers of workers, there is also the need for workforce amenities. The former planning scheme was not as strong as the current one, and therefore different uses were more separated. If the area in question was industrial, no provision was made for retailing function.

The new city plan has recognized that workers and companies require access to the facilities that serve their population. This is one recognizable change in the distribution of retail activities, considering them not only as separated dots on a map called centres, but also as having the ability to spread wherever it is appropriate.

Large format retailing, as the fourth major element of this shift, is another dimension planners recognize within the retail environment, which has ended up in the new planning codes. Large format retailing, which first appeared in the form of big warehouses such as Bunnings, came into the retail market around 10 years ago. The goods these warehouses sell are on a large scale and can include hardware, timber, homeware appliances, food, etc. Another example of this type of bulk retailing is the Costco store, which has recently arrived in the South East Queensland area.

Chapter 9 Professionals’ Insights on Retail Accessibility 234

The likely trips generated by this form of retailing would be car based. When people are making these purchases, it is unlikely that they would catch a bus for it. At Costco you might buy a pallet of something because it is cheap to buy in loads. But you are not going to buy a pallet of toilet rolls and carry them home on a bus (Planner 4).

What is being appreciated in the new city plan is the fact that wherever these bulk retailing centres are located, their positioning is different to where other types of retailing, such as supermarkets are positioned which support multiple trips with possibly small bags that can be carried home on public transport.

The new city plan understands the differences within the retailing market and understands that the way people use the retail services differs depending on whether they access it by public transport or private vehicle. Therefore, these high car-based retailing centres are not placed near already congested roads, or roads that would be ‘chockablock’ (highly congested) in the future. These sorts of high format retail spaces prefer high visibility, access to major roads and operate at a scale with a region-size catchment area rather than at a suburban level. The new city plan shows appreciation of these issues from a market perspective.

9.3.2 Views and perceptions about retail planning (interview questions 8 to 10)

- Key factors affecting planners’ decisions on retail development inquiries What planners generally do is to look at the planning arguments for allowing the material change of use (e.g. from residential to commercial) or even changes in the scale or characteristics of the existing land use (like a change of a planned local shop catering for local catchment to a big Woolworths plus other specialty shops). In terms of retail types of use, both planners and the council try to ensure that any proposal coming through the planning system is not going to have a significant impact on other retailers in that area. They also try to clarify the reasons behind the new proposal in a specific location compared to somewhere else in the city that has been planned for the same purpose. The council asks for certain tests to be done by the applicants to prove that their

Chapter 9 Professionals’ Insights on Retail Accessibility 235 proposals’ impacts on the community are acceptable, if the centres’ characteristics go beyond what is thought to be appropriate within that locality. These tests are also asked for if the characteristics are considered inappropriate or not aligned with what it has been planned for, for example in the sense of scale.

These factors are investigated using the Economic Impact Assessment report, which is required by the planning authorities to be prepared by the developers. Generally, the report looks at different aspects, including environmental, noise and traffic impacts, the catchment area of the centre, the socio demographic characteristics of the surrounding population and the competition within the surrounding environment. The report also looks at where other shopping centres in the city are, how much trade they are likely to have in the future, and if the development goes ahead, will it have a negative impact on the performance of other centres.

The traffic studies from the impact assessment report look at a longer (a ten year) horizon in the area and tries to ensure that the plan is not going to mess up the long term traffic planning in future. If it does, the proposal will be stopped or the developers will have to pay for improvements like putting traffic signals, turn lines, etc. (Planner1).

While the accessibility of the centre is considered to be one of the major issues investigated in these reports, they mostly focus on car accessibility. In large regional centres, public transport accessibility is being only partially investigated, and the focus is still on private transport.

When it comes to the physical characteristics of the centre, planners refer to the Ground Floor Area (GFA) as the first factor that is being looked at whenever a planning application is pending. Based on the GFA, the various types of trips made by people to this kind of land use are being estimated. If it ticks the planners’ box, then it will be followed by other design parameters around car parking provisions. A 1000 square metre of retail space might probably be equal to a potential provision of a minimum of fifty car parking spaces. This is simply the way that planners try to manage the transport impacts on retail

Chapter 9 Professionals’ Insights on Retail Accessibility 236 developments. By providing the appropriate number of car parking spaces, people who are going to drive will not park on the roads.

Another issue that might be looked at is the positioning of the building and how activities are being engaged with the street or the environment around it. These considerations are more in the area of the urban design focus, such as the preference for having the car parking at the rear of the building.

In their submitted proposal the developers also need to illustrate ease of access, or where large trucks might come when making deliveries and other general things that might be asked for from a development sense. To summarize, the current city plan with respect to retail space and how it should be developed, concentrates on the provision of floor space. The distribution of retail space and the centres’ network ranges from one that is serving a local catchment to areas that are bigger are in nominated areas throughout the city. These existing areas have a preferred development of retail or any sort of employment function, for example, within commercial zoning. These areas are usually placed near existing shopping centres. Apart from the floor space area, planners are also interested in making sure that car accessibility to the centres is easy. None of these criteria truly investigates the sustainability solutions for retail space and all are creating a trend towards a more unsustainable environment.

- Critical factors influencing developers’ decisions about the location of retail space Considering the different factors which affect the developers’ decisions about the location of retail spaces, many reasons can be recognized, some of which seem to be more influential than others. One of the most important issues noted by almost everyone within the private sector is the hope that the land use planning policies and regulations will have some ‘market realities’. Interviewees raise the issue that, because of the crucial factors about financial feasibility, yearly turnovers, etc., retail locations and characteristics cannot be simply decided by planners. They believe that these factors are not being appreciated by planners.

Chapter 9 Professionals’ Insights on Retail Accessibility 237

Developers will go ahead with a proposal if it is believed that it will be profitable. “Retail is not a public service”. What developers feel is underestimated by policy makers and planners is the fact that developers have to weigh up different factors every time a new proposal is put forward. These factors include the cost of land, the expensive application fees, the consultant fees, and those such as the presence of a third party submission against this development that can end the proposal in the planning environment court appeal. So on the developers’ side there is much to be considered.

Having a third party against the application is a big expense in itself. The bigger the centres, the more controversies are around them. All through that period developers are paying taxes or option on the site and the land is unproductive, which means racking up a lot of costs until the final decision is released (Developer 3).

Retail developers are usually very protective about their retail surroundings, which means that they tend to oppose to each other when other retail developers are trying to expand or build a new site, particularly if the new site is on land that has not been designated for retail use.

Even if a developer decides that there is enough catchment to put a shopping centre in greenfield land close to some other retail space, they will have quite a hard time in establishing a new centre, firstly because of all the regulatory processes that they have to go through and secondly because the nearest shopping centre will then do everything it can to slow things down or prohibit them (Developer 1).

Time is of the essence for developers, but most city planners do not appreciate the cost and the risks involved in these projects, since they are not the ones who are taking the risk. While big businesses might have the power and money to wait for development approval, it is often not possible for small developers to hold the land and wait for this approval.

When it comes to the accessibility and the form of the centres preferred by developers, it is all about convenience and exposure (visibility). What all developers want is a highly visible location on a main road with enough space for parking so that customers can get in and out quickly. They are also looking

Chapter 9 Professionals’ Insights on Retail Accessibility 238 for proximity to other likeminded businesses, trying for a degree of clustering, in terms of where people choose to locate their businesses. Developers try to choose a retail location that gives them the highest level of profitability and this is generally reflected in the size of the catchment.

Other factors such as integration with public transport and the surrounding land use are increasingly becoming important when it comes to larger centres or centres which are part of a town centre. These factors are referred to as the “success factors”. For smaller neighbourhood centres, the decision is more about profitability but for the larger centres, it is about integration and more incorporation with other uses within larger regional shopping centres.

A good example for planning a large shopping centre as part of a town centre is the Westfield Coomera. As a new regional centre close to the train line, one of the key objectives is to link the shopping centre with the train station and with those other land uses. It isn’t just a shopping centre; it is more of a place for doing the shopping, maybe dropping kids at childcare, doing an education course, etc. It is a vacant site at the moment with the approval for the first stage of development up to 25’000 sq.m. But ultimately it would be 60’000 to 70’000 sq.m. and will have the surrounding land uses that complement the retail centre. So this is the way that developers look at their projects and plan for them. It is not just a one-off development but there is usually a much bigger picture of development in the close future to increase the benefit and expand the power and catchment of the centre (Developer 5).

What can be concluded from the developers’ concerns is that a conglomeration of these problems means that developers usually end up trying to find land which is already zoned and it is quite hard to get that land anywhere else. Therefore, it is normally much easier to look at the issue of convenience before considering other factors.

- Why the developers are expanding their existing centres rather than establishing new ones Looking at the current retail environment in Brisbane, what is quite obvious is the ongoing expansion of large regional centres (for example, Westfield

Chapter 9 Professionals’ Insights on Retail Accessibility 239 shopping centres) which has been going on for at least the last two decades. How can these developments be justified and are they going to continue in future?

What is believed by developers to be an important factor is that “Westfield group expansion in centres is a reflection of the market”. Westfield gets its returns from management fees. All the managers and asset owners are under pressure to build their portfolios because they get their management fees on their entire portfolio, and most of the major retail owners make a lot of money from management fees. The higher the returns from management fees and the property they are managing, the higher their profits.

There are generally two ways for the businesses to expand. The first is to buy a new centre.

Currently, if developers want to buy something with a very good quality they are going to buy another 6 percent yield so it is going to cost them a lot (Developer 2).

The second way is to expand the existing centre. If an existing centre such as Carindale is going to be expanded, a good return is already being created from what has been done so far. Part of that return might be the early mover advantage or the internal rent. So by expanding an existing centre, a powerful asset is made even more powerful and more resistant to competition by making it even larger and more dominant in its catchment. This is the strongest incentive for expansion.

The large centres’ annual reports here in Brisbane already show that they are getting really good returns on their extension projects (Developer 4).

Expanding the existing centre means expanding the existing site area, requiring money for the expansion and the major players to invest in it. These centres are usually in locations that allow them to draw from a very large area. In addition, the construction costs have become a bit lower during the last few years, and it would cost substantially more to do these extensions if the market was strong

Chapter 9 Professionals’ Insights on Retail Accessibility 240 rather than being weak. The centre management can afford to put in tenants such as department stores, which draw shoppers from a very large area.

Up until Carindale only the CBD had Myer and David Jones (DJ). Prior to that Westfield only had a Myer or a DJ now they are all getting both of them. So that is one reason they are getting bigger. They are approaching their peak at about 150’000 because they have to add both department stores. Their flagship department store in the city is 30’000 sqm and the ones out in the suburbs are usually half that like 15’000 or less.

Brookside for example has the department store Myer but they don’t have the location and accessibility to grow to that size.

Chermside and Toombul are very close and Toombul doesn’t actually have a trade area to the east so it basically has a very limited trade area and that is why it lost its DJs because it is not doing well particularly because of Chermside with 150’000 sqm.

Strathpine used to be the biggest centre in the north side until Northlakes opened and they closed the Myer and moved it to Northlakes.

Capalaba has never gotten a department store, one was proposed many years ago but it never happened.

Browns Plains is a well located centre that may get a department store one day and there is a lot of growth that has started to occur in this area.

Chermside, Carindale, Upper Mount Gravatt, Logan Hyperdome are all having department stores. These examples all show the importance of the location of the centres (Developer 3).

For all the above-mentioned reasons, it is strongly believed by the developers that retail space is always being reinvented and must keep the assets fresh and new to draw people in. Without this, the centres are unable to maintain their success.

Chapter 9 Professionals’ Insights on Retail Accessibility 241

9.3.3 Future trends (interview questions 11 to 13)

- The future of shopping centres and possible shifts One of the most important questions about the future of retail space in Brisbane is about the way shopping centres will continue to perform. Will they continue to grow in size? Will they be replaced by other types of retail formats? Discussions with these professional people show almost the same trajectory for the future of these major retail hubs. While most opinions are that there is a significant role for different types of centres in the future, some believe that development of these centres will not happen to the exclusion of other types of retail spaces, including smaller retail developments or online shopping. The key ideas and perceptions around these issues are discussed below.

While the predictions for the future of retail space does not seem to be an easy one, it is expected that there will be expansion in size, number and types of services being provided in larger regional centres. This trend has already started with noticeable expansion in the size of the big centres such as Garden City, Carindale, Chermside and Indooroopilly, through various stages during the last few decades. There is no expectation that there will be new regional centres in new locations since these are not easy to justify. Brisbane is a fairly small city. The large regional centres which are already dominant will continue servicing their particular catchment areas and are not going to be challenged. However, the situation could be different outside of the Brisbane Statistical Division. New regional centres might form on the Sunshine Coast or the Gold Coast, as has already occurred in some locations like Helensvale shopping centres in Gold coast.

These new regional centres would provide a large number of uses, including for retail space and all the ancillary uses which attract people. Regional centres have changed their strategies and due to the appearance and fast expansion of online shopping, have been forced to adapt and add things to their centres that have attracted more people during the last few years. In order to entice people to the centre to shop and spend money, they need to provide things that they did not have in the past, like entertainment, bowling alleys, big eating precincts and more attractions related to leisure and lifestyle rather than just traditional

Chapter 9 Professionals’ Insights on Retail Accessibility 242 retail stores. There are plans that show residential developments, which potentially go hand in hand with regional centres, whether in a form of a residential town on top of the centre or a building next to the centre. These activity centres will include and support various sorts of land uses and retail spaces to go with it.

On the other hand, most of the development activities around Brisbane are happening at the small centres on smaller land areas like supermarket or convenience store based centres. What is obvious is that the number of small centres is increasing in Brisbane and this trend is going to continue. In the future, there will be more sub-regional centres, larger neighbourhood centres, or supermarket centres that have maybe two supermarkets as standard and perhaps up to 10 to 12 shops and this number will grow over time. Sub-regional centres will still be important because they do play a role for those people who do not want to go to the larger regional centres or smaller supermarkets.

Smaller neighbourhood centres (with a Coles/Woolworths) are going to increase in size up to 4,200/4,500 m2. What they have already started to do and are planning to continue to do in the future is to add a bit of clothing for women and children, more ready to eat, bigger delicatessens, pre-prepared meals and more general merchandise on health and beauty. Small neighbourhood centres are trying to get bigger and squeeze out the specialty shops. Some specialty shops, like newsagents for example, are finding it harder to get in to the centres because the supermarkets are taking over more of the functions of the specialty shops.

The overall centre size is falling. There are lots of 4,500 m2 centres, with like 3,200/3,500 sqm Coles and 600/700 m2 specialty shops where before you needed a lot of specialties, because Coles and Woolworth would only pay for 100 m2 for their brand but now they are paying much more of that, so the dynamics has changed, that is because of the competition, they are fighting out there to get the sites, to grab market shares so they will pay more (Developer 4)

As discussed above, the large retail centres will continue to perform. They will be presented as being more experiential and more about lifestyle. Where

Chapter 9 Professionals’ Insights on Retail Accessibility 243 shopping centres previously appeared to be ‘big boxes’, they have begun to develop along their edges to provide more places where people can have lunch, see a movie, and they have a collocation of facilities such as libraries and gyms, etc. This creates a village style multi-use function, which is what high streets used to do, which is to create an experiential, pleasant environment.

Garden City was the first centre to do this and called it a ‘town centre extension’. The Garden City redevelopment was closely followed by redevelopments at Indooroopilly and Chermside Westfield. These sites are usually very well designed landscape focused developments, very considerate of urban design amenities and mostly focused around public transport hubs. What previously was more a closed box with everything internalized has started to become more open, with extensions to a more pleasant surrounding environment. The expectations of retailing today is based on the experience and to make it a joyful experience for the shoppers.

One of the strengths of large centres like Garden City has always been the fact they were anchored by one or two department store, such as Myer and/or David Jones (DJs). These department stores have always been the generative traders; Myer and DJs in higher order centres, down from that were the discount department stores like Big W, Kmart and Target and below them the major supermarkets. However, as the highest order retailers, these department stores have lost a lot of their power over the years. Usually the department stores are only placed in shopping centres with a population catchment of 150 to 200 thousands because in terms of their profitability, that is their benchmark. They used to be the anchor tenants, allowing the specialty stores and everything else to feed off the traffic that they generated.

What has happened, especially within the last two decades, is that specialty shops, particularly fashion and homewares, have become able to do much better than department stores. So now, in a way, the department stores are depending on the traffic that the specialty shops generate. They are not as big an attraction as they were in past years. Department stores used to carry a wide range of white goods and appliances, but that has changed, since retailers like Harvey Norman and JB HiFi and others can do it a lot better. Myer and DJs are

Chapter 9 Professionals’ Insights on Retail Accessibility 244 still recognized as prime traders but their power has diminished. In light of this, perhaps someday some major shopping centres may operate without department stores as the key anchor. It is possible that there might be some other uses, which are better at drawing people such as entertainment, and this will considerably affect the importance and role of large centres in the near future.

It is important to consider the rate of growth of these large centres. Economists usually look at the rates of growth of different types of things in order to make their judgements. When it comes to the major regional centres like Chermside, Garden City, Indooroopilly and Carindale, these centres already exist and they will grow. But in terms of preferences we might see that while they do grow, it might not be at the same rate as other smaller sub-regional centres or even the neighbourhood centres, with a possible higher growth ratio over the next 20 years. This means that we will see more of the sub-regional centres. The higher rate of growth is reflective of the move of the preferences of retailing function. While regional centres like Chermside will continue to grow, the rate of growth will be lessened.

One of the reasons behind this different growth rate ratio is the fact that the big centres are servicing a mega regional catchment area and right now the accessibility to those centres is quite easy. As more people move into the area, it becomes more difficult for them to access these places and the preferences will change; why would anyone go to Mount Gravatt only to buy one thing? To buy just a few items, people will travel to a smaller store. People are shopping more frequently and what they are buying is less in volume, which means doing more daily trips. If it continues, this retailing behaviour is more conducive to smaller stores. In other word, it would be difficult to have that behaviour in mega regional stores. Intuitively consumers are not comfortable finding a car park and walking for 600 metres through a shopping centre to buy one product. Therefore, the faster rate of growth of neighbourhood convenience stores and suburban centres will become noticeable.

Other types of centres will still exist in the future because they perform a function. This function is that if people require multiple items with a greater

Chapter 9 Professionals’ Insights on Retail Accessibility 245 choice in where they can select those items from, that is exactly what large regional centres do, and there will always be a demand for them.

- The role of other types of shopping establishments in the future The next question is what will happen to other types of retail spaces in the future? Are the high streets, which in many cases are struggling, going to disappear? Or there is still hope that they will continue to function, or maybe even to expand in the future?

Planners’ expectations for the future are that there will be great demands for other types of retailing which offer very different experiences for customers. More prominent in cities such as Sydney are traditional types of high street retailing such as a butcher, a baker, a restaurant or café and a fashion store, and these are on the rise in Brisbane.

Even though big centres provide customers with all required services and products, customers are looking for more varied retail spaces. What we will see in the future will be more alternatives, including the high streets, and the smaller retailers. This follows the above discussion regarding retailing experiences, which are growing more quickly and are becoming more evident than ever in the city.

While there are good prospects for growth for these centres, the strip shopping centres cannot be relied on in the large dispersed suburban areas. The only place for these types of developments might be in the inner parts of the city with much more highly populated areas. The outer suburbs cannot afford to have different types of retail outlets because all of the specialty shops are located in large centres. The only way for these street shops to grow in the suburbs might be the ‘eat street’ concept or the food and beverage shops located along streets. These strip shops have started to work well in the inner Brisbane suburbs and will continue to get bigger and stronger.

Despite the fact that the concept of strip shops might not be a prevailing concept of the future of retail in Brisbane as a separate existence, it has started to appear in a different form, strongly linked to the big shopping centres.

Chapter 9 Professionals’ Insights on Retail Accessibility 246

9.3.4 Sustainable transport for retail establishments (interview questions 14 & 15)

- Transit Oriented Development (TOD) plans and its application for Brisbane One of the concepts to achieve a more sustainable future for Brisbane, which has been strongly proposed in the existing planning documents such as the City Plan and the infrastructure plan, is the concept of transit oriented development (TOD). However unfortunately, not too much has been happened when it comes to practice. Discussions with the people in the profession have raised a number of issues. It appears that, in Brisbane, the concept of the TODs has become a slow, or even an unfeasible, process.

The issues of TODs are mostly on the ‘critical mass’ and the ‘required density’. TODs work perfectly when there is critical mass.

Critical mass means when there are enough people living in an area, you will start getting more of TOD behaviour simply because there are enough people on the street to be disturbed and frustrated by the traffic and therefore there is enough people to justify the frequency of public transport (Planner 1).

Some planners are quite optimistic about this happening in Brisbane in the near future, once the city develops, the population increases and there are more major new dots (supporting the concept of corridors) in the city. On the other hand, many developers find the concept of TOD meaningless, unreasonable and not feasible with the current population growth rate and the standing of the market. Economists mostly believe that,

It would be naïve to say that council TOD plans will exactly develop to the intended extent (Developer 5).

Developers believe that TODs will happen in Brisbane when transport becomes a key point of interest for developments and where people prefer to live close to transit routes as opposed to living in big detached houses in dispersed suburbs. That is the reason why Brisbane is still far away from that required density and this will leads the city to develop in a different way (Developer 1).

Chapter 9 Professionals’ Insights on Retail Accessibility 247

This means that while the idea of TOD has been around for quite a number of years, the market has not yet reached the concept. This is still a major issue in Brisbane but might be less of a problem in the future.

An area can always be zoned as something to be fifty stories or more but the developers that might come through only see it as a ten stories potential because the market is not demanding it (Developer 1).

Developers cannot hold the land unless they own it and can afford to land bank the prime site for that period of time. This is only the case when they are a big player such as the developers of Northlakes or Springfield. Where the owners for some reason - have a house, the land has been paid for, and there is no mortgage – are happy just to hold their land and let that land be saved for future development. These cases are quite rare and random and cannot be the basis of future development plans.

Good examples mentioned by a number of interviewees are suburbs such as Milton which is close to the railway line. In the 1990s, the owners bought and put aside a number of lots in row and let the previous office businesses run in smaller old existing buildings because they produced an income. When the SEQ plan identified Milton and Milton railway station as a place for high-rise development to occur, these owners sold their properties to a developer who amalgamated the sites following council’s announcement to allow twenty storey building developments. This process took almost ten years.

For a privately owned piece of land, the owner has the right to develop and he is going to develop whatever he thinks can be profitable for him/her in the short term. This is what is happening to most of the TOD plans. The owners have to sell off their high-rise planned lands for normal housing and the TOD plan will disappear. In a few years’ time, it could have been a great plan, but when the land is developed, it is gone and it is not possible to tear down new houses to put up a higher building any time soon. Timing in these developments is essential and developers believe that planners have no appreciation for timing.

When the market calls for townhouses and the council is unable to make the owners develop them to the maximum, the only way for the council to get the required multi-storey building is by buying and keeping the land for development

Chapter 9 Professionals’ Insights on Retail Accessibility 248 sometime in the future when the market will in fact support that type of development. However, in most cases, the council does not have the funds to do that. Council has done this in the past, and it worked perfectly, for example in the case of the Ipswich CBD where the council bought the land for commercial development. This was a proactive plan.

The deadbeat Taiwanese centre owner of a piece of land in the CBD was offshore, he bought it years ago and put no money in it so it was really dragging the city down and the only way to get around that was for the council to form a development corporation, buy the land and then the state government helped them revitalize the whole of the CBD. They just opened the first 14 storey high-rise in the CBD and forced a bunch of public servants to move from Brisbane to Ipswich (Developer 2).

In most cases, what will happen as a result is that the high-rise and TOD plan opportunity is lost, not because of poor planning, but because the economic cycle is not in the correct phase.

Apart from the non-planning issues mentioned above, there are other concerns mentioned both by planners and developers that are working as a hindrance to the formation of TODs in the city. What is currently happening is that most of the retail developers are developing retail spaces, the residential developers are developing apartments, and the two of them do not mix. Therefore, as soon as mixed use development is asked for on a site, other challenges arise.

Developments such as TODs have not happened in Brisbane yet because there is no delivery framework for them. Developers require the planners to consider issues related to the development of the mixed use hubs. If planners want a place to be developed for a retail space, they should make it as easy as possible for people to do this. By putting the developers through very long complicated process and asking for too many things to be done, TODs will not have the opportunity to be developed.

Another problem from the commercial perspective is the cost of these types of developments. Once a hard edge boundary is drawn around a centre, the value of that land will immediately be uplifted because shopping centres or shops or offices have more value than residential development. In addition, the cost of

Chapter 9 Professionals’ Insights on Retail Accessibility 249 development near railway stations again is prohibitively expensive, making it totally uneconomic for developers.

In reference to the discussion above, what is important to be considered eventually for a TOD development is that the density of population around the centre has to be increased at some point. It is not a matter of if, it is a matter of when. There is no other option but to increase density around existing infrastructure. The way Brisbane has developed up until the current time is that ‘it has gone out’ and now it is being seen that ‘it will go up’. People have already started to make trade-offs. They ask themselves if they would prefer to live far out in the suburbs where there are no shops, no community services and it can take an hour to get to work. Would they sacrifice a bigger house to live in a growth corridor in a smaller house with shops close by and access to work by public transport within twenty minutes? People have started to appreciate the accessibility issue and it is higher in their agenda.

Recent planning intentions are recognizing the fact that the city is not developing around the dots (activity centres, shopping centres, etc.) anymore; rather it is developing along the corridors to access multiple sites and be linked on the transport routes.

Going forward, the market will be far more responsive to max out. As previously mentioned, the new city plan and the realization of the role of the corridors, including railways, busways, etc. as the future growth potential is of great importance for the future of TODs in Brisbane. However, it should be remembered that it often takes a couple of generations for such things to come to fruition.

Sydney and Melbourne with much larger populations than Brisbane, have also been through this phase of critical mass. A city like Sydney with a population of almost five million people could potentially support TODs. Retail spaces centred around the train stations works really well there, because Sydney has the population density and the sheer number of commuters that keeps these centres profitable. If only one in a hundred people goes shopping in the centre, the centre will be viable, but when the population scales down to the level of Brisbane’s population, it does not work so well.

Chapter 9 Professionals’ Insights on Retail Accessibility 250

Examples of TOD attempts in Brisbane

While there are serious issues around the future TOD developments as mentioned above, some examples of this new way of urbanism have already been witnessed in Brisbane. As spaces to market and to sell, new developments within the inner city are usually supported by increasing access to public transport. This can most noticeably be seen with the city glider routes, the blue buses that travel from West End to New Farm to Teneriffe. This high frequency and highly visibile mass transit is being used to sell properties in these areas. These are the areas where the densities of work and housing make the high frequency transport a viable option. These are the examples of the future expectation in terms of how Brisbane will develop. The city will be developed around those corridors and people and the market will choose to live on those development corridors, as their priority will be accessibility.

The suburb of Milton is another good example of existing TOD development. This suburb contains the first apartment building in Brisbane that is built on top of the railway line, and it is likely that the population of this suburb will rapidly increase in the near future. The reason is the proximity to inner city neighbourhoods where the convenience of retail spaces, close by restaurants and transport accessibility that goes to where people work or want to visit. At least in the inner city, the TOD principle will start happening a lot more and be far more visible. In the future, it will start to creep along the transport corridors. Other successful examples are the suburbs of the inner west, Indooroopilly, Taringa and Toowong.

While there are quite a number of success stories about the TOD developments in Brisbane, some of these plans have not been quite so successful. For example, in the case of five storey developments alongside the railway line in Northlakes, which was later changed back to townhouses, because these units had to compete with the market around them, consisting of single dwelling lots and suburban lots. If for the same amount of money people can buy a house with a garden rather than a unit, then an apartment or a high-rise is are not going to be the choice. The five storey development might be a great plan perhaps in 20 years from now, but there is no market for it at present.

Chapter 9 Professionals’ Insights on Retail Accessibility 251

Another example is the original plan for the Moreton Bay rail link, which is the new railway going to Redcliffe. The six stations included in the plan had been planned as TOD developments and that is how they got federal funding in the first place. It was predicted that there would be a large population growth, centre activities and economic activities around these six stations. Later the local area plan showed that these plans were not going to work simply because of the unrealistic predicted densities around these suburban stations.

Another good example would be all the developments along the railway line and Coronation Drive and Milton Road, which is a high frequency public transport route, but it also has little villages along the way where there are lots of supermarkets. In terms of retail space, these villages might not be large shopping centres like Woolworths that you would expect in the suburbs, but that is because they are servicing their immediate catchments. Smaller versions of the supermarkets, such as Coles, Woolworths and IGA convenience stores are expected to be seen a lot more in the future as more people are living in these areas.

As was mentioned above, while TOD plans might seems to be the perfect way to guide the city to a more sustainable retail transport plan, the city might not yet be ready for them.

- How retail trips can be made more sustainable Neither the planners nor the developers have much to recommend when it came to the question of how retail trips could be made more sustainable, and unfortunately their answers were mostly unspecific and indirect.

While there is doubt about how retail trips can ever rely on other means of transport than private cars, there are hopes for behaviour shifts for some specific types of trips or some specific areas within the city that have the capacity. Various groups of people have different ideas about the way they travel. The economists and the retailers have different ideas about how much difference can be made by planners on this issue.

Talking to the developers, the sustainability of retail trips seems to be a very vague and unfamiliar concept.

Chapter 9 Professionals’ Insights on Retail Accessibility 252

Planners drive me crazy. They have this idea on walkability and we should plan for these centres that people can walk to. Only kids and old people will walk to a centre. People are very time-pressed and live very busy lives. They buy more than a handful of things, they take the car. They drive to it even if it is close. But planners still think they should, but they shouldn’t. The same for public transport, you have a handful of things and a child, the bus would be the last thing you consider if you have access to a car. The minute that someone can afford a car you will not see them on the bus doing shopping (Developer 4).

The only way things might change is that petrol goes to 5 dollars per litre or something that might change behaviour (Developer 1).

If the consumers decided that we really want to walk to shops, the market will probably adjust and there is no other way to it (Developer 2).

Many retailers believe that “Cars play big part in retail. Retailing is not a gentle sport, there are winners and losers and it is highly competitive”. It is an industry that is constantly changing. Previously it was easy to place a centre in a location and it functioned very well for a period of ten to twenty years. Today the retail industry is so dynamic that planning schemes find it difficult to keep up with it. There is no way for them to lock in everything and planning schemes have to change with the shift in retail space.

In the early 80s, Australia’s retail used to be around five years behind that in the US. But planning schemes back then didn’t anticipate retail showrooms in centres or the factory outlet centres. They just did not exist here, so they could not plan for them. That is a process that has no end (Developer 3).

Innovations will arise and even people working in the industry cannot anticipate what somebody will come up with in a few years from now. It is a dynamic industry and it has to change to survive as the consumers’ preferences change.

Comparing the previous, more sustainable retail environments with the current circumstances, most of the developer interviewees found it doubtful that the structure of retail environments can ever go back to where it was in the 1930s

Chapter 9 Professionals’ Insights on Retail Accessibility 253 with little grocery stores in walkable distance. The fact is that supermarkets are getting bigger and bigger, they cannot be in close walkable distance every few blocks. Besides, so much has already been invested in places like Westfield that smaller businesses are unable to compete with them.

In a dense city like New York or Chicago, you can get small shops in really densely populated areas where you can’t put together a 1/1.5 hectare site to put a shopping centre but in Australian cities the density is nowhere near those densities even in Sydney, and that density is what drives the public transport (Developer 1).

Within the inner city areas there might be a higher chance of going back to smaller more convenience type centres, rather than having large standalone centres due to the lack of greenfield space available for new developments. This therefore makes it harder for the developer to get approval and for the centre to be financially viable.

Planners believe that in order to return to sustainable transport behaviour, urban density of that residential or local population must be greatly increased. Looking at how retail activities were in the 1930s, they were located in highly accessible, highly visible places surrounded by a dense residential population. People tended to access those destinations by public transport or by walking, and there was a more sustainable transport environment in the community. The land use and density alone cannot guarantee future sustainable transport. Instead, usually when planners undertake the planning process for neighbourhood plans, they need to factor in various things such as what should be zoned, where density should occur, and where the major transport infrastructure such as busways, bus stops or bus routes should go. At the present time in Brisbane, for the neighbourhood plans, the locations of retail spaces are not on the periphery of major public transport routes. Retail spaces are located right in the centre of the neighbourhoods and are only accessible by car. As a result, there is a need to provide car parking spaces.

It is not necessarily our aspiration to have public transport or walking as the preferred and only ways of accessing retail environments, goods and services, but it needs to be provided as part of a sustainable solution. Transport

Chapter 9 Professionals’ Insights on Retail Accessibility 254 infrastructure and land use distribution should be provided in such a way so that people have mode choice. Therefore, when looking at suburban density it can be seen that people are living close to where they can do their shopping and they will not necessarily need to do a car based trip to buy groceries or access other services. If there are more people living in an area, the type of retail services will be more diverse than a petrol station or something like that because there is the density and a catchment, which can support various types of retail outlets and generally, more retail spaces.

Planners do not provide specific planning provisions that say you must not have a car park or should have fewer car parks or that people should use alternate modes of transport. However, planners do say that there should be a minimum number of car parks, that retail and services should be centred around public transport access points. What the planning system should do is to provide options rather than to say it is probably easier to use public transport or cycle or walk as opposed to making car based trips.

Planners might promote more sustainable transport behaviour to access retail spaces but they should always consider the type of retail services provided as well. If the trip is to purchase furniture, for example, there is no sustainable transport option, goods can be delivered. Huge warehouses such as Costco, are located in the city’s outskirts in highly greenfield areas, where the developers can build giant car parks.

The first Costco store that came to Melbourne aimed to be a more compact version than its American counterpart, and to be more urbanised, so the car park was built underneath the store. The developer of the store was trying to promote more traditional car based shopping trips and to promote a more walk up or use of public transport response. It seems that this was an innovative form of that type of retailing which is traditionally built as huge warehouses next to highways. The Melbourne store was an attempt to change this form to a more urban environment to promote public transport accessibility. That is where planners don’t care what the developers do, as long as they conform the planning desires.

Chapter 9 Professionals’ Insights on Retail Accessibility 255

It is hoped that in the future, development will take place where density is the greatest. Future planners and developers need to acknowledge that the form of daily or small trips might need to change to be more conducive to the urban environment, so that car parking spaces can be built beneath the ground, rather than taking up large slabs of land built around residential areas.

To sum up, there is no specific planning policy for making trips to retail areas more sustainable. Land use planners are addressing issues such as where these retail centres are being placed, becoming high trip generators around transport infrastructure and future planning around densities so that people don’t need to use their cars to travel to work or retail centres..

9.4 Conclusion

These interviews were conducted in order to find out more about current retail structure in Brisbane, the reasons behind its formation and how it affects customers’ travel behaviour, the major players who influence retail development, existing barriers and enablers that influence retail transport sustainability, and future plans to overcome the current problems.

It was expected that the interviews would inform us about the realities that are not being extracted from previous directed quantitative analysis and will later form the basis of our proposed scenarios, which will be tested with transport demand model (TDM) in the following chapter. The key findings from the interviews are summarized below.

Planning Regulations and Processes for Retail Development

There are major disagreements between planners and developers about the role of planners and the level of regulation imposed upon by them on retail development. Policies such as those surrounding in-centre developments and EIA submissions for large out-of-centre developments are considered by developers to be a significant hindrance to the approval process, wasting time and money, giving too much power to existing retailers, and limiting competition. Planners, on the other hand, find retail planning policies to be very effective,

Chapter 9 Professionals’ Insights on Retail Accessibility 256 even though there is agreement on the need to make them more flexible and welcoming for developers.

Developers’ plans for retail establishments are concerned only with the issue of profitability. Due to the high price of land, which cannot be simply afforded by everyone, the uncertain nature of the business and the importance of yearly turnover financial feasibility, developers are having a significant say in the final formation of the retail spaces in Brisbane.

The atmosphere of distrust and doubt, which has arisen between planners and developers, has made it difficult to come up with constructive and practical responses to manage and direct the retail environment. Planners believe that City Plan 2014 will help to alleviate some of the developers’ concerns and will make a substantial difference in the near future.

Views and perceptions about retail planning

The current issues of planning are more about floor space provision in nominated areas across the city with preferred retail land use coding and high car accessibility. What the developers are looking for is a highly visible location on a main road with enough space for car parking, easy access for customers, proximity to other similar businesses, a degree of clustering, and overall a higher level of profitability. Other factors, such as a higher level of public transport accessibility and more integration and incorporation with other uses within the larger regional shopping centres, are rising as factors only for the larger regional centres, not for the smaller ones.

While Brisbane already has large regional centres with large catchment areas which cover all the BSD, the population is still expanding, justifying the need for more services for customers. Therefore, large centres are expanding in innovative ways with a focus on lifestyle and recreation. This will bring a high level of profitability for the owners, will follow the council’s in-centre policy, make an existing asset more powerful and will help developers to renovate their businesses. This will make it hard for anyone to oppose this expansion.

Chapter 9 Professionals’ Insights on Retail Accessibility 257

Future trends

The planners and developers interviewed held some common views on the future of retail spaces. Large centres will continue to expand, but this will happen more around the leisure, lifestyle and experience rather than traditional retail. More residential development is expected to be part of these projects. Rather new developments will continue to form as town centre extensions as opposed to isolated big box developments.

Small centres (large neighbourhood centres) will be at the centre of future development activities for the retail environment, will grow in number and will be located on smaller parcels of land in the form of supermarket or convenience store based centres. The focus of such centres is to attract customers who do not want to go to the larger regional centres. The result of this trend is that there will be a reduction in the growth of large shopping centres.

Sustainable transport for retail establishments

While the issue of retail travel sustainability was accepted as being an important matter for future policies, interviewees, especially in the private sector, were doubtful of much change. The retail industry is dynamic, which means that it is difficult for planners to keep up.

However, planners are optimistic about having a more sustainable system due to an increase in population density and having more development along major corridors of growth in the city. This is believed to be essential for the proposed TOD system.

It is time for Brisbane to grow up rather than grow out. People have already begun to trade-off, buying places closer to high frequency public transport rather than having large detached houses in the outer suburbs. Therefore the new city plan and the realization of the role of the corridors, including railways, busways, etc. for growth potential can be of great importance for the future of TODs in Brisbane. The future development of Brisbane will be around corridors and the concept of mass transit. The number of centres, from regional centres down to a small neighbourhood centres, will help these corridors to function in the future as opposed to previous discrete blobs on a map.

Chapter 9 Professionals’ Insights on Retail Accessibility 258

Developers find this plan meaningless, unreasonable and not feasible with the current population growth rate, the standing of the market and the substantial role of private cars in the modern retail environment. TOD development is very expensive and there is yet no clear framework for its delivery in Brisbane. Therefore, available land will be delivered in different ways and the TOD plan will be missed. Brisbane is still far away from that required density and this will lead the city to develop in a different way. What will happen as a result is that in most cases the high-rise and TOD planning opportunity is lost, not because of poor planning but because of the wrong part of the economic cycle.

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 261

10.1 Introduction

This chapter examines how future scenarios of retail distribution in BSD will affect retail travel behaviour -- including the number of shopping trips, the percentage of trips by different transport modes and vehicle kilometres travelled (VKT) by each transport mode. The Brisbane Strategic Transport Model-Multi Modal (BSTM-MM), developed by the Queensland Department of Transport and Main Roads (TMR), is used to perform this analysis, operating under different future scenarios for the retail environment in Brisbane. The previous two chapters provided a better understanding of factors from the retail environment which influence travel behaviour and an assessment of likely future trends in retail development. These understandings will be essential for developing the scenarios under which future travel behaviour will be assessed in this chapter.

The choice model developed in Chapter 8 examined why customers are attracted to specific categories of centres (large, medium or small) and which attractiveness factors affect their destination choices. The interviews in Chapter 9 elicited views and perceptions from experts in the field on how Brisbane’s retail environment might change in the future.

This chapter begins with a review of the concept of scenario planning and its application in the field of land use and transportation. Travel demand models and the use of four-step modelling, as one of the most common methods of scenario testing, will then be described. The strengths and limitations of these types of models are discussed in the light of the literature. A detailed explanation of the BSTM-MM model follows, together with a description of the shopping trip mode choice logit model which is currently embedded in the BSTM-MM travel demand model. The methodology for developing and testing the scenarios will then be discussed. The proposed scenarios generated and defined at city-wide, regional and neighbourhood scales will then be described. The final section explains the results following a transport model run for each future scenario, discusses and compares the differences in travel behaviour outcomes and evaluates the likely impacts of shifts in future retail policies on travel behaviour.

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 262

The results provided in this chapter deliver a view of plausible future forms and distributions of retail centres that will enable planners to anticipate the impacts of their policies on future travel behaviour and to make necessary changes to achieve more sustainable transport patterns.

10.2 Literature review

10.2.1 Scenario planning

Different definitions of ‘scenario planning’ exist. Keough and Shanahan (2008): define it as: “Scenario planning is different from accurate prediction of the future. Instead, it is a process of producing the possible future options which are still being indefinite but at the same time convincing”. Porter (2008) states that: “a scenario is an internally consistent view of what the future might turn out to be - not a forecast, but one possible future outcome” (p.446). Ringland and Owen (2007) state: “Scenario planning is a proven tool for exploring and managing uncertain futures” and it “is a set of processes for creating several scenarios or mental models and using them to aid decision-making” (p.11). (Bartholomew, 2005): p.4) believes that “A process that uses scenarios to assess the future - a ‘scenario planning’ process - utilizes a series of scenarios to gauge possible future conditions. The expectation is that through the process of conceiving, crafting, and evaluating a series of scenarios, an appropriate course, or series of courses, of action can be identified. Hence, through this process, the wide-open question of what the future might bring can be narrowed down to a more manageable set of possibilities”.

Fig 10-1: Scenarios and forecasts / Source: (Ringland and Owen, 2007)

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 263

However defined, these scenarios use existing known information to describe possible changes and conditions in future.

Bartholomew (2007) believes that: “scenario planning can provide important opportunities for citizens to become involved in exploring a wide range of possible futures, leading to the adoption of new policies and implementation strategies that reflect a more desirable future than trend conditions might provide” (p.397). “Scenarios provide an effective mechanism for assessing existing strategies and plans and developing and assessing options. Many organisations have used scenario techniques to help navigate through the uncertainties of their operating environment, in order to make better decisions” (Ringland and Owen, 2007).

The scenario planning method is a way of exploring future conditions. These scenarios will form the basis of more comprehensive future strategies through which organizations, managers and policy makers aim to respond to current and future conditions.

10.2.2 Land-use and transport scenario planning

Land-use and transport scenario planning practice has a long history in the US. The concept was originally transferred from business and the military. A military commanding officer needs to anticipate the enemy’s next move to reduce the level of uncertainty and reduce vulnerability. Similarly a business expert needs to be able to properly anticipate future market conditions and reduce the risk involves in the business (Bartholomew, 2005).

The concept found its way into the field of transport and resource planning in the US in 1960s through the Highway Act in 1962 and later with the formation of the National Environmental Policy Act (NEPA) in 1969 which obliged Federal Government Agencies to provide reports on possible alternative actions and their impacts on the quality of the human environment. Consequently, various methods were developed to examine the future impacts of possible modifications to, or expansions of, natural and man-made systems. Various inputs describing socio-economic characteristics of the population including future location and density of households and employment growth were applied

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 264 in these models, but the interrelated impacts of changes in land use patterns and the efficiency of transport system were ignored. This issue was subsequently overcome with the development of land use transport models in the 1990s (Bartholomew, 2005).

Bartholomew (2007) refers to land use-transport scenario planning as a technique that is being widely applied by metropolitan regions to form their vision for the future. These processes usually focus on regional level growth- related issues such as spatial patterns and urban form, including variations in development density and location of growth around centres. The technique is commonly used to evaluate and compare outcomes under different alternative development patterns. These metropolitan analyses attempt to test different scenarios in order, for example, to decrease VKT by vehicles and emissions of oxides of nitrogen (NOx) in the future. Bartholomew, in her study on the scenario planning process in the US, analysed 80 projects that were undertaken by metropolitan planning organizations in more than 50 metropolitan areas. In Australia the number of such studies is considerably lower, and they tend to focus on commuting trips; (Alexander, 1980; Bell, 1991; Burke et al., 2010).

The results from land use and transport scenarios are typically compared with a base scenario (trend scenario), which portrays the continuation of an existing trajectory into the future. In the trend scenario, the development and transport pattern remains unchanged. The alternative future scenarios typically portray changes in population density, mixed-land uses or developed urban centres, or modifications in transport options or prices compared to the trend scenario. The transport model is then applied to evaluate proposed alternatives to determine their likely impacts on future travel patterns. The last step in the process is to compare the outcomes under these alternative future scenarios to outcomes under the base model in terms of the number of vehicle kilometre (VKT), public transport kilometre travelled (PTKT), the amount of emissions produced or other measurable criteria (Bartholomew and Ewing, 2008).

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 265

10.2.3 Land use and transport models/application of travel demand models

Various mathematical models have been developed to illustrate the interactions between urban land use and transport and to explore likely impacts of land use and transport policies to support the planning process. These models are developed based on existing empirical surveys and help to understand the effects of individual influential factors by keeping all other factors in the model unchanged (Wegener, 2004).

Among the large number of urban and transport models looking at transportation and land use interactions (Wilson, 1998; Wegener, 2004; Sivakumar, 2007), “travel demand models (TDM)” are commonly based on the four-step modelling concept as one of the mostly applied methods. Litman (2014) suggests that:

Travel demand M models (also called traffic models) are designed to evaluate the travel demands of people under specific existing or proposed land use policies, transportation prices and available transit services and consequently measure network traffic volumes and pollution emissions … These models use travel survey and census data to determine Travel Demands, establish baseline conditions and identify trends. Trips are often predicted separately by purpose (i.e., work, shopping, other) and then aggregated into total trips on the network.

The TDM relies on the fact that various types of land uses will generate activities such as housing, working, shopping and leisure. The spatial distribution of these activities necessitates that people travel. It is not only the location of these activities, but also the relative cost of travel and the socioeconomic characteristics of trip makers – such as household income, household size, car affordability etc. – that will affect travel behaviour; this is what the TDM is trying to predict (Gachanja, 2010).

10.2.3.1 Four-step model

The four-step model is the most commonly applied form of TDM that is used to predict the effects of modifications in the transport system. It includes four

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 266 steps: trip generation; trip distribution; mode split; and route assignment as separate components within the model.

While these models are a useful way of exploring the likely future consequences of alternative transport policies, there are a number of limitations that need to be considered before they are applied to a particular setting. Litman (2014) identified a number of limitations associated with four step models that affect the model findings and should be bourne in mind as results are being interpreted by analysts.

Litman refers to the bias of these models towards private cars and how the automobile travel mobility essentially dictates the level of accessibility, while other non-motorized transport modes are very much underestimated. This is not an issue only for the model itself but also for the Household Travel Survey data from which these models are developed. Household Travel Survey data generally do not provide precise records on these trips by other transport mode types. Four step models rely on the concept of congestion; another indication of the dominant role of private cars. The modelled solution to the issue of congestion is to increase the capacity of roads. This ignores the fact that the system could potentially be balanced by trading off time, route, mode type and destination. In reality, trips undertaken using other mode types might be shorter or more comfortable, and thus preferred by trip makers. Consequently, four-step models are not capable of precisely predicting the impacts of planning objectives and policies that try to improve active or public transport. Besides, these models pretty much ignore all qualitative aspects of trips, including the travellers’ security and comfort, and instead only focus on quantitative factors such as travel speed, operating cost (Litman, 2014).

The ”coarse spatial resolution” of traditional four-step models make them relatively incapable of reflecting differences in people’s travel patterns which result from changes in land use patterns especially at smaller neighbourhood scales. Substantial deficiencies and limitations in these models have resulted in more recent development of GIS-based assessment tools and micro-simulation activity-based transport models (Wegener, 2004).

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 267

10.2.4 Objective

This study uses scenario planning to anticipate possible future alternative forms and distributions of retail in the BSD based on the outputs from the analysis and modelling in previous chapters. It aims to model possible swings in the travel behaviour of customers following each of these future shifts in retail land use.

Although there are a considerable number of limitations inherent in the current four-step models, as already discussed, this approach can help us to consider whether any changes in retail land use can have impacts on shoppers’ travel behaviour.

A positive aspect of using the BSTM model is that the model’s strategic scale enables the study to look at the larger area of BSD and thus inform planners/transport planners of the impacts of their city-wide policies on the sustainability of transport in the future. Furthermore, the BSTM model is the only strategic model which is readily available to planners. However, this does not obscure the fact that the BSTM model would benefit from a considerable amount of further development to better represent various types of land uses. To the best of the author’s knowledge, no other land use transport model is currently available for the academic research required to support the aims of this study.

A number of possible future scenarios will be explained in upcoming sections and then run using the existing “Travel Demand Model” in Brisbane’s BSTM, which uses a four-step procedure to analyse and compare travel behaviours. As will be discussed later, BSTM model is based on 2004 HTS data, with additional development to predict outcomes for the years 2011, 2016, 2021, 2026, and 2031, using existing and projected data for these years. The 2016 outcomes are chosen as the base scenario against which the outcomes under alternative scenarios will be compared.

Modelling and analysis in this chapter focuses on the 2016 and 2031 time frames. The 2031 results portray outcomes under predicted and proposed government policies for the future, as defined in the “Connecting SEQ report”. These results will indicate the extent of changes under the proposed scenarios

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 268 and help us to see if proposed policies are seriously considering retail trips in their future predictions or not.

10.3 Methods

The BSTM-MM four-step model is used to estimate transport mode shifts following modifications in retail density and distribution at various scales – city, regional and neighbourhood.

The input data and applied methodology are described in more detail below, followed by an explanation of the model’s variables that can be changed to form the proposed future scenarios.

10.3.1 Brisbane land use and transport model - BSTM_MM

The Brisbane Strategic Transport Model–Multi Model was developed by the Queensland Department of Main Roads. The Brisbane Strategic Transport Model (BSTM) is a fully functional mode choice model capable of estimating transport mode shares in a multi-modal travel environment (Ryan, 2008). The model consists of 1,509 zones covering nine local government areas (LGAs).

The mode choice model consists of two private vehicle modes: ‘car as driver’ and ‘car as passenger’; three public transport modes: ‘walk to public transport’, ‘park and ride’ and ‘kiss and ride’; and two non-motorised modes: ‘walking all the way’ and ‘cycling all-the-way’.

Eight unique transport mode choices were developed for various trip purposes including: home-based work (white collar) (HBW-W); home-based work (blue collar) (HBW-B); home-based education (primary & secondary) (HBE-PS); home-based education (tertiary) (HBE-T); home-based shopping (HBS); home- based 1other (HBO); work-based work (WBW); and other non-home-based trips (ONHB). The models were calibrated using revealed preference data collected in the 2003/04 South East Queensland Travel Surveys (SEQTS) (Ryan, 2008).

1 ‘home-based’ here means trips for the stated purpose from the home as ‘origin’.

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 269

Fig 10-2: A simple multinominal logit (MNL) model specification – Source (Ryan, 2008)

10.3.1.1 Model description

The model specification includes a utility function specification for each transport mode, within an overall random utility model (RUM) framework. The utility function is a linear combination of attributes that describe the attractiveness of a particular transport mode for the specified trip type and destination. These attributes included in the utility function are: travel characteristics, demographic characteristics and land-use characteristics.

Travel characteristics include the time and cost of the trip when undertaken using the seven different modes of transport (Fig 10-2). Demographic characteristics of the trip maker or their household were derived from the origin zone and comprised the number of adults per household and the number of vehicles per household, to indicate the availability of a motor vehicle for a specific trip. Land use characteristics of the destination zone included employment density by various employment categories (Ryan, 2008).

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 270

Fig 10-3: BSTM-MM Study Area (subdivided by Collection Districts), showing constituent local government areas (LGAs)

10.3.1.2 Modelling transport mode choice for Home Based Shopping (HBS)

Home-based shopping (HBS) trips include those made for personal business and shopping. Trips for personal business included trips for banking, health care or seeking other professional services. The definition of shopping did not include travel to a restaurant; these were deemed to be recreation trips and

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 271 were classified as ‘home-based other’ trips. The demographic characteristics, trip characteristics and land use characteristics used to model HBS mode choice for HBS trips are discussed in detail in Appendix 2 section A2-1.

Results from the BSTM, indicate that travel time and cost are the most important attributes affecting mode choice for HBS trips, together with employment density at the destination when public transport mode is the preferred travel mode.

Results also reveal that retail employment density is the most effective parameter in explaining public transport trips, while vehicles per person in the household and vehicles per adult in the household were found to be most effective factors influencing the selection of a private car as the preferred mode of transport. While all the time and cost coefficients in the HBS mode choice model were negative, the results indicated that as the level of private vehicle availability in the household increases so does attractiveness of using private vehicles instead of public transport for a shopping trip.

10.3.2 BSTM results for home-based shopping trips under anticipated future scenarios

10.3.2.1 Scenario development

One way of measuring the impacts of planning policies on urban form and transport is to produce future development scenarios and then study the impacts and scale of change in transport behaviour under those scenarios using a strategic transport model.

Results from the trip makers’ destination choice model (Chapter 8) and the interviews with planning and retail professionals (Chapter 9) revealed that the number of retail opportunities at the destination, the distribution of destinations, population density along transport routes and within the primary catchment of shopping centres, levels of accessibility by various transport modes, shoppers’ socio-demographic characteristics and the clustering/decentralizing of retail activities in the city, are the primary factors that are expected to affect retail travel behaviours. The question that this chapter addresses is “how might these factors affect the mode choice preferences of the trip makers?”

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 272

Fig 10-4: Location of shopping centres within the BSTM zonal system

It is important to note that only a limited number of the potentially influential variables listed in the preceding paragraph are actually included and accessible in the BSTM-MM model which will be used to explore transport behaviour under

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 273 future scenarios. Altering the BSTM to include other potential variables would require significant additional effort that lay beyond the scope of this research.

This chapter seeks to understand transport impacts under future scenarios in which: 1) shopping centres become larger and more discretely clustered in space, i.e. an increased emphasis on the development of large regional centres; or 2) shopping centres become smaller and more evenly distributed spatially, i.e. more emphasis on improving accessibility to small / neighbourhood-sized shopping.

These future scenarios can be enacted in the BSTM-MM by adjusting the location, size and distribution of retail jobs (as a proxy for the size and location of shopping centres) and adjusting the population density around the centres (as a proxy for future highly populated areas around the TODs).

The BSTM will be used to explore travel behaviour under these future scenarios at three different scales –city-wide, regional and neighbourhood. Results will be compared across scenarios and scales.

The first and most important assumption of the BSTM relates to the location of and number of retail jobs at the shopping centres. The exact locations of the centres are available from the SCD dataset. Overlapping these locations in ArcGIS with BSTM zones that contain data for retail employment provides an estimate of the jobs available in those shopping centres.

To implement the alternative future scenarios these retail jobs and the living population in each BSTM zone is either increased or decreased depending on the type of centre they represent and the proposed shifts predicted for the various type of centres under the relevant alternative future scenario. Since the BSTM has been validated and calibrated using HTS data, the total number of jobs/population within Brisbane LGA cannot be changed from the total numbers already allocated in the model. Therefore, all changes in population or jobs in particular zones under alternative future scenarios must be balanced by increasing or decreasing the population sizes or job numbers in other zones. Following any changes in the model input, the BSTM-MM mode choice model is run again to produce trip production-attraction matrices and then the mode

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 274 choice model component is used to assign different trips by various modes to transport networks across the whole city.

10.3.2.2 The study area

As noted above the analysis in this chapter will be conducted at three scales: city-wide across the full BSD, comprising 1,509 transport zones; regional scale comprising almost 500 zones; and neighbourhood scale comprising 30 to 80 zones (Fig 10-5). It is important to undertake analysis at multiple scales so that impacts of changes (and the effectiveness of retail planning policies) can be compared.

The 2016 scenario (embedded in the BSTM) is run as the base scenario. Other future scenarios are then compared against the 2016 results to identify likely future transport impacts. The 2016 scenario is similar to the existing situation in Brisbane in terms of land use, population and network status.

10.3.2.3 Scenario planning

Separate clustering and decentralizing scenarios were developed for the city wide and regional scale. These scenarios were implemented by changing the distribution of retail jobs. - At a city-wide scale, a clustering and a decentralizing scenario were developed. The clustering scenario increased the size of super-regional and major-regional centres and decreased the size of sub-regional centres. On the other hand, the decentralizing scenario increased the size of sub- regional centres and smaller regional centres, whilst reducing the size of the super-regional and major-regional centres. - At a regional scale, three scenarios were developed: larger centres, medium centres and small centres. In the larger centre scenario, the size of super- regional and major-regional centres is increased and the size of the sub- regional and neighbourhood centres is decreased. In medium centre scenarios, the number of sub-regional centres is increased and the number of large regional centres is decreased. In the neighbourhood centre scenario, the number of neighbourhood centres is increased, while the

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 275

number of large and medium centres (super-regional, major-regional and sub-regional centres) is reduced. - At a neighbourhood scale, the regional scenarios were explored within smaller local boundaries to measure and compare the extent of shifts.

In addition, the impact of a population increase scenario was explored at regional and neighbourhood scales (but not at a city-wide scale). The population scenario was implemented by adjusting population distribution. The population scenario, increased the number of people living in the immediate surroundings of large and medium sized shopping centres (Super Regional, Regional and Sub Regional centres) by 15 percent (17,000 individuals), and simultaneously decreased the population in the farther zones outside the boundary of the study area.

A ‘2031’ scenario was also investigated in which the number of retail trips, mode choice decisions and changes in overall VKT, PTKT and number of PT boardings and alightings were estimated for the year 2031. The ‘clustering’ and ‘decentralizing’ and population scenarios at regional and neighbourhood scale were then compared with the 2031 scenario to measure the extent of changes which these scenarios achieved relative to the sustainable future planned for the city in 2031. The 2031 changes will not be only limited to the land use or population distribution, but will also include network and service changes. City- wide comparisons were therefore difficult to interpret and have not been included in here. Details of how the scenarios were developed and implemented are provided in Appendix 2 section 2.3, 2.4 and 2.5. Transport behaviours under each of these scenarios, were compared with the 2016 baseline.

However, the justifications for the selection of the regional and neighbourhood study areas are very briefly discussed below.

There are a number of factors which affected the selection of the regional boundaries:  Proximity to the Central Business District (CBD) comprising smaller BSTM zones, compared to more distant traffic zones with much bigger footprint. The larger the zones get, the less precise the results will be since as discussed before, the model aggregates all the information into the centroid

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 276

of the zones. It is therefore advantageous to choose a region which is not too distant from the CBD.  The presence of all four types of major centres, plus a number of sub- regional and neighbourhood centres in the area, to reflect possible variations in the hierarchy of retail centres’ characteristics.  The location of a number of major transport corridors and major bus stations, identified and planned for future TOD development in the City Plan.  Existing plans for future population increase including infill and new development within the area, which gives us the chance to have more meaningful scenarios for the future  Absence of major barriers such as bus way, river, railway which separate different neighbourhoods from each other and can significantly affect the trip destination decisions of the trip makers.

The two neighbourhood boundaries were selected not only due to the size of the traffic zones, but also based on differences in population, retail density and socio-demographic characteristics. The detailed explanation on the socio- demographic differences of the neighbourhood scenarios are explained in appendix 2, section A2.5. It is assumed that to the great level of similarities between most of the neighbourhoods in the city, these two neighbourhood case studies are likely to be generalizable elsewhere in BSD.

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 277

Fig 10-5: City-wide, Regional and Neighbourhood study areas selected to measure the extent of travel behaviour change under alternative future scenarios for retail development

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 278

10.4 Results and Discussion

10.4.1 Analysis of the results for city-wide scenarios

For the city-wide scenarios, results indicated that the overall number of shopping trips is predicted to increase by almost 2 percent under the clustering scenario compared to the base model (Fig 10-6). This is not unexpected, since further increases in the size of large centres (implemented in the model by increasing the number of retail jobs at those centres) is likely to make them even more attractive as a trip destination.

0 baselinescenario Percentage in change -10

Car PT Cycle Walk Total modesharethe with compare Clustering *1 0.36 38.42 -5.71 -4.13 1.80 Decentralising *2 -5.33 -10.73 1.47 0.32 -5.14

*1 Shows the percentage change in mode share under the clustering scenario compared to the base scenario *2 Shows the percentage change in mode share under the decentralising scenario compared to the base scenario Fig 10-6: Percentage change in mode for home-based shopping trips under the clustering and decentralizing scenarios: modelled at city-wide scale

On the other hand, predictions suggested that the total number of shopping trips would decrease by around five percent, city-wide, under the decentralising scenario. This suggests that more even distribution of mid-sized regional centres across the city could reduce the total number of shopping trips by almost five percent, rather than having a few very large regional centres serving extensive catchment areas. The results might also suggest that the current number of shopping trips to large centres exceeds that necessary merely to satisfy shopping requirements. The large number of trips to this type of centre might therefore be a result the attractiveness of the destination, rather than the need to conduct essential shopping. Thus, these trips might not occur with a

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 279 less centralized pattern of retail shopping opportunities.

While the mode share percentage for walking and cycling is predicted to decrease under the centralization scenario compared to the 2016 base, public transport modes including walking to public transport, kiss and ride and park and ride are all predicted to experience a considerable rise. This increase in the Pt and walking mode share is easier to explain, considering the fact that most of the large regional centres contain a bus station served by high frequency routes. In contrast, keeping the transport network unchanged with more decentralised centres, each with fewer retail jobs compared to the large centres, would be expected to reduce the attractiveness of the centres with good bus service accessibility and thereby reduce the total number of public transport trips. This can indeed be seen in the mode share predictions under the decentralization scenario, where the mode share of PT trips decreases almost 10 percent, compared with the baseline scenario.

An important point to note is that this reduction in public transport mode share is much smaller than the increase in public transport mode share predicted under the clustering option. This is an indication that good public transport services are available for sub-regional centres, so while the large regional centres attract fewer shoppers under decentralization, many trips to (relatively) bigger sub- regional centres are undertaken by public transport.

While a higher percentage of total trips are undertaken by public transport under the clustering scenario compared to the 2016 baseline, there is almost no change in the percentage of trips undertaken by private vehicles. In contrast, predictions under the decentralizing scenario suggest a reduction of almost five percent in private car usage, and a ten percent reduction in PT mode share, while cycling and walking are predicted to show a small increase in their mode share compared to the baseline scenario. The scenario results for trip characteristics (Fig 10-7) predict: the number of kilometres travelled by car or public transport (VKT and PTKT, respectively); the number of hours travelled by car (VHT); and the number of boardings and alightings for public transport, respectively. While the clustering scenario shows an increase in VKT, VHT and PTKT and also PT boardings and alightings at the

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 280 city-wide scale, the decentralising scenario reveals the opposite trend, for all metrics (Fig 10-7).

-5 withbaseline scenario Percentage underchange -10 alternativescenario compared VKT VHT PTKT Boarding Alighting Clustering *1 1.59 7.33 6.92 7.14 7.14 Decentralising *2 -3.09 -7.83 -1.93 -1.42 -1.42

*1 Shows the percentage change for each metric under the clustering scenario compared to the baseline scenario *2 Shows the percentage change for each metric under the decentralizing scenario compared to the baseline scenario Fig 10-7: Percentage change in travel metrics for all trips under Clustering and Decentralising scenarios, compared with the baseline scenario: City-wide predictions

The overall trip length by car is predicted to be approximately 1.5 percent higher under the clustering scenario compared to the 2016 baseline, while this metric is predicted to reduce by three percent under the decentralising scenario. An increase of almost seven percent is predicted in public transport kilometers travelled under the clustering scenario. This is likely to be due in part, to the location of public transport nodes.

The results suggest that even though clustering increases the percentage of PT trips, while decreasing the percentage of active modes trips, it also increases the total number of shopping trips - resulting in an increase in VKT and VHT. On the other hand moving towards a decentralized scenario not only decreases the total number of shopping trip compare to the baseline, but also considerably reduces VKT and VHT.

10.4.2 Analysis of Regional-Scale Scenario Results

Results from regional-scale modelling (Fig 10-8) predict that the total number of retail trips will increase by almost ten percent under the large centre scenario and by around five percent under the population increase scenario, compared

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 281

to the 2016 baseline scenario. Only very minor changes in mode shares are predicted under the 2031 scenario. The total number of retail trips is predicted to decrease by almost six percent under the medium centre and small centre scenario. This may be due to the fact that some trips to large centres are non- essential and only occur because of the attractiveness of the destination rather than an actual need to purchase a specific product.

Under the larger centre scenario, public transport is predicted to increase its mode share by more than 35 percent. This is not surprising, given the increased attractiveness of large centres together with good public transport connections to these locations. A nine percent increase in private vehicle mode share to these large centres is also predicted, while cycling and walking mode shares are predicted to remain almost unchanged from the 2016 baseline.

10 baselinescenario

Percentage in change 0 modeshare compare withthe

-10 Medium-C-S Large-C-S *1 Small-C-S *3 Pop-I-S *4 2031-S *5 *2 Car 8.97 -6.03 -5.88 -5.37 -0.02 PT 36.2 -10.34 -7.47 0.9 0.21 Cycle 0.78 -2.46 -3.9 4.2 -0.21 Walk -0.77 -1.08 -1.11 4.74 -0.05 Total 10.35 -6.15 -5.79 5.19 -0.02

*1 Ratio for the increase in the number of retail jobs for large centre scenario (Large-C-S) to the base scenario *2 Ratio for the increase in the number of retail jobs for medium centre scenario (Medium-C-S) to the base scenario *3 Ratio for the increase in the number of retail jobs for small centre scenario (Small-C-S) to the base scenario *4 Ratio for the increase in the population surrounding the centres to the base scenario *5 Ratio for the 2031 scenario proposed in BSTM model to the base scenario Fig 10-8: Mode shifts in shopping trips, under different future scenarios, expressed as percentage changes from the 2016 baseline

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 282

Having more people living closer to the larger centres would be expected to encourage more walking and cycling trips. It is not surprising therefore that percentage of walking and cycling trips increase by almost five percent under the increased population scenario, compared to the 2016 baseline. What is perhaps important is that the private car mode reduces by around five percent. Therefore, encouraging people to live closer to large centres could be one way to shift trips from private vehicles to walking and cycling.

The two scenarios with more sub-regional and neighbourhood centres (medium and small centre scenarios) are predicted to have similar impacts on mode preferences, with only a small difference between them in the mode share of public transport and cycling. Public transport’s role is slightly stronger in the medium centre scenario. This may be related to the location of PT stations and frequency of bus routes servicing larger centres compared to the services provided for neighbourhood centres.

Predicted changes in trip metrics under future scenarios in the regional scale analysis are shown in Figure 10-9. Unfortunately, the regional-scale model can only report these metrics for all types of trips in combination, rather than for shopping trips by themselves.

Considerable increases in all trip metrics are predicted under the 2031 scenario, but it is important to note that no change in the total number of retail trips was predicted under this scenario in the retail mode share model (Figure 10-8). This suggests that the significant increases in the ‘traffic’ metrics are unlikely to be associated with shopping trips, but rather with other categories of trip such as commuting.

Interestingly, although a larger increase in the number of retail trips by car compared to PT, walking and cycling is predicted under the population increase scenario, the trip metric predictions show a lower rate of increase in VKT compared with PTKT (Fig 10-9). This might be related to shorter distances travelled by private cars when people are living closer to retail centres. Even though the trip metric predictions for population scenario shows a larger percentage of PTKT (15.83), the percentage rate for the trips by PT (Fig. 10-8) does not show significant variation (0.9), therefore the increase in the PTKT

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 283 might result from higher number of people who travel within that area.

Under the larger centre scenario, VKT is predicted to increase only by less than 2 percent, whereas PTKT is predicted to decrease slightly. This shows no significant changes in the overall number of kilometres travelled. The medium centre scenario is predicted to produce a decrease in both VKT and PTKT, while the changes in these metrics under the neighbourhood scenario do not go further than 2.5 percent.

To sum up, predictions from the regional-scale modelling appear to follow the same pattern to those found previously for the city-wide scenarios but the scale of the shifts are not considerable:  the decentralizing scenarios (medium and small centre scenarios) reduce the number of overall shopping trips and VKT  medium and small centre scenarios and larger populations living around the shopping centres appears to encourage more sustainable transport behaviour, even though the predicted change is small for active modes - most likely due to the model limitations noted previously.

10 baselinescenario

Percentage underchange 0 alternativescenario compared with -10 Large-C-S *1 Medium-C-S *2 Small-C-S *3 Pop-I-S *4 2031-S *5 VKT 1.58 -1.54 0.41 11.48 18.40 VHT 4.53 -1.86 1.91 17.41 52.76 PTKT -0.30 0.21 2.27 15.83 38.86 Boarding -0.40 0.45 1.22 22.09 42.02 Alighting 0.85 -0.13 0.80 15.00 40.74

* Future scenarios are the same as those reported in Figure 10-8 Fig 10-9: Percentage changes in overall trip characteristics, under different future scenarios, expressed as percentage changes relative to the 2013 baseline: modelled at a regional scale

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 284

10.4.3 Analysis of Neighbourhood Scale Scenario Results

For the neighbourhood-scale analysis, no new scenarios were developed; instead, the regional scenarios were simply applied again at the neighbourhood scale. Two case study neighbourhoods where selected due to their different socio-demographic and urban form characteristics. The results for the neighbourhood scale analysis are reported as percentage changes in transport mode choice for shopping trips and as percentage changes in the number of VKT, VHT, PTKT, and PT boarding and alighting for all types of home-based trips.

10.4.3.1 Neighbourhood Boundary-1(Fig 10-5)

For the first neighbourhood case study around Garden City as the main super- regional centre, results under the large centre scenario (Figure 10-10) predict a considerable increase in the total number of shopping trips, together with sizeable increases in the car, public transport and cycling mode shares and a drop in the walking mode share compared to the 2016 base line. It is interesting to note that increased concentration of a large shopping centre with very good PT accessibility is predicted to increase the percentage of retail trip rates undertake by PT and cycling, but does not increase the percentage of walking trips (Fig 10-10).

Under the more distributed and less concentrated small and medium centre scenarios, predictions suggest a large reduction in the total number of retail trips compared to the 2016 baseline scenario (7 and 16 percent, under the respective scenarios). The population growth scenario is predicted to generate almost a 20 percent increase in the total number of retail trips in the area compared to 2016, but it should be highlighted that cycling and walking modes become very important under this scenario, by increasing to 20 and 27 percent respectively, compare to the baseline - even though substantial growth in the number of private vehicles trips is also predicted.

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 285

-10 Mode share rate percentage percentage rate share Mode

-20 Large-C-S *1 Medium-C-S *2 Small-C-S *3 Pop-I-S *4 2031-S *5 Car 14.6 -8.1 -16.7 16.8 -0.8 PT 25.1 -8.5 -11.4 2.3 -2.0 Cycle 7.1 -4.5 -15.5 20.6 -1.4 Walk -4.7 0.2 -11.0 27.7 -1.8 Total 13.5 -7.4 -16.4 18.1 -0.9

* All the asterisks are similar to what is explained in figure 10-8 Fig 10-10: Percentage change in transport mode share and total number of shopping trips, relative to the 2016 baseline – Neighbourhood case study-1 [Garden City]

The neighbourhood-scale results are consistent with the regional-scale results which were described earlier, while the changes for neighbourhood scenarios are much more considerable.

Figure 10-11 shows the rate of trip characteristics compare to the baseline scenario for the neighbourhood case study 1 (Garden City). While the medium and small centre scenarios shows a reduction in the number of hours on private car (VHT) and the VKT for all trip types, an almost 7 percent increase is predicted for the PTKT. Making Garden City larger than it currently is, on the other hand, increases the number of VKT and surprisingly reduces the PTKT. Population scenario also shows a considerable increase both for the number of kilometres travelled with PT and car. The results once again show a more sustainable trend for having smaller and medium size centres compare to larger ones.

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 286

percentage percentage 0

Tri[p characteristics' rate rate characteristics' Tri[p -20

-40 Medium-C-S Large-C-S *1 Small-C-S *3 Pop-I-S *4 2031-S *5 *2 VKT 10.29 -3.50 -3.85 16.79 17.12 VHT 51.87 -17.09 -16.41 17.76 56.44 PTKT -24.41 7.00 7.76 15.86 56.65 Boarding -35.29 10.15 9.77 51.53 53.81 Alighting -10.48 7.44 0.68 -28.34 57.70

 All the asterisks are similar to what is explained in figure 10-8 Fig 10-11: Overall trip characteristics’ rate to the baseline - neighbourhood boundary-1[Garden City]

10.4.3.2 Neighbourhood Boundary-2(Fig 10-5)

When it comes to the neighbourhood boundary-2 [Capalaba], the total number of shopping trips is predicted to increase by almost 34 percent under the larger centre scenario (Figure 10-12). Mode share predictions under the same scenario follow a similar pattern to the regional-scale analysis, suggesting more trips by public transport and private car, and fewer trips by cycling and walking compared to the 2016 base scenario. Car trips are predicted to show a 35 percent increase, whereas public transport is predicted to increase by almost 54 percent. Cycling trips are predicted to remain almost unchanged, but walking trips are predicted to reduce by 20 percent.

The medium centres and small centres scenarios are both predicted to generate significant reductions in the total number of retail trips, together with substantial reductions in the number of trips by car and public transport. The medium centre scenario is predicted to generate increases in the number of

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 287 cycling and, particularly, walking trips. Somewhat surprisingly, the number of cycling and walking trips are predicted to decrease substantially under the small centres scenario. This might be related to the lower population density within the close proximity of the centre.

The population increase scenario does not predict considerable influences on the shopping travel behaviour. While similar to the Garden City neighbourhood scenario the role of public transport is not significant, private cars and active mode of transport show a very small rise in the rate of trips with each mode share.

50 40 30 percentage percentage 20 10 0 -10

Mode share rate rate share Mode -20 -30 Medium-C-S Large-C-S *1 Small-C-S *3 Pop-I-S *4 2031-S *5 *2 Car 34.22 -19.76 -27.12 3.41 -0.83 PT 53.67 -21.73 -25.11 -0.20 -1.35 Cycle -0.47 3.96 -9.96 4.64 -1.36 Walk -20.52 16.21 -4.95 1.63 -2.10 Total 33.15 -18.55 -26.64 3.27 -0.90

* All the asterisks are similar to what is explained in figure 10-8 Fig 10-12: Percentage change in transport mode share and total number of shopping trips, relative to the 2016 baseline – Neighbourhood case study-2 [Capalaba]

The medium and small centre scenarios predict a considerable decline in the total number of shopping trips compared to the 2016 baseline. Besides, the medium centre scenario predicts an increase in the number of walking and cycling trips while the active modes have a minus sign for the smaller centre scenario. Under the large centre scenario, the total number of shopping trips is predicted to increase, as is the number of shopping trips undertaken by car and

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 288 public transport. Once again, the results show that having more attractive medium or smaller size centres for people in different locations, rather than aggregated shops and services in one place, will attract more people to use the active modes of transport. Having more people living around major centres will not only increase the total number of retail trips, but will also increase the number of walking and cycling trips. These findings were consistent across both neighbourhood case studies.

When it comes to the overall trip metrics for this neighbourhood, there is a reduction in all metrics of the medium and small centre scenarios, while the population and large centre scenario follows an increase for all different trip characteristics’ rate.

It is important to note that while socio-demographic variations in the two boundaries were expected to influence on the way people travel in the city, the results do not support this hypothesis.

100

20 percentage percentage 0

Tri[p characteristics' rate rate characteristics' Tri[p -20

-40 Medium-C-S Large-C-S *1 Small-C-S *3 Pop-I-S *4 2031-S *5 *2 VKT 29.19 -20.11 -11.26 42.13 12.87 VHT 44.12 -25.40 -13.08 77.71 16.14 PTKT 32.09 -15.53 -10.34 65.83 68.69 Boarding 46.30 -21.16 -8.43 10.00 75.83 Alighting 80.61 -30.39 -33.40 -20.68 61.51

* All the asterisks are similar to what is explained in figure 10-8 Fig 10-13: Shopping trip characteristics’ rate to the baseline - neighbourhood boundary-2 [Capalaba]

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 289

10.5 Conclusions

While the scale of shifts in shopping trips’ mode share and the trip metrics for the city-wide scenarios appear to be quite small, the decentralizing scenario shows more promising results in terms of reducing the total number of retail trips, reducing VKT and PTKT and generating a small increase in cycling and walking trips. For the clustering scenario, larger centres are predicted to generate large increases in the number of PT trips, as would be expected given that bus stations are located at the large shopping centres. There was no significant growth in the number of cycling or walking trips. Thus, for the decentralizing scenario, even though smaller regional and sub-regional centres might be less attractive for customers, they do provide sustainable travel outcomes.

For the other two sets of scenarios, regional and neighbourhood scales, results also suggest a reduction in the percentage of shopping trips for the more decentralized scenarios – i.e. increased numbers of neighbourhood and sub- regional centres - compared to the 2016. This is also typically accompanied by a decrease in the number of trips undertaken for all transport modes. These results support the idea that increasing the number of smaller and more accessible shopping destinations can help promote a more sustainable retail transport environment. Comparing trip characteristics such as VKT, VHT and PTKT, the decentralised medium centre scenarios (having larger number of sub-regional/medium size centres) is predicted to generate higher reductions in all of these characteristics than the neighbourhood expansion scenario.

While overall trends in within neighbourhood boundaries seems to be similar to the regional and city-wide scale analyses, the effects are generally predicted to be stronger and the predicted changes in travel behaviour are more significant. While transport behaviours are modelled at neighbourhood scale, the socio- demographic differences do not show significant influences on the mode choice these groups of people.

Concentrating the population around major centres, also appeared to be effective in encouraging active transport, increasing the number of walking and

Chapter 10 Predicting the Impacts of Retail Future Scenarios on Travel Behaviour 290 cycling trips, but also increasing the number of car trips while not having much impact on public transport usage.

The results for 2031 scenario shows strong shifts in VKT and PTKT at the regional and neighbourhood scales, but the results for the total number of shopping trips and the shifts in the mode share rate do not show considerable differences to the 2016 baseline. This might be because the predicted and planned shifts in 2031 are mostly related to other types of land use such as commuting or changes in the transport network rather than modifications in the distribution and location of retail in the city. This is another indication that shopping trips are not properly considered by transport planners and effective policies affecting the travel behaviour of customers are not yet predicted for future.

Chapter 11 Summary, conclusion and policy recommendations 291

Chapter 11 Summary, conclusion and policy recommendations

Chapter 11 Summary, conclusion and policy recommendations 293

11.1 Introduction

‘How can retail trips become a more sustainable activity in future’ is the question that I have asked and discussed a considerable amount during the last three years with different people sitting in my office, my friends, people whom I’ve met at conferences and meetings, urban/transport planners and professional developers. Raising this question has always evoked a surprise, followed by the instant question of ‘so, what have you found?’ Attempting to change the way people travel to access their retail necessities seems to be implausible to most people due to the nature of these trips or because of existing cultural expectations attached to these types of trip.

Looking back through the literature on retail travel behaviour, a number of solutions have been mentioned quite frequently to overcome the unsustainable transport implications of retail trips. These include the concept of transit- oriented developments, ideas around a more mixed-use urban form and of course providing a stronger, more frequent and comfortable system of public transport and a cycling network to support retail establishments. While these solutions might work appropriately in specific contexts, there are a huge number of other issues which influence the success of these policies such as population density, socio-demographic characteristics of the trip-makers, and cultural aspects of the environment.

Considering what has happened to retail structure of Brisbane over recent decades, the increasing number of shopping centres, more concentrated and car-reliant retail developments and the huge role of private cars, there is legitimate concern regarding how retail trips are affecting transport sustainability in the city.

This research has attempted to construct a platform of understandings around the formation and function of retail in Brisbane and the transport behaviours of retail customers to come up with possible future guidelines that can direct retail travel towards a more sustainable environment.

It is important to appreciate that retail trips are not like commuting trips for which morning trip destinations and evening trip origins are concentrated in only a few

Chapter 11 Summary, conclusion and policy recommendations 294 major locations in the city. Therefore, a simple concentrated or distributed format of jobs planned and tested by transport models cannot easily identify the best and most efficient distribution of transport infrastructure to support retail travel requirements. Retail and its spatial distribution are much more complicated.

So while retail trip patterns might be significantly affected by transport infrastructure policies (such as more frequent and less expensive PT options), they may also be influenced by the results of land use planning policies including the distribution and level of services and products provided at each retail location. Retail travel behaviour could be heavily influenced by the spatial distribution of retail centres that leads people to behave in specific ways. Consequently, this research has focused on planning redistribution policies for retail rather than on policies to reconfigure transport infrastructure.

Retail location is not only a matter of concern for the city planners; it is also very much related to the priorities of developers/owners of land. Besides, customers’ preferences are playing a key role on the decisions made both by developers and planners.

This final chapter is going to review the findings from the applied methods in this research and come up with recommendations on possible land use and planning approaches for the rearrangement or redistribution of retail to assist in delivering a more sustainable trip pattern.

The chapter will start by re-stating the main research questions and presenting a concise summary of the findings obtained from each of the data analysis chapters (Chapters 5 – 10 inclusive). It will be then summarise how these findings contribute new knowledge to the research field. Policy recommendations for practice (both in Australia and within the broader international context) are then put forward. A discussion around the existing limitations of this research follows. The chapter closes with some suggestions for possible avenues for future research.

Chapter 11 Summary, conclusion and policy recommendations 295

11.2 Summary

Three major research questions addressed the behaviour, preferences and decisions of key retail participants (developers, planners, customers).. The findings are discussed briefly here.

As previously explained in the methodology chapter, the methods applied in Chapters 5 and 6 investigated the travel behaviour of customers, their trip decisions, mode choice, travel distances, type of products purchased, and chosen destinations. The influence which socio-demographic characteristics of dominant groups of trip makers exerted over trip attributes was then explored using revealed preferences from the HTS data. Findings from these two chapters indicated that the scope of the research could be restricted only to the hierarchy of the shopping centres in the city. Chapter 7 provided a more detailed investigation of the spatial characteristics of shopping centres as retail destinations.

While the data clearly indicated that most retail trips are undertaken by private car, interesting questions remained concerning the factors which influence destination choice. Chapter 8 aimed to answer this question. Potentially influential factors were identified in Chapters 5, 6 and 7 regarding trip attributes, individual attributes, spatial and destination attributes, and these factors were used to develop a discrete choice model of customers’ preferences for retail destinations.

Planning professionals were the focus of Chapter 9, in which interviews with planners were analysed to determine underlying strategies and policies for ongoing retail development. Chapter 10 then attempted to test the possible impacts of these policies on future retail mode share and the total number of shopping trips. Table 11-1 provides a summary of the main findings from Chapters 5 to 10.

Chapter 11 Summary, conclusion and policy recommendations 296

Table 11-1: Main findings from Chapters 5 to 10

Chapter Findings Five:  Substantial weekly share of retail trips and a high level of reliance on Travel private cars for this trip type, make it a major contributor in the behaviour unsustainable travel pattern of Brisbane.  Public transport and walking are having an almost minor role in the overall

retail mode share  Getting closer to the CBD and the inner Brisbane, substantial shifts in the percentage of motorized and non-motorized shifts are traceable in terms of retail trips which bring up the matter of considering other aspects such as socio-demographic factor or other retail spatial distribution  While daily household requirements especially the grocery and food trips comprise the destination of a large number of retail trips, people still seems to be more interested to carry on these trips on weekdays compare to weekends.  Even for the distances closer than 1 km for the trips to shopping centres and supermarkets, the role of walking and PT trips are very insignificant. Walking is becoming more noticeable for smaller type of centres or supermarkets by a very small rise while PT is hardly used for the trips closer than 5 km.  While the HTS shows that almost more than 50 percent of retail trips are going to shopping centres and supermarkets every week, larger regional centres form the destination of almost one fifth of retail trips.  While more than 40 percent of the retail trips to the CBD are work-based shopping trips, the same amount is allocated to the home-based trips for other types of shopping centre destinations. Six:  Retail trips are forming a large number of overall trips for almost 20 Socio- percent. demographic characteristic  School students and working couples with no kids have the lowest s of trip reliance on private motorised vehicle trips and the highest dependency on makers PT and walking mode of transport. This is then followed by the people living alone both in working age or as retired.  While these groups have a lower percentage of motorized trips, they have the longest distances travelled with private cars  public transport is mostly not applied by households’ who have children, elderly people and female whose main activities is keeping house or are unemployed.  Percentage of trips for groceries and food is similar across most groups mostly followed by the trips to purchase personal or household goods.  The results show that socio-demographic factors including the person’s occupation, having kids in the household, number of people in the household (sole, couple, having kids) and the age (to a lower extent) not only affects the types of modes being selected by the person but also the distances they are happy to travel.  While private cars are still the most preferable mode type among all the groups for at least 75 percent of retail trips, working couples with children and non-worker females have the highest reliance on private motor vehicles by approximately 90 percent.

Chapter 11 Summary, conclusion and policy recommendations 297

Seven:  Following the concept of critical mass, shopping centres are located in Spatial approximately similar distances in terms of the number of people living characteristic around them. Therefore the number of the centres increases in the areas s of shopping as the populated increases. destinations  As the level of the centre increases, its mixed-use score also increase which shows the tendency of having various types of land uses which can serve and attract people.  In terms of travel time, neighbourhood centres show the highest level of accessibility from various CDs within less than 15 minute drive. When it comes to the other type of modes, neighbourhood centres are still ahead of the other centre types but there is a huge drop in the percentage of highly accessible CDs. This is then followed by the Sub-Regional centres with a huge gap in all various mode group compare to neighbourhood centres. Eight:  The results show that the most important factor affecting people’s Destination destination choices between the large, medium and small scale centres is choice model the required travelled distance, followed by the size of the centre, the type of product being purchased and the person’s occupation.  The attractiveness of a destination increases as its gross lettable area- retail (GLAR) increases  The accessibility of shopping centres via walking from a particular trip origin affects trip makers’ destination decisions.  Retail density and mixed use level (entropy) both appear to exert a positive, but statistically insignificant, effect on destination choice  Population density in the locality of a centre appears to exert a negative effect on destination choice which is again hard to explain  Among the interacted factors with distance such as age, occupation, type of products, day of the week , etc. some showed significant aversion to distance: o distance aversion reduces by almost 50% for trips to purchase clothing and household goods, compared with trips to purchase groceries and food o Distance aversion approximately doubles for students compared with the baseline ‘employed’ occupational category. o having children in the household increases distance aversion, this is not a significant influence on trip-makers’ destinations o the model results indicate an increase in distance aversion for weekend trips, although difficult to explain o Young or old trip makers are more averse to distance than 'standard age' trip makers but the results are not significant o Householders which no vehicle also show a non-significant increase in distance aversion, compared with car-owning households Nine:  Retail is mainly formed by developers due to the uncertain nature of the Practitioners’ business and the important issues of yearly turnover and financial visions of the feasibility, etc. rather than the planners. Changes in retail are mostly future unpredictable and not easy to plan for.  Large centres will continue to expand, but more around the leisure, lifestyle and experience rather than traditional retail and more residential

Chapter 11 Summary, conclusion and policy recommendations 298

development is expected to be part of these projects.  New developments will continue to form as town centre extensions as opposed to isolated big box developments.  High street will be still significantly around but less for fashion and traditional retail and more for food and restaurant.  Small centres (large neighbourhood centres) followed by sub-regional centres would be at the centre of future development activities for retail and will grow in number. They will be established on smaller parcels of land in form of supermarket or convenience store based centres.  Future development of Brisbane would be around corridors and the concept of mass transit. The number of centres with a different scale from a regional centre down to a small neighbourhood centres will help these corridors to function in future as oppose to previous discrete blobs on a map.  The high-rise and TOD plan opportunity might be lost not because of poor planning but because of wrong part of economic cycle which doesn’t let them to be financially feasible to happen. Ten:  The centralization and decentralization scenarios for various scale of city- Travel wide, regional and neighbourhood show that as the number of mid-size behaviour centres increase and the number of larger regional centres decrease, the under future total number of shopping trips compare to the baseline scenario (2016) planning will decrease as well as the retail trip rate for each type of mode. scenarios  It might also be a representative of the fact that a number of shopping trips happening to large centres are excessive and unessential trips and might happen because of the higher attractiveness of the destination  While the centralization scenarios decrease the rate of walking and cycling trips to retail, the rate of PT trips will increase. This might be due to the location of major bus stations at the large regional centre.  For the rate of the PT share to the base year, the decentralizing scenario shows some reduction but it is not a major one while the active modes show a small increase in their rate.  Having larger population in the surrounding of the centres, seems influential in increasing the rate of the retail trips compare to the base scenario for the trips happening by car and non-motorised trips, while it doesn’t make considerable changes in terms of PT usage.  Running the 2031 scenario, there is hardly any shift in the rate of mode share compare to the 2016 base year for retail trips. This shows the low level of consideration for the shopping trips in the overall sustainability plan. Instead there is a large swing in the rate of Kilometres Travelled by private cars or public transport for all different trip types which is an endorsement on the significant role that is being given to commuting trips to improve sustainability.  Comparing the socio-demographic characteristics of the people within the two neighbourhood boundaries shows that households with the higher number of employed adults, more children and less people in secondary or tertiary education will be less affected in terms of their non-motorised retail trips while households with larger number of young students and less children seems to be more encouraged by shifts in retail land use distributions.

Chapter 11 Summary, conclusion and policy recommendations 299

In order to briefly reflect on the findings of the research, each research question and the findings related to that question are discussed separately below.

Research question 1: The first set of questions considered the existing retail environment and investigated the behavioural implications of this environment: o What is the current form and structure of retail spaces in Brisbane? o What is the current travel behaviour of Brisbane’s retail shoppers? o Which transport modes are preferred for retail trips in Brisbane?

The answers to this set of research questions can be found in Chapters 5, 6 and 7 and are briefly represented here separately:

Chapter 5: The HTS data showed that amongst all trips, almost a quarter of weekly trips are allocated to retail. This is a very important fact, considering the existing gap in the study of retail trips in Australia. Public transport, cycling and walking are not much used for retail trips in Brisbane, suggesting that retail trips in Brisbane could be considered to be ‘unsustainable’. However, more indepth analysis showed that even though private cars are regarded as the preferred mode of transport for shopping destinations, there is a substantial difference in retail travel behaviour within the CBD and inner Brisbane boundaries compared to the outer Brisbane suburbs. A much lower reliance on private cars was noticeable in inner areas. This might be related to different factors including the urban form differences in the two areas, socio demographic characteristics of the trip makers or even the differences in the nature of shopping trips (work- based or home-based). For example, almost half of retail trips within the CBD are work-based shopping trips, which are very different from home-based shopping trips in terms of the distances travelled or the ease of travelling by private cars such as finding available and cheap parking spaces.

More than half of the retail trips reported in the HTS for the BSD are to supermarkets and shopping centres. The large shopping centres (super and major regional centres) account for about 15 percent of weekly retail trips. The results also revealed that, a substantial number of retail trips, almost 50 percent, were grocery and food trips which – surprisingly – were undertaken mainly during working days rather than the weekend. This is very much in contrast with the concept of one stop grocery shopping at the weekend. Many of

Chapter 11 Summary, conclusion and policy recommendations 300 these trips during working days for daily requirements are undertaken on the return trip from work, or after dropping off children at school, or as a separate trip during the day. Contrary to what appears to be happening in some other countries such as U.S or U.K, these trips could happen anywhere within a shopping centre or a standalone supermarket due to the availability of supermarkets within and outside the centres.

When it comes to the influence of trip distance on mode choice, surprisingly, the results do not show a significant shift in mode choice as trip distance increases. While the broader literature refers to distances less than 1 km as walkable, in Brisbane (Burke and Brown, 2007), walking is only rarely used for shopping trips. PT is even experiencing a worse situation compare to walking.

The results reinforce the a high preference for the private-car amongst shopping customers and indicate the importance of studying other factors – apart from mode choice - when considering options for improving the sustainability of retail travel behaviour.

Chapter 6: Interestingly, the analysis of socio-demographic characteristics of trip makers showed that different groups of people with different status in terms of household size, household income, number of children in the household, age, gender, occupation and license-holding status, show considerable similarities in their mode choice and travel behaviour. All socio-demographic groups were highly reliant on private cars (more than 75 percent) as expected, however factors such as having children in the household, the occupation of the trip maker and to some extent their age and their sex appeared to affect preferences for selecting more sustainable travel modes.

Working couples with no children and school students showed the lowest reliance on private motorized vehicles and the highest dependency on PT and walking modes, followed by people who live alone or are retired. On the other hand, PT is not generally used by households who have children, elderly people and by females whose main activity is keeping house or by people who are unemployed. These findings might seem to be more helpful for planning at a local or neighbourhood scale rather than the regional scale; however, they could also have significant implications in the larger regional scale considering

Chapter 11 Summary, conclusion and policy recommendations 301 the type and locations of the houses that these different groups select for their living. Looking at the existing Australian context, young couples with no children and students generally prefer to live in smaller units in highly populated areas which are closer to PT corridors, services and shops. Larger households with children and usually one or two employed adults typically prefer larger lots with detached or townhouse types of dwellings.

Chapter 7: As previous chapters identified that shopping centres were the destination for more than 50 percent of retail trips, shopping centres formed the focus of the research from Chapter 7 onwards. Chapter 7 analysed the spatial distribution of almost 200 shopping centres within BSD, and quantified spatial characteristics of these centres and their settings. The results identified the minimum feasible population density required for the catchment of shopping centres, while looking at the distances between centres showed approximately similar spaces in terms of the number of people living in their catchments. Therefore, the number of centres in an area would be expected to increase as the population in that area increased.

As shopping centres get bigger the level of mixed-use within their surroundings was also shown to increase to support the needs of the population that travelled to that area. Conversely, the vicinity of smaller centres showed a lower level of mixed-use.

Small centres were shown to have the highest level of accessibility by cars from the collection districts around the city. Accessibility by PT, walking and cycling was much lower than accessibility by private cars, although accessibility by these other transport modes is still highest for neighbourhood centres, followed by sub-regional centres.

For half of BSDs collection districts, it takes more than 60 minutes to get to large centres by PT. Medium sized centres are more accessible by PT, with less than 45 minutes accessibility from 47 percent of CDs. However, 45 minutes is still quite long time to spend on PT. Accessibility by PT is better for small centres, which can be accessed by PT in less than 15 minutes from 11 percent and in less than 30 minutes for almost 40 percent of collection districts. This is still considered a poor level of PT accessibility for a daily trip destination.

Chapter 11 Summary, conclusion and policy recommendations 302

Walking accessibility is very poor for almost all types of centres. It is not surprising that walking is not considered a viable option for the vast majority of retail trip makers.

Research question 2: The second set of questions aims to model shopping destination choice and identify the factors that influence destination choice: o How do Brisbane’s shoppers choose between available retail destinations? o Which factors affect shoppers’ destination choice? o Which types of destinations are generally preferred and why?

The answer to these research questions can be found in Chapters 8,

Chapter 8: The findings from previous chapters on retail structure and travel behaviour justified limiting the scope of the research to shopping centres and provided detailed information about mode choice and trip characteristics. The earlier chapters also provided the basic information required to address the next question about shoppers’ destination preferences.

Chapter 8 presented a destination choice model for retail trips. The results showed that distance plays a significant role in trip makers’ decision, with shoppers much more inclined to select closer destinations. The size of the centre, and therefore the number of opportunities available to customers at the destination, was also found to influence destination choice. The attractiveness of a destination increased as its gross lettable area-retail (GLAR) increased, probably reflecting the additional attractiveness of a broader range of shopping possibilities. Distance aversion was found to be very much weaker when shopping for clothes, household goods or other types of products, than for grocery and food trips.

A strong aversion was also identified for walking distance. The longer it takes to walk to a particular size of shopping centre, the less likely a shopping centre of this type will be chosen as the shopping destination.

Other factors such as person’s occupation also affect the choice of destination, but less significantly than the factors mentioned previously. Among other occupational groups, students were found to display the strongest aversion to

Chapter 11 Summary, conclusion and policy recommendations 303 travel distance. This is consistent with previous findings about this group of trip makers who are less likely to travel by private car.

Other socio-demographic factors which had been shown in Chapter 6 to be associated with mode usage were found to have the expected impact on distance aversion. For example, young and old trip makers showed a greater aversion to travel distance than standard age group trip makers.

Research question 3: The third set of questions focuses on the potential future of retail environments. They ask whether under current trajectories particular policies could help to facilitate a shift towards a more sustainable retail transport future: o What appears to be the likely future of retail spaces in Brisbane – given existing trends and probable future policies? o How could planning policies potentially help to facilitate a shift towards a more sustainable future for retail environments in Brisbane?

To answer these final questions, two methods were applied: interviews and travel demand modelling as discussed in Chapters 9 and 10. The first approach involved collecting ideas from the planners and retail developers/owners regarding future possible trajectories for retail in Brisbane. The second method used a pre-existing strategic transport model to estimate the impacts of potential future policies on travel behaviour.

Chapter 9: Talking to people in the profession, it became clear that developers and shopping centre owners play a very significant role in the future of sustainable retail transport, and that risk, profitability and innovation are key issues for the future of retail in the city. While Brisbane’s retail future will continue to be based around its existing shopping centres, it appears likely that these centres will continue to expand to offer more opportunities for leisure, lifestyle and experience in the form of town centre extensions as opposed to isolated ‘big box’ developments. The ‘high street’ will continue to exist, but will be orientated more towards food and restaurant outlets and less towards fashion and traditional retail. A much more significant role (in both size and numbers) is predicted for the smaller neighbourhood and sub-regional centres based around a supermarket or convenience stores.

Chapter 11 Summary, conclusion and policy recommendations 304

More residential developments are expected in retail projects along main transport corridors or within large regional centres, in line with concepts of mixed-use and critical mass, while high-rise building and TOD plans are not currently being considered by developers.

Development corridors and the concept of mass transit are believed to be central to Brisbane’s retail development. Shopping centres of all sizes will help these corridors to function in the future, in contrast to the previous approach of ‘discrete dots’ around the city and next to major roads.

Chapter 10: While the findings from previous chapters are essential in developing ideas, plans and policies to guide the future direction of retail, it is also important to look into possible shifts in retail travel behaviour that will follow from the implementation of these plans in future.

Possible future scenarios and new policies are modelled in Chapter 10 using a pre-existing travel demand model - BSTM-MM. The results showed some very interesting findings on retail mode shift and overall trip attributes. Results under different retail distribution scenarios (more centralized or more decentralized retail distribution) showed that within various scales, having a larger number of mid-size and small centres will decrease the total number of shopping trips and the number of trips by each transport mode compared to the baseline scenario (2016). A future scenario with larger more attractive centres showed that many retail trips might not be considered as essential, but as discretionary. More centralized scenarios for retail distribution could help to increase the number of PT trips, while the decentralizing scenario was predicted to produce an insignificant reduction in PT usage and a slight increase in the percentage of active mode use compared to 2016.

Larger populations around the centres were predicted to increase the rate of the retail trips by car and active transport modes compared to the base scenario, without making considerable changes in PT usage. This again suggests that people do not consider PT options for shopping trips, even for very short distances, with private cars and walking or cycling preferred.

Hardly any change in the mode share for retail trips was predicted under the Brisbane 2031 development scenario compared to the 2016 base year. This

Chapter 11 Summary, conclusion and policy recommendations 305

suggests that shopping trips have received little attention in the overall ‘2031’ sustainability plan. Instead, predictions suggest a large decrease in the number of kilometre travelled by private cars or public transport for all different trip types. This emphasises the importance of considering retail trips when planning to improve transport sustainability in the city.

Comparing the socio-demographic characteristics of residents in the two neighbourhoods revealed some differences in their shopping trips. For the proposed scenarios, households with larger number of young students and less children are predicted to make more non-motorised retail trips than the ones with higher number of employed adults, more children and less people in the secondary/tertiary education per household.

11.3 Contribution to knowledge

This study was an attempt to look at retail as one of the most important elements of the urban environment that has major transport implications, but has evidently been ignored in Australian cities. The review of the literature showed a number of gaps especially among academic studies of retail. This study aimed to address these gaps by answering relevant research questions through a case study in Brisbane. In this section, the contribution of the thesis is summarized under a number of themes related to existing knowledge gaps.

Urban planning, economic geography and marketing studies

Looking back through the literature, urban planning studies mostly focus on elements of urban form and their impacts on non-work/retail trips. The literature has addressed the impacts of the road network, mixed use, population density, accessibility and urban design (Tracy et al., 2011) (Agyemang-Duah et al., 1995; Krizek and Johnson, 2006).

On the other hand, economic geography has focussed on the location of economic activities to service the population, with considerable attention devoted to the scale, level of function, level of accessibility and catchment populations that shape each retail destination. This strand in the literature began back in the early 19th century and developed considerably from

Chapter 11 Summary, conclusion and policy recommendations 306

Christaller’s simple CPT to the complicated discrete choice models of today (Fotheringham, 1988; Yrigoyen and Otero, 1998b; Bates, 2003; Roy and Thill, 2003).

While there are considerable overlaps between these approaches, the variations in their preferences and priorities, resulted in an environment that planners put forward policies (such as zoning and requirement for EIA or the TOD plans) which business does not support and found them counter- productive.

This study aimed to come up with a more comprehensive approach considering the preferences and desires of various retail players including developers, customers and planners. While the number of studies for each category is quite considerable, to the knowledge of author, there are no academic studies that have considered these various factors simultaneously.

The scale of study

The majority of studies in the literature are limited to small neighbourhood scales rather than regional scale (Rapoport, 1987; Owens, 1993a; Whyte, 2012), although Handy’s 1992 study is an exception (Handy, 1992). This study on the other hand, focused on the transport implications of retail at a regional scale, even though the future scenarios were studied across a range of scales from city-wide to regional and neighbourhood. The TDM results observed stronger effects on travel behaviour at the smaller scales; however, considering the shortcomings and biases existing in the four-step BSTM-MM model, important shifts in mode choice for retail travel were also detectable at larger regional scales.

Study of the customers retail preferences

The existing literature provides different examples of studies on the non- work/retail mode choice and the factors affecting customers’ mode decisions (Richards and Ben-Akiva, 1974; McCarthy, 1980; Hausman and McFadden, 1984; Limanond et al., 2005). However, the destination choice of customers has rarely been addressed in existing research.

Chapter 11 Summary, conclusion and policy recommendations 307

Cities like Brisbane can provide their shoppers with easy access to almost all types of retail destinations, so a good understanding of customers’ destination decisions and preferences can help to inform city-wide sustainability planning. As car based trips are typically much less expensive than PT alternatives, mode choice is dominated by the private car. In this situation, mode choice is likely to be influenced by destination choice, rather than vice versa.

The application of mixed of methods to cover various aspects of the issue

Previous researchers have used different methods (such as aggregated/disaggregated analysis, choice modelling, TDM) to study retail trips (Handy, 1996a). Each of these methods has advantages and disadvantages that might limit the findings. The selection of method is partly related to the availability of datasets and partly a consequence of the research objectives and the type of factors being investigated.

This research instead has attempted to consider a broader range of factors which influence the decisions of developers, planners and customers and has therefore used a broad set of methods to address and study these different aspects of the retail environment. Therefore, considering the limitation of these methods (partly discussed in Chapter 2: literature review), five different methods of analysis were applied to cover the variety of these issues.

Focusing only on shopping centres as key elements of the retail environment

While the literature refers to some studies on the differences between traditional and modern forms of retail (Friedman et al., 1994), there is no existing study on shopping centres and the significant role that they play in customers’ travel behaviour in Australia.

Retail travel behaviour has changed enormously since the 1950s when the first shopping centres were established in Brisbane and Melbourne. This is largely related to the distribution and the function of these centres in the city. Chapter 5 showed that more than half of retail trips end up in various types of centres.

Chapter 11 Summary, conclusion and policy recommendations 308

Therefore, this study has focused on the significant influences which these retail destinations exert on the travel behaviour of Australian city dwellers.

Various types of products rather than only the groceries and food

Studies on retail have, in many cases, focused on a specific type of product such as grocery and food trips, rather than trying to obtain an overall understanding of retail customers’ trip behaviour (Recker and Kostyniuk, 1978; Guy, 1987; Carrasco, 2008). Given that shopping centres dominate as retail destinations, and that many different types of product are available in most shopping centres, it is better to consider shopping trips for all different types of products to obtain a better understanding of retail transport behaviour. .

11.4 Limitations

When it comes to the study of retail there are quite a large number of limitations that arise from a lack of information and restricted datasets. Examples include: 1- Retail transport data are usually collected by the owners of retail establishments for specific marketing purposes which do not necessarily match planners’ research requirements. Apart from that other types of datasets such as the SCD dataset on the shopping centre detail and the POI (Point of Interest) which shows the location of various types of retail are also accessible for a number of these retail destinations. However, the problem is the focus, scale and the purpose for the collection of these datasets which are mostly not consistent and are hard to be joined and merged and make a reliable precise applicable dataset. This is one of the major reasons which led the research into lots of inevitable assumptions to make it possible. 2- Existing transport models have mostly overlooked the major role of retail trips and have not addressed retail trips with the same level of detail or rigour as commuting trips. 3- Existing LUPTAI model to measure the accessibility (used in this study as the best way of measuring the accessibility level) cannot consider the locational characteristics (various types of centres for example) and is only based on the geospatial location of the activity.

Chapter 11 Summary, conclusion and policy recommendations 309

4- While online shopping in Australia is not yet as dominant as it is in some other countries such as U.S., it is increasingly growing in number. The lack of reliable information yet make it really hard to investigate the real influence of the online shopping on people’s travel behaviour. While many professionals (due to the interviews) are quite sceptical about major influences of online shopping on the dominance of shopping centres’, this research couldn’t find reliable resources to support testing the internet shopping influences on customers’ travel behaviour.

Overcoming the above-mentioned limitations should help to produce much stronger and more reliable results from the methodological framework and methods that have been developed through the thesis.

11.5 Recommendations

11.5.1 For Australian practice

From what was discussed above, it is hard to expect the urban/transport policies to make significant changes in the travel behaviour of people in terms of their mode choice. People seems to be very much interested in using their own cars to go for their shopping rather than carrying everything on public transport or walk to home with plastic bags in their hands and possibly one or two children accompanying them and also considering the expensive fees of PT.

PT, walking and cycling as other means of transport supporting the sustainability concept should be developed and invested for, but it should be also noticed that having them available by itself, might not make a huge difference in the travel behaviour of people. Other factors including the supportive environment and developing the culture of sustainability among the trip makers should also be seriously considered.

Specific groups of people including young couples with no kids and the secondary or tertiary students followed by sole people, could be the supporting group for the other mode types including PT, walk or cycling. These are the groups which seems to be happy to live in smaller units in more densely populated areas closer to the PT corridors and are usually more interested in

Chapter 11 Summary, conclusion and policy recommendations 310 living in more mixed-use types of developments with retail at the ground level and housing units on the top. Providing more village type of shopping centres in smaller neighbourhood or sub-regional scale around these corridors will help them to strengthen/improve their sustainability travel behaviour as it is recommended in the new city plan (2014).

However, when it comes to families with children, living within the outer dispersed suburbs’ area, it is much harder to encourage them to change their reliance on private cars. For PT, making any changes in number and frequency of services provided for these trip makers needs strong support from the population who are going to hop on and off alongside the PT routes. Developing TODs along these PT corridors is reliant on a supportive critical mass to maintain the economic side of the developments. Brisbane population and its population growth rate are not strong supportives of these types of developments. While there are specific spots in the city mostly within inner Brisbane, that these developments seems to be feasible and have already partially happened, many planned TODs have been a failure and have later experienced a shift back to lower less densely populated plans. So currently, the market does not match the required planning shifts but it might be ready in a few years from now. However, when the opportunity of vacant undeveloped land is gone, changing the land use would be very hard for at least the next 30 to 50 years which is the average life of a building.

Therefore, planners might start thinking and focusing on reducing the number of kilometres that is being travelled by private cars rather than only considering the shifts to non-motorized or public modes of transport. They can start to think about redistribution of retail more frequently and in closer distances instead of large in-centre developments with greater number of shops and activities and a bigger catchment area.

This trend not only matches the desires of the developers in having bigger neighbourhood centres which has already started to happen, but also make customers happier to travel shorter and closer distances for a large part of their weekly trips which is going for groceries and food.

Chapter 11 Summary, conclusion and policy recommendations 311

The new city plan1 has already started to support the increase in the competition by omitting the obligations for an EIA for new developments while there are a basic number of requirements that the developers should provide. Therefore, for example a new Aldi store can be established anywhere as long as the developers think it would be a profitable business and Coles and Woolworth has no power to stop Aldi by taking the case to the court and hindering the development process.

While existing large regional centres will continue to work, a larger number of neighbourhood centres with smaller catchment areas around them not only can encourage people to consider other available options such as cycling and walking but will also decrease the number of kilometres travelled by customers’ private cars to access these destinations.

11.5.2 For international practice

The findings of this research are based on Brisbane and its specific socio- demographic, economic, geographic and planning context. However, people with a similar suburban lifestyle in other countries could be sharing very common issues and experiences.

Low density suburban areas where private cars are essential for accessing activities and services, do not provide many transport mode alternatives for daily trips. Large shopping centres service these suburban areas in almost all of these communities, worldwide. While there might be differences between the characteristics of these centres such as having a supermarket within the centre or not, they all follow very similar objectives and patterns of development to provide customers’ requirements. The form of the centre might be open or closed due to differences in weather and environment, but they will still aim to deliver a broad range of shopping, recreational and socializing opportunities to encourage customers to spend more time there.

Socio-demographic characteristics might differ from one community to another and may influence preferences for using public transport, walking or cycling,

1 Brisbane City Plan 2014, Source: http://eplan.brisbane.qld.gov.au/

Chapter 11 Summary, conclusion and policy recommendations 312

however, these factors are much more likely to influence transport behaviour at small neighbourhood scales. Therefore, the recommendations presented above for Brisbane would seem to be applicable to similar car-based unsustainable cities elsewhere. While the concept of transit oriented developments (TODs) relies on a number of different factors, more concentrated and compact populated areas along transit routes will increase the possibility of more sustainable transport usage.

11.5.3 For further research

Aside from recommendations for planning practice, this research has opened the discussion for a large number of studies essential for understanding retail travel behaviour and ways to encourage more sustainable retail travel in the future. Some of these research opportunities are:  Conducting explicit interviews with retail trip makers to identify the factors which drive their retail preferences and choices of destination and transport mode  Focus on the issues of transit oriented development and linkages between the developers and planners and how to make it happen  Running more specific analysis of travel behaviour and preferences of major identified socio-demographic groups of customers  Developing a better understanding of the importance of accessibility by looking at the accessibility of individual shopping centres rather than considering the overall accessibility of a specific category of shopping centre  Collecting more detailed data on the exact origin and destination of retail trips (rather than the existing aggregated HTS data)  Developing simultaneous destination and mode choice models  Looking at the impacts of change in PT frequency and network configuration  Investigating the impacts of PT fares and parking charges

References

References 315

(1981) Retailing in Canberra, NCDC Technical Paper 31, Canberra (2006) Living in Brisbane 2026- our shared vision, Brisbane City Council. (2012) Trave in south-east Queensland-An analysis of travel data from 1992 to 2009, The State of Queensland, Department of Transport and Main Roads. Adler, T. J. & Ben-Akiva, M. (1976) Joint-choice model for frequency, destination, and travel mode for shopping trips. Transportation research record, (569). Agyeman, J., Bullard, R. D. & Evans, B. (2003) Just sustainabilities: development in an unequal world, MIT Press. Agyemang-Duah, K., Anderson, W. P. & Hall, F. L. (1995) Trip generation for shopping travel. Transportation Research Record, (1493), 12-20. Akiva, M. E. B. & Lerman, S. R. (1985) Discrete choice analysis: theory and application to predict travel demand, MIT press. Aldous, T. (1992) Urban villages: a concept for creating mixed-use urban developments on a sustainable scale, Urban Villages Group. Alexander, I. (1980) Office dispersal in metropolitan areas II: Case study results and conclusions. Geoforum, 11(3), 249-275. Applebaum, W. (1966) Methods for determining store trade areas, market penetration, and potential sales. Journal of Marketing Research, 127- 141. Arentze, T., Timmermans, H. & Janssens, D. (2006) Modeling short-term dynamics in activity-travel patterns: from Aurora to Feathers. Athey, S. & Imbens, G. W. (2007) Discrete choice models with multiple unobserved choice characteristics. International Economic Review, 48(4), 1159-1192. Australian Government (2009) Urban public transport: recent bus transport statistics, Department of Infrastructure, Transport, Regional Development and Local Government. Badoe, D. A. & Steuart, G. N. (1997) Urban and travel changes in the greater Toronto area and the transferability of trip‐generation models. Transportation Planning and Technology, 20(4), 267-290. Baker, R. G. V. & Wood, S. (2010) Towards Robust Development of Retail Planning Policy: Maintaining the Viability and Vitality of Main Street Shopping Precincts. Geographical Research, 48(1), 65-74. Bartholomew, K. (2005) Integrating land use issues into transportation planning: Scenario planning. Bartholomew, K. (2007) Land use-transportation scenario planning: promise and reality. Transportation, 34(4), 397-412. Bartholomew, K. & Ewing, R. (2008) Land use–transportation scenarios and future vehicle travel and land consumption: a meta-analysis. Journal of the American Planning Association, 75(1), 13-27.

References 316

Bates, C. (2003) New approaches to retail forecasting. Leading Diversity. Adelide Planning Institute of Australia (PIA). Batty, M. (1978) Reilly's challenge: new laws of retail gravitation which define systems of central places. Environment and planning A, 10(2), 185-219. Bekhor, S. & Prashker, J. N. (2008) GEV-based destination choice models that account for unobserved similarities among alternatives. Transportation Research Part B: Methodological, 42(3), 243-262. Bell, D. A. (1991) Office location—city or suburbs? Transportation, 18(3), 239- 259. Bernard, P. O. (1987) Modelling shopping destination choice behaviour using the basic multinominal logit model and some of its extensions. Transport Reviews: A Transnational Transdisciplinary Journal, 7(1), 17-51. Boarnet, M. & Crane, R. (2001) The influence of land use on travel behavior: specification and estimation strategies. Transportation Research Part A: Policy and Practice, 35(9), 823-845. Boarnet, M. & Handy, S. (2010) DRAFT Technical Background Document on the Impacts of Residential Density Based on a Review of the Empirical Literature. Boarnet, M. G. & Sarmiento, S. (1998) Can land-use policy really affect travel behaviour? A study of the link between non-work travel and land-use characteristics. Urban Studies, 35(7), 1155-1169. Bordoloi, R., Mote, A., Sarkar, P. P. & Mallikarjuna, C. (2013) Quantification of Land Use diversity in the context of mixed land use. Procedia-Social and Behavioral Sciences, 104, 563-572. Bromley, R. D. F. & Thomas, C. J. (1993) The retail revolution, the carless shopper and disadvantage. Transactions of the Institute of British Geographers, 222-236. Bromley, R. D. F. & Thomas, C. J. (2002) Retail Change, Taylor and Francis. Brown, S. (1995a) Christaller knew my father: Recycling central place theory. Journal of Macromarketing, 15(1), 60-72. Brown, S. (1995b) Christaller knew my father: Recycling central place theory. Journal of Macromarketing, 15(1), 60. Bruton, M. J. (1985) Introduction to transportation planning. Bucklin, L. & Gautschi, D. (1983) The importance of travel mode factors in the patronage of retail centers. Patronage Behaviour and Retail Management, 45-55. Burke, M. & Brown, A. (2007) Distances people walk for transport. Road & Transport Research: A Journal of Australian and New Zealand Research and Practice, 16(3). Burke, M., Dodson, J. & Gleeson, B. (2010) Employment decentralisation in South East Queensland: Scoping the transport impacts. Research paper, 29.

References 317

Calthorpe, P. (1993) The next American metropolis: Ecology, community, and the American dream, Princeton Architectural Press. Cao, X., Mokhtarian, P. L. & Handy, S. L. (2009) Examining the impacts of residential self‐selection on travel behaviour: a focus on empirical findings. Transport Reviews, 29(3), 359-395. Cao, X. J., Xu, Z. & Fan, Y. (2010) Exploring the connections among residential location, self-selection, and driving: Propensity score matching with multiple treatments. Transportation research part A: policy and practice, 44(10), 797-805. Carlsson-Kanyama, A. & Linden, A. L. (1999) Travel patterns and environmental effects now and in the future:: implications of differences in energy consumption among socio-economic groups. Ecological Economics, 30(3), 405-417. Carrasco, N. (2008) Deciding where to shop: disaggregate random utility destination choice modeling of grocery shopping in canton zurich, Master Thesis, IVT, ETH Zurich, Zurich. Cervero, R. (1989) Jobs-housing balancing and regional mobility. Journal of the American Planning Association, 55(2), 136-150. Cervero, R. (2002) Built environments and mode choice: toward a normative framework. Transportation Research Part D: Transport and Environment, 7(4), 265-284. Cervero, R. (2005) Accessible Cities and Regions: A Framework for Sustainable Transport and Urbanism in the 21st Century. Urban Transport, Berkeley. Cervero, R. & Kockelman, K. (1997a) Travel Demand and the 3Ds: density, diversity and design. Transportation Research D: Transport & Environment, 3(2), 199-219. Cervero, R. & Kockelman, K. (1997b) Travel demand and the 3Ds: density, diversity, and design. Transportation Research Part D: Transport and Environment, 2(3), 199-219. Christaller, W. (1966) Central places in southern Germany, Prentice-Hall. Consulting, P. D. (2008) Land Use And Public Transportation Accessibility Index (LUPTAI) model improvements, Queensland Transport Integrated Transport Planning Division. Corporation, Q. S. B., Queensland. Dept. Of Business, I. a. R. D. & Government, Q. D. O. H. a. L. (1992) A Discussion paper on the proliferation of shopping centres and related matters, 0724243682, Dept. of Business, Industry and Regional Development, with the assistance of the Dept. of Housing and Local Government and the Queensland Small Business Corporation, . Crane, R. (1996) Cars and drivers in the new suburbs: linking access to travel in neotraditional planning. Journal of the American Planning Association, 62(1), 51-65.

References 318

Crane, R. & Crepeau, R. (1998) Does neighborhood design influence travel?: A behavioral analysis of travel diary and GIS data. Transportation Research Part D: Transport and Environment, 3(4), 225-238. Cubukcu, K. M. (2001) Factors affecting shopping trip generation rates in metropolitan areas. Studies in Regional and Urban Planning, 9, 51-68. Curtis, C. & Perkins, T. (2006) Travel behaviour: A review of recent literature. Perth, WA: Urbanet, Curtin University of Technology. Davison, G. & Yelland, S. (2004) Car Wars: How the Car Won Our Hearts and Conquered Our Cities, Allen & Unwin. Dennis, C., Marsland, D. & Cockett, T. (2002) Central place practice: shopping centre attractiveness measures, hinterland boundaries and the UK retail hierarchy* 1. Journal of Retailing and Consumer Services, 9(4), 185-199. Dieleman, F. M., Dijst, M. & Burghouwt, G. (2002) Urban form and travel behaviour: micro-level household attributes and residential context. Urban Studies, 39(3), 507-527. Dodson, J., Burke, M., Evans, R., Gleeson, B. & Sipe, N. (2010) Travel Behavior Patterns of Different Socially Disadvantaged Groups. Transportation Research Record: Journal of the Transportation Research Board, 2163(-1), 24-31. Doust, K. & Black, J. (1996) A holistic assessment framework for urban development and transportation with innovative triple bottom line sustainability metrics. Projections-MIT Journal of Planning 9, 11-24. Downs, R. & Desouza, A. (2006) Learning to think spatially: GIS as a support system in the K-12 curriculum. Committee on the Support for the Thinking Spatially, National Research Council, Publisher: The National Academies Press, URL: http://books. nap. edu/catalog. php. Dramowicz, E. & Magazine, D. (2005) Retail Trade Area Analysis Using the Huff Model. Directions Magazine. Drechsler, P. (2014) Metropolitan Activity Centre Planning in Australia: Implications of Millennial Consumption Practices. Urban Policy and Research, 32(3), 271-287. Ewing, R. (1995) Beyond Density, Mode-Choice, and Single-Purpose Trips. Transportation Quarterly, 49(4), 15-25. Ewing, R. & Cervero, R. (2001) Travel and the built environment: a synthesis. Transportation Research Record: Journal of the Transportation Research Board, (1780), 87-114. Ewing, R. & Cervero, R. (2010) Travel and the built environment: a meta- analysis. Journal of the American planning association, 76(3), 265-294. Ferreira, L. (1999) The changing nature of demand for public transport in Brisbane: unmet community transport needs-towards a rapid appraisal methodology, Working Paper 3. Field, A. (2013) Discovering statistics using IBM SPSS statistics, Sage.

References 319

Fotheringham, A. S. (1988) Note—Consumer Store Choice and Choice Set Definition. Marketing Science, 7(3), 299-310. Frank, L., Bradley, M., Kavage, S., Chapman, J. & Lawton, T. K. (2008) Urban form, travel time, and cost relationships with tour complexity and mode choice. Transportation, 35(1), 37-54. Frank, L. & Pivo, G. (1994a) Impacts of Mixed Use and Density on Utilization of Three Modes of Travel: Single-Occupant Vehicle, Transit and Walking. Transportation Research Record, 1466, 44-52. Frank, L. & Pivo, G. (1994b) Impacts of mixed use and density on utilization of three modes of travel: single-occupant vehicle, transit, and walking. Transportation research record, 44-44. Friedman, B., Gordon, S. P. & Peers, J. B. (1994) Effect of neotraditional neighborhood design on travel characteristics. Gachanja, J. N. (2010) Towards integrated land use and transport modelling: Evaluating accuracy of the four step transport model-the case of Istanbul, Turkey. Enschede, The Netherlands: International Institute for Geo- Information Science and Earth Observation (PhD Dissertation). Gautschi, D. (1981) specification of patronage models for retail centre choice. Journal of Marketing Research 18(2), 162-174. Getis, A. & Getis, J. (1966) Christaller's central place theory. Journal of Geography, 65(5), 220-226. Geurs, K. T. & Van Wee, B. (2004) Accessibility evaluation of land-use and transport strategies: review and research directions. Journal of Transport geography, 12(2), 127-140. Ghosh, A. (1984) Parameter nonstationarity in retail choice models. Journal of Business Research, 12(4), 425-436. Ghosh, A. & Mclafferty, S. (1984) A model of consumer propensity for multipurpose shopping. Geographical Analysis, 16(3), 244-249. Giuliano, G. (2005) Low income, public transit, and mobility. Transportation Research Record: Journal of the Transportation Research Board, 1927(- 1), 63-70. Goodman, R. & Coiacetto, E. (2009) Retail form in Melbourne and Brisbane: a preliminary investigation into the differences between the two cities. Guy, C. (1987) Recent advances in spatial interaction modelling: an application to the forecasting of shopping travel. Environment and Planning A, 19(2), 173-186. Habib, K. M. N. & Kockelman, K. (2008) Modeling choice of residential location and home type: recent movers in Austin Texas. 87th Annual Meeting of the Transportation Research Board, Washington, DC. Citeseer. Hägerstraand, T. (1970) What about people in regional science? Papers in regional science, 24(1), 7-24.

References 320

Handy, S. (1992) Regional versus local accessibility: neo-traditional development and its implications for non-work travel. Built Environment, 18(4), 253-267. Handy, S. (1996a) Methodologies for exploring the link between urban form and travel behavior. Transportation Research Part D: Transport and Environment, 1(2), 151-165. Handy, S. (1996b) Understanding the link between urban form and nonwork travel behavior. Journal of planning education and research, 15(3), 183. Handy, S. (2002) Travel behaviour--land use interactions: An overview and assessment of the research. In: In perpetual motion: Travel behavior research opportunities and application challenges. Hanink, D. M. (1996) Principles and applications of economic geography: economy, policy, environment, John Wiley & Sons. Hanson, S. (1997) Ten geographic ideas that changed the world, Rutgers University Press. Hanson, S. & Hanson, P. (1981) The travel-activity patterns of urban residents: dimensions and relationships to sociodemographic characteristics. Economic Geography, 57(4), 332-347. Hanson, S. & Schwab, M. (1987) Accessibility and intraurban travel. Environment and Planning A, 19(6), 735-748. Hausman, J. & Mcfadden, D. (1984) Specification tests for the multinomial logit model. Econometrica: Journal of the Econometric Society, 1219-1240. Hay, A. & Shaz, K. (2012) Parking and mode choice in Sydney: evidence from the Sydney Household Travel Survey. 35th Australasian Transport Research Forum. Hensher, D. A., Rose, J. M. & Greene, W. H. (2005) Applied choice analysis: a primer, Cambridge University Press. Hobbs, F. (1979) Traffic Planning And Engineering, Headington Hill Hall. Oxford OX3 OBW, England. Holtzclaw, J. (1994) Using Residential Patterns and Transit To Decrease Auto Dependence and Costs. Hu, P. & Pooler, J. (2002) An empirical test of the competing destinations model. Journal of geographical systems, 4(3), 301-323. Hubbard, R. (1978) A review of selected factors conditioning consumer travel behavior. Journal of Consumer Research, 1-21. Huff, D. L. (1963) A probabilistic analysis of shopping center trade areas. Land economics, 81-90. Iacono, M., Krizek, K. J. & El-Geneidy, A. (2010) Measuring non-motorized accessibility: issues, alternatives, and execution. Journal of Transport Geography, 18(1), 133-140.

References 321

Ibrahim, M. F. (2002) Disaggregating the travel components in shopping centre choice: an agenda for valuation practices. Journal of Property Investment & Finance, 20(3), 277-294. Innes, J. D., Ircha, M. C. & Badoe, D. A. (1990) Factors affecting automobile shopping trip destinations. Journal of Urban Planning and Development, 116(3), 126-136. Jepson, E. J. (2001) Sustainability and planning: diverse concepts and close associations. Journal of planning literature, 15(4), 499-510. Jones, K. & Simmons, J. (1993) Location, Location, Location Kattiyapornpong, U. & Miller, K. E. (2011) Understanding travel behavior using demographic and socioeconomic variables as travel constraints. [Queensland University of Technology, School of Advertising, Marketing and Public Relations]. Keefer, L. E. (1966) Urban travel patterns for airports, shopping centers, and industrial plants. NCHRP Report, (24). Keough, S. M. & Shanahan, K. J. (2008) Scenario planning: toward a more complete model for practice. Advances in Developing Human Resources, 10(2), 166-178. Kiel, G. & Haberkern, G. (n.d.) Issues in shopping centre trade areas. Kitamura, R., Chen, C. & Narayanan, R. (1998) Traveler destination choice behavior: Effects of time of day, activity duration, and home location. Transportation Research Record: Journal of the Transportation Research Board, 1645(1), 76-81. Kitamura, R. & Kermanshah, M. (1984) Sequential model of interdependent activity and destination choices. Transportation research record, (987). Koenig, J.-G. (1980) Indicators of urban accessibility: theory and application. Transportation, 9(2), 145-172. Koppelman, F. & Pas, E. (1984) Estimation of disaggregate regression models of person trip generation with multiday data. Papers presented during the Ninth International Symposium on Transportation and Traffic Theory held in Delft the Netherlands, 11-13 July 1984. Koppelman, F. S. & Bhat, C. (2006a) A self instructing course in mode choice modeling: multinomial and nested logit models. US Department of Transportation, Federal Transit Administration, 31. Koppelman, F. S. & Bhat, C. (2006b) A self instructing course in mode choice modeling: multinomial and nested logit models. Prepared for US Department of Transportation Federal Transit Administration. Krizek, K. J. & Johnson, P. J. (2006) Proximity to trails and retail: effects on urban cycling and walking. Journal of the American Planning Association, 72(1), 33-42. Kulash, W., Anglin, J. & Marks, D. (1990) Traditional neighborhood development: will the traffic work?

References 322

Lamb, R. (2010 ) State Government Control of Planning Law in Queensland: A gradual creep towards domination? Leck, E. (2006) The Impact of Urban Form on Travel Behavior: A Meta- Analysis. Berkeley Planning Journal, 19, 37–58. Lee, M.-S. & Goulias, K. G. (1997) Accessibility indicators for transportation planning using GIS, Transportation Research Board. Lee, Y. & Mccracken, M. (2012) Centripetal and Centrifugal Movement: Shopping Centres in Denver, USA, and Brisbane, Australia. Urban Studies, 49(7), 1489-1506. Levinson, H. S. (1976) Urban travel characteristics. Transportation and Traffic Engineering Handbook, 138-206. Limanond, T., Niemeier, D. & Mokhtarian, P. (2005) Specification of a tour- based neighborhood shopping model. Transportation, 32(2), 105-134. Limanond, T. & Niemeier, D. A. (2004) Effect of land use on decisions of shopping tour generation: A case study of three traditional neighborhoods in WA. Transportation, 31(2), 153-181. Litman, T. (2005) Land use impacts on transport. Victoria Transport Policy Institute (www. vtpi. org). Litman, T. (2014) Transport Model Improvements, Improving Methods for Evaluating The Effects and Value of Transportation System Changes. Victorian Transport Policy Institute. Litman, T. (2015) Evaluating accessibility for transportation planning, Measuring People’s Ability to Reach Desired Goods and Activities. Victoria Transport Policy Institute, Victoria, Canada. Lord, J. D. (1988) Retail decentralisation and CBD decline in American cities, University of Stirling Institute for Retail Studies. Maat, C. (2000) Travel reduction ‘built in’: the role of land use planning. IN BANISTER, D. & MARSHALL, S. (Eds.) Encouraging Transport Alternatives: Good Practice in Reducing Travel. London, The Stationery Office. Maat, K., Van Wee, B. & Stead, D. (2005) Land use and travel behaviour: expected effects from the perspective of utility theory and activity-based theories. Environment and Planning B: Planning and Design, 32(1), 33- 46. Marshall, S. & Mclellan, A. (1998) Reducing the need to travel: A European guide to good practice. Policy, planning and sustainability. Proceedings of seminar B held at AET European Transport Conference, Loughborough University, England, 14-18 September 1998. Volume P421. Martin, B. V., Memmott, F. W. & Bone, A. J. (1961) Principles and techniques of predicting future demand for urban area transportation. Mass Inst Tech Dept Pub Wks Jt Hwy Res.

References 323

Mccarthy, P. S. (1980) A study of the importance of generalised attributes in shopping choice behaviour. Environment and Planning A, 12(11), 1269- 1286. Mcfadden, D. (1980) Econometric models for probabilistic choice among products. Journal of Business, S13-S29. Mcgeoch, C. A. (2011) 30 years of travel in Melbourne: 1978/79 and 2007/08. Australasian Transport Research Forum (ATRF), 34th, 2011, Adelaide, South Australia, Australia. Mcnally, M. G. & Ryan, S. (1992) A comparative assessment of travel characteristics for neo-traditional developments. University of California Transportation Center. Meurs, H. & Haaijer, R. (2001) Spatial Structure and Mobility. Transportation Research Part D: Transport and Environment, 6(6), 429-446. Miller, H. J. (1991) Modelling accessibility using space-time prism concepts within geographical information systems. International Journal of Geographical Information System, 5(3), 287-301. Mitchell Hess, P., Vernez Moudon, A. & Logsdon, M. (2001) Measuring land use patterns for transportation research. Transportation Research Record: Journal of the Transportation Research Board, (1780), 17-24. Mooi, E. & Sarstedt, M. (2011a) Cluster analysis, Springer. Mooi, E. & Sarstedt, M. (2011b) A Concise Guide to Market Research, The Process, and Methods Using IBM SPSS Statistics, Springer. Moran, A. (2007) Land Regulations, Housing Prices and Productivity. Agenda, 14(1), 35-50. Murdie, R. A. (1965) Cultural differences in consumer travel. Economic Geography, 41(3), 211-233. Nelson, D., Niles, J. & Hibshoosh, A. (2001) A new planning template for transit- oriented development, Mineta Transportation Institute, San José State University. Nerella, S. & Bhat, C. R. (2004) Numerical analysis of effect of sampling of alternatives in discrete choice models. Transportation Research Record: Journal of the Transportation Research Board, 1894(1), 11-19. Niles, J. & Nelson, D. (1999a) Measuring the Success of Transit-Oriented Development: Retail Market Dynamics and Other Key Determinants. Niles, J. & Nelson, D. (1999b) Measuring the Success of Transit-Oriented Development: Retail Market Dynamics and Other Key Determinants. APA National Planning Conference. Seattle, WA. Retrieved January. Norusis, M. (2008) SPSS 16.0 statistical procedures companion, Prentice Hall Press. O'sullivan, S. & Morrall, J. (1996) Walking distances to and from light-rail transit stations. Transportation research record: journal of the transportation research board, 1538(1), 19-26.

References 324

Olaru, D., Smith, N. & Peachman, J. (2005) Whereabouts from Monday to Sunday? Australasian Transport Research Forum (ATRF), 28th, 2005, sydney, new south wales, australia. Owens, P. M. (1993a) Neighborhood form and pedestrian life: Taking a closer look. Landscape and urban planning, 26(1-4), 115-135. Owens, P. M. (1993b) Neighborhood form and pedestrian life: Taking a closer look. Landscape and urban planning, 26(1), 115-135. Pacione, M. (2005) Urban geography: a global perspective, Psychology Press. Pas, E. (1984) The effect of selected sociodemographic characteristics on daily travel-activity behavior. Environment and Planning A, 16(5), 571-581. Patterson, Z., Kryvobokov, M., Marchal, F. & Bierlaire, M. (2010) Disaggregate models with aggregate data: Two UrbanSim applications. The Journal of Tranport and Land Use, 3(2), 5-37. Pikora, T., Giles-Corti, B. & Donovan, R. (2001) How far will people walk to facilities in their local neighbourhoods [?]. Australia: Walking the 21st century, international conference, 2001, perth, western australia, vol 3. Ploeger, J. & Baanders, A. (1995) Land Use and transport Planning in the Netherlands. PLANNING FOR SUSTAINABILITY. PROCEEDINGS OF SEMINAR B HELD AT THE 23RD EUROPEAN TRANSPORT FORUM, UNIVERSITY OF WARWICK, ENGLAND, SEPTEMBER 11-15, 1995. VOLUME P389. Porter, M. E. (2008) Competitive advantage: Creating and sustaining superior performance, Simon and Schuster. Prillwitz, J., Harms, S. & Lanzendorf, M. (2006) Impact of life-course events on car ownership. Transportation Research Record: Journal of the Transportation Research Board, 1985(-1), 71-77. Property Council of Australia (2007) Shopping centre facts. Property Council of Australia. Rapoport, A. (1987) Pedestrian street use: Culture and perception. Public streets for public use, 80-94. Recker, W. W. & Kostyniuk, L. P. (1978) Factors influencing destination choice for the urban grocery shopping trip. Transportation, 7(1), 19-33. Reilly, M. & Landis, J. (2002) The Influence of Built-Form and Land Use on Mode Choice. University of California Transportation Center Research Paper, 669. Reilly, W. J. (1931) The law of retail gravitation, WJ Reilly. Richards, M. G. (1974) Disaggregate simultaneous urban travel demand models: A brief introduction. Transportation, 3(4), 335-342. Richards, M. G. & Ben-Akiva, M. (1974) A simultaneous destination and mode choice model for shopping trips. Transportation, 3(4), 343-356. Ringland, G. & Owen, M. (2007) Scenario as a Tool in Management Development. Organizations and People 14(1).

References 325

Robinson, R. & Vickerman, R. (1976) The demand for shopping travel: a theoretical and empirical study. Applied Economics, 8(4), 267-281. Rouhani, O. M. (2009) Road privatization and sustainability. Sustainable Transportation: An International Perspective, Projections 9, MIT Journal of Planning. Roy, J. R. & Thill, J.-C. (2003) Spatial interaction modelling. Papers in Regional Science, 83(1), 339-361. Roy, J. R. & Thill, J.-C. (2004) Spatial interaction modelling, Springer. Ryan, M. (2008) BSTM Multi-Modal Model Development, Queensland Governemnt, Queensland Transport Main Roads. Ryan, S. & Mcnally, M. G. (1995) Accessibility of neotraditional neighborhoods: A review of design concepts, policies, and recent literature. Transportation Research Part A: Policy and Practice, 29(2), 87-105. Ryley, T. (2006) Use of non-motorised modes and life stage in Edinburgh. Journal of Transport Geography, 14(5), 367-375. Şchiopu, D. (2010) Applying TwoStep cluster analysis for identifuing bank customers' profile. Buletinul, 62, 66-75. Schneider, R. J. (2011) Understanding sustainable transportation choices: Shifting routine automobile travel to walking and bicycling. University of California Transportation Center. Scott, A. (2002 ) Shopping Centre Development Regulation Asia Pacific Shopping Centres Conference. Scott, D. M. (2006) Constrained Destination Choice Set Generation: Comparison of GIS-Based Approaches. Transportation research board 85th annual meeting. Segal, D. B. (1999) Retail trade area analysis: concepts and new approaches. Journal of Database Marketing, 6, 267-278. Shaz, K. & Corpuz, G. (2012) The 0km Journey to Work–Travel Behaviour of Sydney’s Home-based Workforce. Australasian Transport Research Forum (ATRF), 35th, 2012, Perth, Western Australia, Australia. Shih, M. Y., Jheng, J. W. & Lai, L. F. (2010) A Two-Step Method for Clustering Mixed Categroical and Numeric Data. Tamkang Journal of Science and Engineering, 13(1), 11-19. Shobeirinejad, M., Burke, M. & Sipe, N. (2012) Analysing retail Travel behaviour using an Australian dataset. Australasian Transport Research Forum (ATRF) Shopping Centre Council of Australia (2011) Productivity Commission Inquiry into the Economic Structure and Performance of the Australian Retail Industry, Shopping Centre Council of Australia. Simmons, J., Jones, K., Yeates, M. & Kamikihara, S. (1998) International comparisons of commercial structure and public policy implications. Progress in Planning, 50(4), 291ą313.

References 326

Sivakumar, A. (2007) Modelling transport: a synthesis of transport modelling methodologies. Imperial College of London. Sivakumar, A. & Bhat, C. R. (2007) Comprehensive, unified framework for analyzing spatial location choice. Transportation Research Record: Journal of the Transportation Research Board, 2003(1), 103-111. Song, Y., Merlin, L. & Rodriguez, D. (2013) Comparing measures of urban land use mix. Computers, Environment and Urban Systems, 42, 1-13. Spiggle, S. & Sewall, M. A. (1987) A choice sets model of retail selection. The Journal of Marketing, 97-111. Srinivasan, S. (2002) Quantifying spatial characteristics of cities. Urban Studies, 39(11), 2005-2028. Stead, D. & Marshall, S. (2001) The relationships between urban form and travel patterns. An international review and evaluation. EJTIR, 1(2), 113- 141. Stimson, R. J. & Avery, B. (1990) The Brisbane Plan-A City Strategy. Strategy paper. Stopher, P., Zhang, Y., Armoogum, J. & Madre, J.-L. (2011) National Household Travel Surveys: The Case for Australia. Australasian Transport Research Forum 2011 Proceedings. Swait, J. (2001) Choice set generation within the generalized extreme value family of discrete choice models. Transportation Research Part B: Methodological, 35(7), 643-666. Termansen, M., Mcclean, C. J. & Skov-Petersen, H. (2004) Recreational site choice modelling using high-resolution spatial data. Environment and Planning A, 36, 1085-1100. The Urban Transport Institute, T. (2010) South-East Queensland Travel Survey 2009, Queensland Transport and Main Roads. Thill, J.-C. (1992) Choice set formation for destination choice modelling. Progress in Human Geography, 16(3), 361-382. Timmermans, H. & Golledge, R. G. (1990) Applications of behavioural research on spatial problems II: preference and choice. Progress in Human Geography, 14(3), 311-354. Tracy, A. J., Su, P., Sadek, A. W. & Wang, Q. (2011) Assessing the impact of the built environment on travel behavior: a case study of Buffalo, New York. Transportation, 38(4), 663-678. Train, K. (2009) Discrete choice methods with simulation, Cambridge University Press. Transport and Main Roads (2014) Land Use and Public Transport Accessibility Index (LUPTAI). Tsai, C.-H. P., Mulley, C. & Clifton, G. (2012) The spatial interactions between public transport demand and land use characteristics in the Sydney

References 327

Greater Metropolitan Area. Road & Transport Research: A Journal of Australian and New Zealand Research and Practice, 21(4), 62. Un General Assembly (2012) The Future we want (Resolution adopted by the General Assembly on 27 July 2012). Vickerman, R. & Barmby, T. (1984) The structure of shopping travel: some developments of the trip generation model. Journal of Transport Economics and Policy, 109-121. Walmsley, D. & Weinand, H. (1990) Is Central Place Theory relevant to an understanding of retail provision and planning? Urban Policy and Research, 8(2), 69-75. Warner, L. (2013) The evolution of Brisbane retailing. Jones Lang LaSalle. Warnes, A. & Daniels, P. (1979) Spatial aspects of an intrametropolitan central place hierarchy. Progress in Human Geography, 3(3), 384. Wegener, M. (2004) Overview of land-use transport models. Handbook of transport geography and spatial systems, 5, 127-146. Whyte, W. H. (2012) City: Rediscovering the center, University of Pennsylvania Press. Wilson, A. G. (1998) Land-use/transport interaction models: Past and future. Journal of Transport Economics and Policy, 3-26. Yamashita, H., Fujii, T. & Itoh, S. (2006) The Development Of Diverse Suburban Activity Centres In Melbourne, Australia. APPLIED GIS, 2(2), 09.1-09.26. Yang, J., Sun, D., Du, Y. & Zhao, Y. (2009) The effect of sampling of alternatives on MNL models: An empirical analysis in the context of shopping-destination choice models. Computer Science and Information Technology, 2009. ICCSIT 2009. 2nd IEEE International Conference on. IEEE. Yigitcanlar, T. (2010) Rethinking Sustainable Development: Urban Management, Engineering, and Design, IGI Global. Ying He, S. (2006) Modeling destination choice for shopping using a GIS-based spatio-temporal framework: An investigation of choice set generation and scale effects, Master ofArts thesis, School ofGraduate Studies, McMaster University. Yrigoyen, C. C. & Otero, J. V. (1998a) Spatial interaction models applied to the design of retail trade areas. 38th Congress of the European Regional Science Association. Yrigoyen, C. C. & Otero, J. V. (1998b) Spatial interaction models applied to the design of retail trade areas. Zhao, J., Peters, A. & Rickwood, P. (2013) Effect of Raising Fuel Price on Reduction in Household Transport Greenhouse Gas Emissions: A Sydney Case Study. State of Australian Cities (SOAC). Sydney.

Appendix I

Appendix I 330

Table A1-1: Revealed results for the base model

Discrete choice (multinomial logit) model Dependent variable Choice Log likelihood function -1126.23581 Estimation based on N = 1033, K = 2 Inf.Cr.AIC = 2256.5 AIC/N = 2.184 Model estimated: Mar 16, 2015, 14:52:02 R2=1-LogL/LogL* Log-L fncn R-sqrd R2Adj Constants only must be computed directly Use NLOGIT ;...;RHS=ONE$ Response data are given as ind. choices Number of obs.= 1033, skipped 0 obs ------+------|Standard Prob. 95% Confidence CHOSEN| Coefficient Error z |z|>Z* Interval ------+------ASC1| .25552*** .07494 3.41 .0006 .10865 .40240 ASC2| -.02240 .08001 -.28 .7795 -.17921 .13441 ------+------Note: ***, **, * ==> Significance at 1%, 5%, 10% level.

Table A1-2: Revealed results for the selected MNL model

|-> NLOGIT ;lhs = CHOSEN, CSET, OPTIONS ;CHOICES = 1, 2, 3 ;CheckData ;Model: U(1,2,3) = ConsLC * DumLC + ConsMC * DumMC + P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_WESD * WESD + P_NVEHSD * NVEHSD + P_KIDSD * KIDSD + P_AGE1D * AGE1D + P_AGE3D * AGE3D + P_EXP2D * EXP2D + P_EXP3D * EXP3D + P_MAINA2D * MAINA2D + P_MAINA3D * MAINA3D + P_MAINA4D * MAINA4D + P_MAINA5D * MAINA5D$ ------

Discrete choice (multinomial logit) model Dependent variable Choice Log likelihood function -236.94251 Estimation based on N = 1033, K = 22 Inf.Cr.AIC = 517.9 AIC/N = .501 Model estimated: Mar 16, 2015, 14:52:08 R2=1-LogL/LogL* Log-L fncn R-sqrd R2Adj

Appendix I 331

Constants only must be computed directly Use NLOGIT ;...;RHS=ONE$ Response data are given as ind. choices Number of obs.= 1033, skipped 0 obs ------+------|Standard Prob. 95% Confidence CHOSEN| Coefficient Error z |z|>Z* Interval ------+------CONSLC| -3.58690 3.48620 -1.03 .3035 -10.41973 3.24593 CONSMC| -3.55936 3.67924 -.97 .3333 -10.77054 3.65182 P_DISTKM| -.21686*** .03081 -7.04 .0000 -.27726 - .15647 P_RETDEN| .00351 .00262 1.34 .1812 -.00163 .00865 P_POPDEN|-.19787D-05 .3811D-04 -.05 .9586 -.76674D-04 .72716D-04 P_PT| -.01221 .01129 -1.08 .2798 -.03434 .00993 P_CAR| .00997 .03328 .30 .7646 -.05526 .07519 P_WALK| -.00574** .00233 -2.46 .0139 -.01031 - .00117 P_BIKE| .01578 .01227 1.29 .1984 -.00827 .03984 P_ENT| .10471D-05 .1249D-05 .84 .4018 -.14009D-05 .34951D-05 P_GLAR| .20786D-04*** .6850D-05 3.03 .0024 .73606D-05 .34211D-04 P_WESD| -.06492* .03698 -1.76 .0791 -.13740 .00756 P_NVEHSD| -.23709 .19595 -1.21 .2263 -.62114 .14696 P_KIDSD| -.04738 .03413 -1.39 .1651 -.11428 .01952 P_AGE1D| .09299 .10303 .90 .3668 -.10894 .29492 P_AGE3D| .02077 .06454 .32 .7476 -.10572 .14726 P_EXP2D| .10520** .04728 2.22 .0261 .01252 .19787 P_EXP3D| .07781** .03640 2.14 .0325 .00647 .14915 P_MAINA2| -.21296* .11726 -1.82 .0694 -.44278 .01687 P_MAINA3| -.09563 .07097 -1.35 .1778 -.23473 .04346 P_MAINA4| -.09789 .06528 -1.50 .1337 -.22582 .03005 P_MAINA5| -.08816 .07876 -1.12 .2630 -.24253 .06620 ------+------Note: nnnnn.D-xx or D+xx => multiply by 10 to -xx or +xx. Note: ***, **, * ==> Significance at 1%, 5%, 10% level.

Appendix I 332

Table A1-3: Results for the Hausman-test including the unrestricted model and also both restricted models (Excluding the LC alternative and Excluding the MC alternative)

a)The results for unrestricted model run

|-> NLOGIT ;lhs = CHOSEN, CSET, OPTIONS ;CHOICES = 1, 2, 3 ;CheckData ;Model: U(1) = ConsLC + P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_GLAR * GLAR + P_ENT * ENTROPY + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_BIKE * BIKE / U(2) = ConsMC + P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_GLAR * GLAR + P_ENT * ENTROPY + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_BIKE * BIKE / U(3) = P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_GLAR * GLAR + P_ENT * ENTROPY + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_BIKE * BIKE $ ------

Discrete choice (multinomial logit) model Dependent variable Choice Log likelihood function -247.90772 Estimation based on N = 1033, K = 11 Inf.Cr.AIC = 517.8 AIC/N = .501 Model estimated: Mar 16, 2015, 15:12:16 R2=1-LogL/LogL* Log-L fncn R-sqrd R2Adj Constants only must be computed directly Use NLOGIT ;...;RHS=ONE$ Chi-squared[ 9] = 1756.65619 Prob [ chi squared > value ] = .00000 Response data are given as ind. choices Number of obs.= 1033, skipped 0 obs ------+------|Standard Prob. 95% Confidence CHOSEN| Coefficient Error z |z|>Z* Interval ------+------CONSLC| -3.47419 3.40121 -1.02 .3070 -10.14043 3.19205 P_DISTKM| -.26677*** .01731 -15.41 .0000 -.30070 - .23284 P_RETDEN| .00333 .00256 1.30 .1923 -.00168 .00834 P_POPDEN|-.11643D-04 .3685D-04 -.32 .7520 -.83866D-04 .60580D-04

Appendix I 333 P_GLAR| .20032D-04*** .6704D-05 2.99 .0028 .68928D-05 .33171D-04 P_ENT| .98660D-06 .1220D-05 .81 .4187 -.14046D-05 .33778D-05 P_PT| -.01347 .01098 -1.23 .2200 -.03499 .00805 P_CAR| .00990 .03263 .30 .7616 -.05405 .07385 P_WALK| -.00600** .00233 -2.57 .0100 -.01056 - .00143 P_BIKE| .01779 .01218 1.46 .1442 -.00608 .04166 CONSMC| -3.37436 3.59311 -.94 .3477 -10.41674 3.66801 ------+------Note: nnnnn.D-xx or D+xx => multiply by 10 to -xx or +xx. Note: ***, **, * ==> Significance at 1%, 5%, 10% level.

b) The variance and covariance measures for unrestricted model

|-> MATRIX; J1 = [0,1,0,0,0,0,0,0,0,0,0 / 0,0,1,0,0,0,0,0,0,0,0 / 0,0,0,1,0,0,0,0,0,0,0 / 0,0,0,0,1,0,0,0,0,0,0 / 0,0,0,0,0,1,0,0,0,0,0 / 0,0,0,0,0,0,1,0,0,0,0 / 0,0,0,0,0,0,0,1,0,0,0 / 0,0,0,0,0,0,0,0,1,0,0 / 0,0,0,0,0,0,0,0,0,1,0] $ |-> MATRIX; Bu = J1 * B; Vu = J1 * VARB * J1'$

c) Removal of LC alternative and the model run results

Appendix I 334

|-> REJECT; DumLC=1 $ |-> REJECT; OPTIONS = 1 |-> NLOGIT ;lhs = CHOSEN, CSET2, OPTIONS2 ;CHOICES = 2, 3 ;CheckData ;Model: U(2) = ConsMC + P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_GLAR * GLAR + P_ENT * ENTROPY + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_BIKE * BIKE / U(3) = P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_GLAR * GLAR + P_ENT * ENTROPY + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_BIKE * BIKE $ ------

Discrete choice (multinomial logit) model Dependent variable Choice Log likelihood function -72.02353 Estimation based on N = 625, K = 10 Inf.Cr.AIC = 164.0 AIC/N = .262 Model estimated: Mar 16, 2015, 15:22:00 R2=1-LogL/LogL* Log-L fncn R-sqrd R2Adj Constants only must be computed directly Use NLOGIT ;...;RHS=ONE$ Chi-squared[ 9] = 722.30851 Prob [ chi squared > value ] = .00000 Response data are given as ind. choices Number of obs.= 1033, skipped 408 obs ------+------|Standard Prob. 95% Confidence CHOSEN| Coefficient Error z |z|>Z* Interval ------+------CONSMC| -5.94894 8.87077 -.67 .5025 -23.33533 11.43746 P_DISTKM| -.25923*** .02883 -8.99 .0000 -.31573 - .20272 P_RETDEN| -.01900** .00961 -1.98 .0480 -.03784 - .00017 P_POPDEN|-.67059D-04 .5862D-04 -1.14 .2526 -.18195D-03 .47834D-04 P_GLAR| .00012*** .3862D-04 3.09 .0020 .00004 .00020 P_ENT| .16014D-05 .3011D-05 .53 .5948 -.42993D-05 .75021D-05 P_PT| .01305 .02063 .63 .5270 -.02738 .05348

Appendix I 335 P_CAR| -.00670 .07330 -.09 .9271 -.15037 .13696 P_WALK| -.00659 .00653 -1.01 .3127 -.01939 .00621 P_BIKE| -.03203 .04646 -.69 .4906 -.12310 .05903 ------+------Note: nnnnn.D-xx or D+xx => multiply by 10 to -xx or +xx. Note: ***, **, * ==> Significance at 1%, 5%, 10% level. ------

d) The variance and covariance measures for restricted model

|-> MATRIX; J2 = [0,1,0,0,0,0,0,0,0,0 / 0,0,1,0,0,0,0,0,0,0 / 0,0,0,1,0,0,0,0,0,0 / 0,0,0,0,1,0,0,0,0,0 / 0,0,0,0,0,1,0,0,0,0 / 0,0,0,0,0,0,1,0,0,0 / 0,0,0,0,0,0,0,1,0,0 / 0,0,0,0,0,0,0,0,1,0 / 0,0,0,0,0,0,0,0,0,1] $ |-> MATRIX; Br = J2 * B; Vr = J2 * VARB * J2'$

e) Measuring the model test-statistic

f) Removal of MC alternative and the model run results

|-> REJECT; DumMC=1 $ |-> REJECT; OPTIONS = 2

Appendix I 336 |-> NLOGIT ;lhs = CHOSEN, CSET3, OPTIONS3 ;CHOICES = 1, 3 ;CheckData ;Model: U(1) = ConsLC + P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_GLAR * GLAR + P_ENT * ENTROPY + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_BIKE * BIKE / U(3) = P_DISTKM * DISTKM + P_RETDEN * RETDEN + P_POPDEN * POPDEN + P_GLAR * GLAR + P_ENT * ENTROPY + P_PT * PT + P_CAR * CAR + P_WALK * WALK + P_BIKE * BIKE $ ------Discrete choice (multinomial logit) model Dependent variable Choice Log likelihood function -111.83442 Estimation based on N = 724, K = 10 Inf.Cr.AIC = 243.7 AIC/N = .337 Model estimated: Mar 16, 2015, 15:38:14 R2=1-LogL/LogL* Log-L fncn R-sqrd R2Adj Constants only must be computed directly Use NLOGIT ;...;RHS=ONE$ Chi-squared[ 9] = 768.28601 Prob [ chi squared > value ] = .00000 Response data are given as ind. choices Number of obs.= 1033, skipped 309 obs ------+------|Standard Prob. 95% Confidence CHOSEN| Coefficient Error z |z|>Z* Interval ------+------CONSLC| 5.43947 5.77256 .94 .3460 -5.87454 16.75347 P_DISTKM| -.25356*** .02462 -10.30 .0000 -.30182 - .20530 P_RETDEN| .45294D-04 .00381 .01 .9905 -.74315D-02 .75221D-02 P_POPDEN|-.24172D-04 .6314D-04 -.38 .7018 -.14792D-03 .99572D-04 P_GLAR| .26913D-04*** .9506D-05 2.83 .0046 .82828D-05 .45544D-04 P_ENT|-.22574D-05 .2065D-05 -1.09 .2744 -.63056D-05 .17909D-05 P_PT| -.00115 .01619 -.07 .9436 -.03289 .03060 P_CAR| .00283 .04482 .06 .9496 -.08501 .09068 P_WALK| -.00317 .00357 -.89 .3757 -.01017 .00384 P_BIKE| .00409 .01741 .24 .8141 -.03002 .03821 ------+------Note: nnnnn.D-xx or D+xx => multiply by 10 to -xx or +xx. Note: ***, **, * ==> Significance at 1%, 5%, 10% level. ------

|-> MATRIX; J2 = [0,1,0,0,0,0,0,0,0,0 / 0,0,1,0,0,0,0,0,0,0 /

Appendix I 337

0,0,0,1,0,0,0,0,0,0 / 0,0,0,0,1,0,0,0,0,0 / 0,0,0,0,0,1,0,0,0,0 / 0,0,0,0,0,0,1,0,0,0 / 0,0,0,0,0,0,0,1,0,0 / 0,0,0,0,0,0,0,0,1,0 / 0,0,0,0,0,0,0,0,0,1] $ |-> MATRIX; Br = J2 * B; Vr = J2 * VARB * J2'$

g) Measuring the model test-statistic

|-> MATRIX; bd = Bu-Br; Vd = Vr -Vu $

|-> MATRIX; vdinv = [vd] $

|-> MATRIX; list; q = bd'*vdinv*bd $

Q| 1

------+------

1| 5.11348

|-> CALC; list; p = 1-chi(q,9) $

[CALC] P = .8243057

2 Appendix Il

Appendix Il 340

A2.2 Brisbane land use and transport model – BSTM-MM

A2.2.1 Home Based Shopping Trip Model

The number of demographic characteristics, trip characteristics and land use characteristics used in the HBS mode choice model are shown in table A2-1.

Additional household characteristics were synthesised by appropriate combinations of these basic characteristics. For example vehicles per household divided by persons per household or number of adults per household minus number of vehicles per household. This provides an indication of competition for car use or the maximum number of people in the household who could drive simultaneously. This provides the potential to address some aspects of the issue of captivity results in the circumstances where the individual choice set contains only one, or at most a small number of alternatives, as it has defined within the context of the travel demand modelling.

The trip time and cost as part of the trip attributes (Table A2-1) for each mode type included variables such as the in-vehicle travel time, parking cost, public transport fare, etc. Employment density and retail employment density (based on the attraction end of the trip) were selected and applied as the land use attributes for each individual trip.

The model with results included the following parameters: The highway travel time, parking cost, highway cost and vehicle availability for car driver mode choice, the highway travel time and vehicle availability for car as passenger mode choice, the mode-specific constants for park and ride and kiss and ride mode choice, travel time, waiting time, access time, vehicle availability and the employment density at the attraction end for the walk to public transport mode choice and finally selection of cycle or walk were found to be affected by travel times for the same. The cost parameter had no relevance; which may be due to shopping trip lengths being very small.

Appendix Il 341

Table A2-1: List of all different attributes considered in the developed mode choice logit models

List of household characteristic considered in the model and attribute notation Household Characteristic Attributes Notation of the Attribute Persons per household PERS Adults per household ADUL Workers per household WORK White collar workers per household WWRK Blue collar workers per household BWRK Licence holders per household LIC Tertiary students per household TERT School students per household SCHL Vehicles per household VEHS List of trip attributes considered in the model and attribute notation Travelling Mode Attributes Notation of the Attributes Car as Driver In-vehicle travel time TTCAD (CAD) Vehicle operating cost TCCAD Parking cost at the attraction end PCCAD Mode-specific constant CCAD Car as Passenger In-vehicle travel time TTCAP (CAP) Mode-specific constant CCAP Walk to Public In-vehicle (PT) travel time TTW2PT Transport (W2PT) Public transport fare TCW2PT Waiting time WTW2PT Combined walk access and egress time to and ATW2PT from the public transport stop or station Mode-specific constant CW2PT Park & Ride In-vehicle (PT) travel time TTPR (PR) Public transport fare TCPR Waiting time WTPR Car travel time between the public transport stop CATPR or station and the production end of the trip Walk travel time between the public transport stop WATPR or station and the attraction end of the trip Mode-specific constant CPR Kiss & Ride In-vehicle travel time TTKR (KR) Public transport fare TCKR Waiting time WTKR Car travel time between the public transport stop CATKR or station and the production end of the trip Walk travel time between the public transport stop WATKR or station and the attraction end of the trip Mode-specific constant CKR Cycle all-the-way Cycling time TTC (C) Mode-specific constant CC Walk all-the-way Walking time TTW (W) Mode-specific constant CW List of employment density attributes considered in the mode choice model and attribute notation Household Characteristic Attributes Notation of the Attribute Employment density (jobs per hectare) EMPLDENS Retail employment density (retail jobs per hectare) REMPLDEN

Appendix Il 342

A2.3 City wide Scenario planning

The first sets of scenarios are focusing on clustering and decentralizing retail jobs within BSD. Possible shifts in the number of retail trips, mode choice decisions and changes in overall Vehicle Kilometre Travelled (VKT), Public Transport Kilometre Travelled (PTKT) and number of PT passengers’ boarding and alighting will be investigated based on the model run results. Table A2-2 describes details of the two proposed scenarios.

Table A2-2: Macro scale scenarios within BSD (city-wide)

Scenario Increase in the No. of Ret- Decrease in the No. of Ret- Jobs % Jobs %

Clustering 50% for all the Super- 30% for the Regional centres Regional and Major-Regional 15% for the Sub-Regional centres centres

Decentralizing 50% for the Sub-Regional 40% for Super-Regional and centres Major-Regional &

& some small Regional some very large Regional centres centres

1-Clustering Scenario

The first scenario looks at bigger clusters of retail in future around the super- regional and major-regional centres. It considers the situation in which these currently large regional centres continue to grow and cover a much bigger number of retail jobs by 50 percent. Subsequently, the catchment areas of the centres will also expand while the function of other smaller regional or sub- regional centres is assumed to get affected considerably.

As it had been mentioned in the literature review, shopping centres in Australia follows the same trend similar to Christallers’ central place theory. Larger centres cover almost all the activities and services within the smaller centre types. It means that larger catchment for the big super-regional and major- regional centres will affect the service area of the other centre types and reduce their supremacy. Consequently, the size and number of their retail services and

Appendix Il 343 jobs will shrink. That being said, a 30 percent reduction in the number of retail jobs for regional centres and 15 percent for sub-regional centres are predicted in this scenario. This is done not only to show the impacts of the catchment but also to keep the balance for the total number of jobs in the model.

2-Decentralizing Scenario

The second scenario is focusing on the decentralization policy and tries to look at conditions in which people are not very much attracted to large centres and big department stores. They might mostly prefer to go to smaller centres to resist the heavy traffic around large centres and have easier parking options rather than the limited number of parking spaces at the large centres which keeps them marching around for long time. In this scenario, smaller centres are being preferred over the larger ones for most of their shopping purposes. These smaller centres will let the customers to get in and out more easily and quickly while they will yet experience a good access to large supermarkets, food courts, and all the other essential products that they require.

These centres would be categorized more as a sub-regional and small regional centre rather than large super-regional centres. In this scenario, the big centres comprise almost 40 percent of retail job loss in the large super-regional centres. Instead an almost 50 percent increase within the smaller regional and sub- regional centres will make them more attractive and at the same time restricted in a more reasonable size which are much easier to be found in closer distances from customers’ houses.

Appendix Il 344

Fig A2-1: Location and distribution of shopping centres within BSD

Appendix Il 345

A2.4 Regional Scenario planning

As discussed previously, the second sets of scenarios focus on part of the BSD with 467 BSTM zones located in the south-east part of the Brisbane river (Fig A2-2) including both parts of inner and outer suburbs in addition to 28 BSTM zones outside of Brisbane city council boundary which makes the total number of 495 zones.

Fig A2-2: The location of the regional study area with almost 500 traffic zones from BSTM

There are a number of factors which affected the selection of this area as the focus of this study including:  The proximity of the area to the Central Business District (CBD) which comprises more of the smaller BSTM zones in comparison to the farther zones from CBD with a much bigger footprint which will create more

Appendix Il 346

accurate results. The larger the zones get, the less precise the results will be since as discussed before, the model aggregates all the information into the centroid of the zones.  The presence of all four types of major centres - Super-Regional, Major- Regional and Regional centres – including Westfield Carindale, Westfield Garden City and Capalaba Park Shopping Centre and Capalaba Central Shopping Centre plus a number of sub-regional and neighbourhood centres in the area which help us to reflect possible variations in the hierarchy of retail centres’ characteristics.  The location of a number of major transport corridors and also the major bus stations, identified and planned for future TOD development in the city plan.  The existing current plans for future population increase including infill and new development within this area such as in Rochedale which gives us the chance to have more meaningful scenarios for future  The location of major cut-off lines such as bus way, river, railway which separate different neighbourhood from each other and can significantly affect the trip destination decisions of the trip makers.  The origin and destination of the retail trip that has ended up into big shopping centres extracted from the Household Travel Survey Data 2009 (a large number are located within the study boundary)

The results for the model run will be studied within the boundaries of the study area. Different aspects of the shopping trips similar to what was studied in the BSD scale including percentage of retail trips with various mode types, Vehicle Kilometre Travelled (VKT), Public Transport Kilometre Travelled (PTKT), number of boarding and alighting will be then extracted and measured.

To keeps the total numbers of population and retail jobs unaffected two different approaches are selected and applied, one within and the other outside of the defined boundary which will be discussed in detail later in the chapter.

Before developing the scenarios, in order to get a better sense of where the location of retail and population is going to go and where might be the possible locations for future growth within almost 15 years from now, it is good to look at

Appendix Il 347 the available data (projected in the BSTM model) for various years and make sense of possible changes in the location and distribution density of retail jobs and population.

The BSTM model has developed for a number of years in future including the 2016, 2021, 2026 and 2031 and contains future projections for population and employment as well as changes in the transport system. Two scenarios including the 2016 as the base model and 2031 as the planed future scenario were selected and are discussed in here.

A2.4.1 Future predicted plans (2016 – 2031) and the scale of the shifts

Fig A2-3 shows the density of retail jobs in 2016. It is noticeable that in many cases the density of retail is higher around the existing shopping centres in different levels of hierarchy from Super-Regional to neighbourhood and also around the current and future growth corridors next to major bus routes, railways and major roads. This is obviously in regards to the current planning policies supporting the in centre development approach and limiting out of centre development.

It should be remembered that as discussed in chapter 7, the various datasets extracted from the SCD 2011 and BSTM model are not perfectly matched but are currently the best available types of data that can be available to us.

While the retail density is showing to be very high in the areas close to the river it should be noticed that these areas are mostly the location of the warehouses related to the ship industry. Likewise, transport zones within close distance from a CBD are obviously experiencing a different type of growth and need to follow different planning strategies in future. As we go farther from the CBD and the Brisbane city LGA the area of the BSTM zones are getting bigger so the average increase in the retail job density is actually representing a larger number of retail jobs in those areas.

Appendix Il 348

Fig A2-3: Retail density for 2016 within the regional study area

When it comes to the population density, parts of the study area mostly located closer to the transportation corridors (main roads, major public transport routes and railway) are experiencing a higher density of population as it can be seen in Fig A2-4.

It is also quite obvious that centres are surrounded by the areas of higher population density. Wherever the population density is getting lower, there is no sign of shopping centres. Super-Regional and Regional centres including Westfield Garden City, Westfield Carindale and Capalaba centres followed by Sub-Regional centres such as Mount Gravatt Plaza, Sunny bank Plaza Shopping centre, Wynnum Plaza Shopping Centre, etc. are mostly located within the areas of higher population density outside the smaller CBD or Inner Brisbane area and are usually servicing quite a large catchment.

Appendix Il 349

Fig A2-4: Population Density 2011 within the regional study area

Therefore it is quite important to notice that even with a larger number of population living around these centres, the percentage of public and active transport trips are still very much low. This might be relevant to the fact that the critical mass hasn’t yet occurred in these areas in a way it needs to happen or it might be relevant to a large number of other factors regarding people’s mode preferences towards private cars, etc. It is important to study any possible considerable improvement in other modes apart from private automobiles if the population density continues to increase around these centres following some highly pitched approaches in the previous and new city plan including the Transit Oriented Development strategies. The second scenario is focusing on the population living within each zone and the future increase in the household number in some areas and the possible impact on the retail travel behaviour of

Appendix Il 350 people.Besides, considering the existing policies and proposed directions for a more sustainable transportation-land use system, it is important to compare the future distribution of retail jobs as it is reported in the BSTM for 2031 with the current 2016 base model scenario and see where the future growth in retail is heading.

Fig A2-5 compares the retail job density distribution for 2016 and 2031. Based on the input data for BSTM model, 5,646 more retail jobs will be created within the area between the periods of 2016 to 2031. It doesn’t only include the number of added jobs, it also include the total number of reductions and increases in the number of retail jobs in various zones in the area. The differences are represented in three categories for retail employment density defined as no change areas which means the number of retail job density hasn’t changed between this period, more than average (more than 34 retail jobs per km2) and less than average (less than 34 retail jobs per km2). 34 retail jobs per km2 is the average number of retail jobs per km2. it is resulted from the reduction of number of retail jobs per km2 for 2031 and 2016 divided by the area of each zone followed by the sum of them all divided by the total number of 495 transport zones.

The maps reveals highly possible future retail job growth areas and gives us a good understanding to make a comparison between the current situations and possible future trends as predicted by the TMR planning predictions for 2031. As it can be seen areas alongside the river is showing considerable shifts in the number of retail jobs followed by some areas around Capalaba regional centre, Garden city Sub-regional centre and the sub-regional Mount Gravatt plaza centre. There are also other areas mostly close to an existing neighbourhood centre which are predicted to experience a substantial increase in the number of their retail jobs. Based on the above mentioned analysis and what has been discussed previously on the possible future direction of retail jobs and population distribution centres around them, different scenarios will be explained in this stage.

Appendix Il 351

Fig A2-5: Predicted areas of retail job increase from 2016 to 2031

The same plan has also been developed for the future areas of population growth as intended by government planners presented in Fig A2-6. The areas of population growth are colour-coded based on similar legend of less than average growth, more than average and areas with no change in the number of population. Looking closer the number of areas with no change is very limited and only contains 18 zones.

Apart from that other zones are mostly experiencing a less than average shifts in their population density. Areas around Capalaba, Garden City and Carindale shopping centres are experiencing a higher than average level of population growth. Sub-regional centres including Stockland Cleveland, Wynnum Plaza shopping centre, Sunnybank Plaza shopping centre, Mt Gravatt Plaza, Cannon Hill K-Mart Plaza are coming as the next group of centres experiencing some

Appendix Il 352 population growth in their vicinity. Interestingly other mostly scattered areas with larger population density than average are located alongside the transport corridors or in some cases close to some already existing neighbourhood centres.

Fig A2-6: Predicted areas of population growth from 2016 to 2031

A2.4.2 Proposed scenarios for the regional scale

Based on the above mentioned analysis, various zones are selected as possible areas of growth for future retail employment for our scenarios. Fig A2-7 and Fig A2-8 show various zones around large centres including super-regional, major-regional and sub-regional centre which are being selected for the shift in the number of retail job opportunities and the number of the people living in these areas. A number of scenarios have been developed within the regional

Appendix Il 353 boundaries following the same concept of clustering and decentralizing of retail jobs as well as increase in the number of residents living within the zones around existing centres that will be explained below.

Fig A2-7: Selected traffic zones for the shifts in the future scenarios Retail employment

1- Centralizing Scenario

Following the same trend in the macro scale scenario, the first sets of the regional scale scenarios is looking at the higher level of concentration for retail. This scenario is considering the increase in the retail job number for big Super- Regional and Regional centres including Westfield Carindale, Westfield Garden City and Capalaba Park Shopping Centre and Capalaba Central Shopping Centre. There are two ways for developing the scenario including:

Appendix Il 354

 In one case as the centres get bigger, their trade area will get bigger as well and since all the larger centres are comprising all the products and services provided in smaller centres in Australia, therefore the smaller centre can’t compete with the larger ones and the number of jobs in the smaller sub-regional and neighbourhood centres will start to decrease.  In the second case the increase in the size of the large centres won’t be followed by the decrease in the surrounding other types of centres but since the model requires the same total number of retail jobs, the added numbers will be balanced by decreasing the number of retail jobs in farther zones close to the boundary of BSD with possibly no direct impact on the study area.

2- Decentralizing Scenario

The second sets of scenarios include the decentralizing scenarios being considered in two categories focusing on the smaller and more distributed types of centres in the region and its possible impacts on the travel behaviour of people:  The first groups of scenarios investigate further expansion in the size of the sub-regional centres and more reduction in the size of large super- regional centres.  The second group study the impacts of expanding the neighbourhood centres and reducing the size of the retail job numbers mostly in sub- regional centres which are more similar to these centre types in terms of services. It means larger number of small centres rather than the sub regional ones.

3- Population growth

The third group of scenarios related to the number of people living closer to the super regional, regional and sub-regional centres and how and if it might affect their travel behaviour.

A number of zones around these centres have been selected based on the previous analysis on the population growth in the area and will be tested in terms of their population increase (Fig A2-8).

Appendix Il 355

The population has only been balanced by decreasing the number of people in farther areas rather than moving the existing population closer or farther from the centres since there are lots of socio-demographic issues attached t the number of people in each zone that makes it quite hard to decide about their possible movements around the selected boundary.

Fig A2-8: Selected traffic zones for future scenarios population growth

A sensitivity analysis is done by considering 15%, 30% and 50% increase in the retail job numbers to see what would be the possible impacts of the shifts on the trip makers’ mode choice resulted in both cases. Therefore 6 different scenarios were developed (Table A2-3 ) and run for each case plus a population growth scenario limited only to 15% increase in the number of people living in the selected areas.

Appendix Il 356

Results were then extracted from the model and were compared. The comparison between model results showed a consistent trend in terms of the increase in the percentages of trips in each case. Therefore it was decided that only the 50% increase scenarios and their following results are going to be discussed in this chapter.

Besides, while both sets of scenarios including the ones in which the increase in the number of retail jobs for specific type of centres are followed with a decrease in the number of retail jobs in the other type of existing centres and the second group in which the number of retail jobs were balanced with changes in the number of retail jobs in farther areas with minimum or no impacts on the study area, are run and their results are presented here, only the first group will be explained in details. This is due to a number of reasons including the fact that retail jobs can’t only increase in an area by 50 percent with no justification.

The first sets of scenarios are explained briefly in here:

Large C Scenario: 50 percent increase in number of retail jobs for large centres followed by 25% decrease in the number of retail jobs for Sub-Regional or medium large centres. This means larger clustered oversize centres such as Westfield Garden City with almost 3000 to 3500 retail jobs and smaller types of centres similar to medium size neighbourhood shopping outlets with almost 200 to 400 retail jobs.

SubR C Scenario: 50 percent increase in the number of retail jobs in Sub- Regional centres which brings the number of retail jobs to around 800 to 1200, followed by 30% decrease in the number of jobs for Super Regional and large Regional centres with almost 1200 to 1500 retail jobs. This scenario will represents centres almost similar to smaller regional centres distributed more commonly around the city with higher possible accessibility level for the population rather than having the unbeatable fewer super large Regional ones

Neigh C Scenario: 50 percent increase in the Neighbourhood centre job numbers which leaves them with 200 to 1200 retail jobs and requires almost 40% reduction in the retail job availability within Sub Regional and Super Regional centres and the final number of 500 to 1500 jobs. This makes the Sub-

Appendix Il 357

Regional centres almost the same size as Neighbourhood type of destinations besides moderately big Regional centres almost half size of the current large Super Regional ones.

All the reported results are then compared to the 2016 base scenario:

(New scenario-Base scenario)/ Base scenario*100

While the above-mentioned scenarios are focusing on the possible changes in the travel behaviour of people by making changes in the distribution of retail job locations in the area, the population scenario have also been developed to investigate how proximity to the centres and higher level of accessibility to the centres might affect people’s travel behaviour and shift the mode share percentages. All the results were also compared to the 2031 scenario proposed and developed by TMR as a trajectory to a more sustainable future transport system in the city.

Population Scenario: 15 percent increase in the number of people living around the large and medium size shopping centres (super-regional, regional and sub-regional centres) followed by a decrease in the number of population in the farther zones outside the boundary of the study area. This will increase the population by almost 17,000 around the larger centres.

2031 scenario: This scenario was developed by TMR for 2031 to improve a better and more sustainable transport system that increase the number of trips by public and active transport modes and decrease the reliance on private motor vehicles. Table A2-3 shows the results for the various proposed scenarios.

Appendix Il 358

Table A2-3: Proposed future scenarios to be tested using BSTM to better understand the travel behaviour of customer

Ret. Job No. Increase Ret. Job No. Decrease Within the Outside the Within the Outside the defined defined defined defined boundary*1 boundary*2 boundary boundary 1 Increase in the Re 15% 25% for sub- Job No. for super regional regional and regional centres 2 30% 35 % for sub regional and 10% for neighbourhood 3 50% 50 % for sub regional and 22% for neighbourhood 4 15 % 2.7 % 5 30 % 5.3 % 6 50 % 8.9 % 7 Increase in the Re 15 % 9 % - super- Job No. for sub regional and regional centres regional 8 30 % 17 % 9 50 % 30 % 10 15 % 1.5 % 11 30 % 3 % 12 50 % 5 % 13 Increase in the Re 15 % 25 % - sub Job No. for regional neighbourhood centres 14 30 % 52 % - sub regional

15 50 % 40 % - sub regional & 37 % - super- regional and regional 16 15 % 2.5 % 17 30 % 5.2 % 18 50 % 8.7 % *1within the defined boundary means that the number of jobs will be balanced in the boundary. Increasing of one is equal to decrease of the other one *2Outside of the boundary means changes in far zones which are believed to have none or very small impacts on the studied area so it means like only an increase in the area with no balancing within the study area

Appendix Il 359

A2.5 Neighbourhood scale Scenario planning

As previously discussed, in order to better understand the scale of changes resulted from each scenario, the results were also examined in the neighbourhood scale. The difference is that new scenarios were not defined and the only difference is that impacts from the previously made scenarios will be studied in a smaller area of 30 to 50 traffic zones to see if the significance of the changes is more considerable in smaller areas.

Fig A2-9: Neighbourhood boundaries 1 and 2

These further analyses will help us to see how effective the retail policies can be in the smaller scale. Two areas (Fig A2-9) are selected to be studied for the previously defined measures of number of shopping trips by each mode, and all

Appendix Il 360 trip type VKT, VHT,PTKT, number of passengers boarding and alighting form the public transport in the study area.

The first boundary 1 comprises the Westfield Garden City shopping centre, two sub-regional centres and almost 80 surrounding zones. Boundary 2 includes Capalaba Central shopping centre and Capalaba Park shopping centres, one sub-regional centre within the catchment area and its surrounding area with almost 30 BSTM zones.

The two areas were not only different in terms of the size of the transport zones that they contain (much bigger for boundary 2 compare to finely divided zones within boundary 1) but also have different socio demographic population living within and close to the catchment of the centres which might be the source of some variations in the travel behaviour of trip makers and the studied parameters. Fig A2-10 shows the different socio demographic characteristics of the people living in the area including total number of people under 17 per household, number of people in secondary study per household, number of employed people per household, number of dependent per household.

It can be seen from the maps that the boundary 1 area is having a larger number of households with dependent people living in them, while the aggregated results for the GW zones shows a lower than average condition in there. In terms of the number of employed people per household, the employment rate is higher in the boundary 2 area which might be a replication of more middle age fulltime types of people who are living with their family in this area compare to garden city with younger and more like student groups of people who are not having a permanent job.

Some other reasons behind the selection of these two boundaries is related to the fact that both centres in the study area have almost a good level of accessibility to public transport even though Garden city is having more variety and frequency of buses compare to Capalaba. Both centres are categorised as large type of regional centres (super regional and regional centres) with 9 and 11 major tenants and 60 and 80 thousands GLA.

Results are also compared with the projected retail employment and population increase projected within the model for 2031 scenarios.

Appendix Il 361

a) Average dependent per household b) Average employed per household

c) Average aged under 17 per household d) Average Secondary/Tertiary student per household

Fig A2-10: Socio-demographic characteristics of the study areas