Dr. Juan de Dios Ortúzar is Emeritus Professor in the School of Engineering at Pontificia Universidad Católica de Chile and also Key Researcher at Instituto Sistemas Complejos de Ingeniería (ISCI) and the BRT+ Centre of Excellence. He specializes in discrete choice modelling and survey design with particular focus on transport demand modelling and the valuation of transport externalities.

Dr. Luis G. Willumsen is an internationally recognised authority in transport and traffic modelling and has over 35 years of experience in this area. He previously lectured at Leeds University and University College London, and was also a Director of Steer before leaving in 2009 to set up his own independent practice. He is also Managing Partner of Nommon Solutions and Technologies, a company processing big data to provide location and mobility intelligence.

Preface xviii

About the Companion Website xxii

1 Introduction 1

1.1 Background 1

1.2 Models and Their Role 2

1.3 Characteristics of Transport Problems 3

1.3.1 Characteristics of Transport Demand 3

1.3.2 Characteristics of Transport Supply 4

1.3.3 A View of Transport Problems 6

1.3.4 A Simple Model 7

1.3.5 Classic and New Modes of Transport 9

1.4 Modelling and Decision-Making 9

1.5 Issues in Transport Modelling 12

1.5.1 General Modelling Issues 12

1.5.1.1 The Roles of Theory and Data 12

1.5.1.2 Model Assumptions 13

1.5.1.3 Model Specification 14

1.5.1.4 Model Calibration, Validation, and Use 15

1.5.1.5 Modelling, Forecasting, and Judgement 16

1.5.2 Aggregate and Disaggregate Modelling 17

1.5.3 Homo Sapiens and Homo Economicus 18

1.5.4 Cross-Section and Time Series 20

1.5.5 Revealed and Stated Preferences 21

1.6 The Structure of the Classic Transport Model 22

1.6.1 The Classic 4/5 Stage Model 22

1.6.2 Granularity 24

1.6.3 Macro, Meso, and Micro Models 27

1.7 Transport Planning and Uncertainty 27

1.8 Theoretical Basis Versus Expedience 31

1.9 Becoming a Better Modeller 33

Exercises 33

2 Data 35

2.1 Basic Sampling Theory 36

2.1.1 Statistical Considerations 36

2.1.1.1 Basic Definitions 36

2.1.1.2 Sample Size to Estimate Population Parameters 38

2.1.1.3 Obtaining the Sample 40

2.1.2 Practical Considerations in Sampling 43

2.1.2.1 The Implementation Problem 43

2.1.2.2 Finding the Size of Each Subpopulation 43

2.2 Errors in Modelling and Forecasting 44

2.2.1 Different Types of Error 45

2.2.1.1 Measurement Errors 45

2.2.1.2 Sampling Errors 46

2.2.1.3 Computational Errors 46

2.2.1.4 Specification Errors 46

2.2.1.5 Transfer Errors 47

2.2.1.6 Aggregation Errors 47

2.2.2 The Model Complexity/Data Accuracy Trade-off 48

2.2.3 Forecasting Errors 51

2.3 Basic Data-collection Methods 53

2.3.1 Practical Considerations 53

2.3.1.1 Length of the Study 53

2.3.1.2 Study Horizon 53

2.3.1.3 Limits of the Study Area 54

2.3.1.4 Study Resources 54

2.3.2 Types of Surveys 54

2.3.2.1 Survey Scope 55

2.3.2.2 Home Interview Travel Surveys 57

2.3.2.3 Other Important Types of Surveys 66

2.3.3 Data Correction, Expansion, and Validation 68

2.3.3.1 Data Correction 69

2.3.3.2 Imputation Methods 71

2.3.3.3 Sample Expansion 72

2.3.3.4 Validation of Results 72

2.3.4 Longitudinal Data Collection 73

2.3.4.1 Basic Definitions 73

2.3.4.2 Representative Sampling 74

2.3.4.3 Sources of Error in Panel Data 75

2.3.4.4 Relative Costs of Longitudinal Surveys 76

2.3.5 Travel Time Surveys 76

2.3.6 Digital Data Sources 77

2.4 Stated Preference Surveys 79

2.4.1 Introduction 79

2.4.1.1 Contingent Valuation and Conjoint Analysis 79

2.4.1.2 Stated Choice Methods 81

2.4.2 The Survey Process 83

2.4.2.1 Clarifying Study Objectives and Defining Objects of Interest 84

2.4.2.2 Defining Experimental Assumptions 86

2.4.2.3 Generating the Experimental Design 92

2.4.2.4 Conduct Post Design Generation Testing 97

2.4.2.5 Conduct Questionnaire 98

2.4.2.6 Nothing is Important 99

2.4.2.7 Realism and Complexity 100

2.4.2.8 Use of Computers in SP Surveys 101

2.4.2.9 Quality Issues in Stated Preference Surveys 102

2.4.3 Case Study Example 103

2.4.4 Limitations of Stated Preference Methods 115

Exercises 115

3 Zones and Networks 119

3.1 Zoning Design 120

3.2 Road Network Representation 122

3.2.1 Traffic Flow 123

3.2.2 Network Details 123

3.3 Link Properties and Functions 125

3.3.1 Link Properties 125

3.3.2 Network Costs 126

3.3.3 Definitions and Notation 127

3.3.4 Speed-Flow and Cost-Flow Curves 127

3.3.5 Public Transport Networks 131

Exercises 132

4 Trip Generation Modelling 133

4.1 Introduction 134

4.1.1 Some Basic Definitions 134

4.1.2 Characterisation of Journeys 135

4.1.2.1 By Purpose 135

4.1.2.2 By Time of Day 135

4.1.2.3 By Person Type 136

4.1.3 Factors Affecting Trip Generation 136

4.1.3.1 Personal Trip Productions 137

4.1.3.2 Personal Trip Attractions 137

4.1.3.3 Freight Trip Productions and Attractions 137

4.1.4 Growth-Factor Modelling 138

4.2 Regression Analysis 139

4.2.1 The Linear Regression Model 139

4.2.2 Zonal-Based Multiple Regression 148

4.2.3 Household-Based Regression 149

4.2.4 The Problem of Non-Linearity 151

4.2.5 Obtaining Zonal Totals 152

4.2.6 Matching Generations and Attractions 153

4.3 Cross-Classification or Category Analysis 153

4.3.1 The Classical Model 153

4.3.1.1 Introduction 153

4.3.1.2 Variable Definition and Model Specification 154

4.3.1.3 Model Application at Aggregate Level 155

4.3.2 Improvements to the Basic Model 156

4.3.2.1 Equivalence Between Category Analysis and Linear Regression 156

4.3.2.2 Regression Analysis for Household Strata 158

4.4 Other Trip Generation Formulations 159

4.4.1 Alternative Model Formulations 159

4.4.1.1 The Negative Binomial (NB) Approach 159

4.4.1.2 The Ordinal Probit Model 160

4.4.1.3 Comparing the Performance of Count Data and Linear Regression Models 160

4.5 Trip Generation and Accessibility 161

4.6 The Frequency Choice Logit Model 162

4.7 Tour Generation 164

4.8 Forecasting Variables in Trip Generation Analysis 165

4.9 Stability and Updating of Trip Generation Parameters 167

4.9.1 Temporal Stability 167

4.9.2 Geographic Stability 168

4.9.3 Bayesian Updating of Trip Generation Parameters 168

Exercises 171

5 Trip Distribution Modelling 173

5.1 Definitions and Notation 174

5.2 Growth-Factor Methods 176

5.2.1 Uniform Growth Factor 176

5.2.2 Singly Constrained Growth-Factor Methods 177

5.2.3 Doubly Constrained Growth Factors 178

5.2.4 Advantages and Limitations of Growth-Factor Methods 180

5.3 Synthetic or Gravity Models 180

5.3.1 The Gravity Distribution Model 180

5.3.2 Singly and Doubly Constrained Models 182

5.4 The Entropy-Maximising Approach 183

5.4.1 Entropy and Model Generation 183

5.4.2 Generation of the Gravity Model 185

5.4.3 Properties of the Gravity Model 187

5.4.4 Production-Attraction Format 189

5.4.5 Segmentation 190

5.5 Calibration of Gravity Models 190

5.5.1 Calibration and Validation 190

5.5.2 Calibration Techniques 191

5.6 The Tri-Proportional Approach 192

5.6.1 Bi-Proportional Fitting 192

5.6.2 A Tri-Proportional Problem 194

5.6.3 Partial Matrix Techniques 195

5.7 Other Synthetic Models 197

5.7.1 Generalisations of the Gravity Model 197

5.7.2 Intervening Opportunities Model 198

5.7.3 Disaggregate Approaches 200

5.8 Practical Considerations 200

5.8.1 Sparse Matrices 200

5.8.2 Treatment of External Zones 201

5.8.3 Special Generators 201

5.8.4 Intra-Zonal Trips 201

5.8.5 Journey Purposes 202

5.8.6 K Factors 202

5.8.7 Adjusting Trip Matrices 203

5.8.8 Errors in Modelling 203

5.8.9 The Stability of Trip Matrices 204

5.8.10 Sense Checks 206

Exercises 206

6 Modal Split and Direct Demand Models 209

6.1 Introduction 209

6.2 Factors Influencing the Choice of Mode 209

6.3 Trip-End Modal-Split Models 211

6.4 Trip Interchange Modal-Split Models 211

6.5 Synthetic Models 213

6.5.1 Distribution and Modal-Split Models 213

6.5.2 Distribution and Modal-Split Structures 215

6.5.3 Multimodal-Split Models 216

6.5.4 Calibration of Binary Logit Models 219

6.5.5 Calibration of Hierarchical Modal-Split Models 220

6.6 Direct Demand Models 222

6.6.1 Introduction 222

6.6.2 Direct Demand Models 222

6.6.3 An Improvement on Direct Demand Modelling 224

6.7 Sense Checks 225

Exercises 227

7 Discrete Choice Models 231

7.1 General Considerations 231

7.2 Theoretical Framework 234

7.3 The Multinomial Logit (MNL) Model 236

7.3.1 Specification Searches 238

7.3.2 Universal Choice Set Specification 239

7.3.3 Some Properties of the MNL 240

7.4 The Nested Logit Model (NL) 241

7.4.1 Correlation and Model Structure 241

7.4.2 Fundamentals of Nested Logit Modelling 242

7.4.2.1 The Model of Williams and of Daly-Zachary 243

7.4.2.2 The Formulation of McFadden: The GEV Family 244

7.4.3 The NL in Practice 246

7.4.3.1 Limitations of the NL 247

7.4.4 Controversies About Some Properties of the NL Model 247

7.4.4.1 Specifications Which Address the Non-Identifiability Problem 247

7.4.4.2 On the Limits of the Structural Parameters 249

7.4.4.3 Two Further Issues 251

7.5 The Multinomial Probit Model 253

7.5.1 The Binary Probit Model 253

7.5.2 Multinomial Probit and Taste Variations 254

7.5.3 Comparing Independent Probit and Logit Models 255

7.6 The Mixed Logit Model 255

7.6.1 Model Formulation 255

7.6.2 Model Specifications 256

7.6.2.1 Basic Formulations 256

7.6.2.2 More Advanced Formulations 258

7.6.3 Identification Problems 259

7.6.3.1 Theoretical Identification 260

7.6.3.2 Empirical Identification 260

7.7 Other Choice Models and Paradigms 261

7.7.1 Other Choice Models 261

7.7.2 Choice by Elimination and Satisfaction 262

7.7.2.1 Compensatory Rule 263

7.7.2.2 Non-Compensatory Rules 263

7.7.3 Habit and Hysteresis 264

7.7.4 Modelling with Panel Data 265

7.7.4.1 Panel Data Models 266

7.7.4.2 Efficiency and Repeated Observations 267

7.7.4.3 Dealing with Temporal Effects 269

7.7.5 Hybrid Choice Models Incorporating Latent Variables 271

7.7.5.1 Modelling with Latent Variables 272

7.7.5.2 Hybrid Discrete Choice Model 272

7.7.6 Attribute Non-Attendance and Other Heuristics 273

Exercises 276

8 Specification and Estimation of Discrete Choice Models 279

8.1 Introduction 279

8.2 Choice-Set Determination 280

8.2.1 Choice-Set Size 280

8.2.2 Choice-Set Formation 281

8.3 Specification and Functional Form 282

8.3.1 Functional Form and Transformations 282

8.3.1.1 Basic Box-Cox Transformation 283

8.3.1.2 Box-Tukey Transformation 283

8.3.2 Theoretical Considerations and Functional Form 283

8.3.3 Intrinsic Non-Linearities: Destination Choice 284

8.4 Statistical Estimation 285

8.4.1 Estimation of Models from Random Samples 285

8.4.1.1 The t-test for Significance of any Component theta¯*k of theta¯* 288

8.4.1.2 The Likelihood Ratio Test 291

8.4.1.3 The Overall Test of Fit 292

8.4.1.4 The rho² Index 293

8.4.1.5 The Percentage...