household activity-travel behavior: implementation of within

Household Activity-Travel Behavior: Implementation of Within-Household

Interactions

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de

Technische Universiteit Eindhoven, op gezag van de

rector magnificus, prof.dr.ir. C.J. van Duijn, voor een

commissie aangewezen door het College voor

Promoties in het openbaar te verdedigen

op dinsdag 1 december 2009 om 16.00 uur

door

Renni Anggraini

geboren te Banda Aceh, Indonesië

Dit proefschrift is goedgekeurd door de promotor:

prof.dr. H.J.P. Timmermans

Copromotor:

dr. T.A. Arentze

Copyright © 2009 R. Anggraini

Technische Universiteit Eindhoven,

Faculteit Bouwkunde, Urban Planning Group

Photo by: Aldy Fithrico

Cover design: Tekenstudio, Faculteit Bouwkunde

Printed by the Eindhoven University of Technology Press Facilities

BOUWSTENEN 141

ISBN 978-90-6814-623-4

NUR-code 955: Bouwkunde

i

PREFACE

This thesis is the result of my PhD study that I have accomplished as a member

of the Urban Planning Group, Eindhoven University of Technology. Without

the help, contribution and support of many people, family, friends and

colleagues, I would not have been able to complete this PhD research project. I

would like to thank everyone who has supported and assisted me during this

time, and especially express my gratitude to those who have assisted me by

providing valuable feedback on my work at various stages.

First of all, it is a great honor for me to have worked under the supervision of

Professor Harry Timmermans. I acknowledge and show my profound respect to

him as a highly reliable advisor. He is a very encouraging and inspirational

advisor, always providing interesting and promising research directions. I

would also like to thank my co-promoter, Theo Arentze, for his considerable

support. Throughout my study, Theo provided me very detailed technical and

conceptual support both in theoretical and practical aspects, especially in

computer programming. As my research concerned the refinement of the

ALBATROSS system, it was not an easy task for me to understand somebody

else’s work and algorithms. The bi-weekly meetings with Harry and Theo have

improved my knowledge of activity-based analyses and research in general.

Their feedbacks and comments on papers and the thesis manuscript were very

impressive and improved my English writing skills. Without their assistance, it

would have been impossible for me to finalize this PhD research. Thanks to

Harry and Theo! I really loved working with both of you.

I would like to thank the University of Syiah Kuala for financing my PhD

research through the TPSDP Project-Dikti during my first two years. Special

gratitude goes to Prof. Dr. Samsul Rizal, Dr. Alfiansyah Yulianur, Dr. Mustanir,

Dr. Ismail AB, Dr. Moch. Afiffudin, Danker Schaareman, and all staff for their

efforts to the successful of my research. I would also like to express thank the

Eindhoven University of Technology for financially supporting me for the

second half of my PhD research.

ii

I would also like to thank my colleagues in the Urban Planning Group. I

enjoyed pie time, lunch time and chatting on many occasions. In particular, I

show my appreciation to Mandy van de Sande-van Kasteren, Anja van den

Elsen-Janssen and Ingrid Dekkers-de Bruijn for their splendid secretarial

support and kindness, and other colleagues including Astrid Kemperman and

Aloys Borgers for their inspiration. I will never forget the help of Peter van der

Waerden and Leo van Veghel who picked me up at Schiphol airport on day one.

Thanks also go to my colleagues, Marloes Verhoeven, Claudia Pelizaro, Linda

Nijland, and Han Qi who were very generous giving away their home stuff. I

would also like to thank my dear friends and family, Ina Rosyid, Inne Harjanto,

Dianti-Oki, and Luluk-Nandra who were welcoming my children to their

homes, especially when my husband was away to Indonesia and I could not pick

up the children from school. Thanks also to Ella Meilinda and Rinaldi Husin

families who visited us frequently in Eindhoven and made our stay in Holland

more cheerful. It was also a sweet memorable time with Vivi, Desi, Dianti, Runi

and Leila for cooking together during Ramadhan. Especially to Desi and Ferdi:

thanks a lot for guiding me in computer programming.

Special thanks also go to my brothers and sister, Yudi Kurnia, Susi Andriani,

and M. Fadhilla Ismali who always supported me in every possible way, and to

my mother and my late father, for giving me everlasting support and pray for

my education and life. I thank God for having all of you in my life. Thanks also

to my parents-in-law for support and kindness. Last of all, I would like to thank

my dearly-beloved husband, Aldy Fithrico, and our lovely kids, Alyauma

Akmal Kalani and Alzhira Hana Fitriani. Their presence and love were really

delightful and allowing me to enjoy our time in Holland. Thanks for all support,

especially during the injury time of finalizing the thesis, when my husband and

son helped me to produce the author and subject indexes.

Finally, I thank the many people who contributed to my life and ask forgiveness

from those I have omitted unintentionally. Thank you all!

iii

TABLE OF CONTENTS

Preface

List of Figures

List of Tables

CHAPTER 1

INTRODUCTION 1

1.1 Shifting Paradigms in Travel Demand Modeling 1

1.2 Household Decision Making 3

1.3 Aims and Outline of the Thesis 3

References 6

CHAPTER 2

LITERATURE REVIEW 7

2.1 Introduction 7

2.2 Analytical Studies on Household Decision Making 8

2.2.1 Car Allocation and Usage Decisions 8

2.2.2 Task and Time Allocation Decisions 8

2.2.3 Joint Activity Participation 13

2.3 Partial Models of Household Decision Making 15

2.3.1 Car Allocation 15

2.3.2 Task Allocation 16

2.3.3 Joint Activity Participation 23

2.3.4 Travel Arrangements 24

2.4 Household Decision Making in Comprehensive Activity-Based

Models 26

2.4.1 Constraint-based Models 27

2.4.2 Simulation Models 27

2.4.3 Utility-Maximizing Models 29

2.4.4 FAMOS 30

2.4.5 CEMDAP and CEMDAP-2 31

2.5 Conclusions 32

References 32

iv

CHAPTER 3

FRAMEWORK 44

3.1 Introduction 44

3.2 ALBATROSS and Household Decision Making 45

3.3 The New Version of ALBATROSS 46

3.3.1 Activity-Travel Diary Data 48

3.3.2 The ALBATROSS Process Model and Extension to

Include Household Decision Making 51

3.3.2.1 The Mandatory Activity Module 57

3.3.2.2 The Non-Work Activity Module 59

3.4 Derivation of Decisions from Decision Tree 60

3.4.1 Discrete Choices 60

3.4.2 Continuous Choices 61

3.4.3 Goodness-of-Fit Measures 63

3.4.3.1 Discrete Choices 63

3.4.3.2 Continuous Choices 65

3.5 Conclusions and Discussion 65

References 66

CHAPTER 4

CAR ALLOCATION BETWEEN HOUSEHOLD HEADS IN

CAR-DEFICIENT HOUSEHOLDS: A DECISION MODEL 68

Abstract 68

4.1 Introduction 69

4.2 ALBATROSS Process Model 71

4.3 Data 74

4.4 Car Allocation Model Specification 74

4.5 Empirical Analysis 77

4.5.1 Descriptive Analysis 77

4.5.2 Decision Tree Induction 79

4.5.3 Deriving Impact Tables 80

4.5.4 Condition and Action Variables 82

4.5.5 Results 84

4.6 Summary and Conclusions 90

References 91

v

CHAPTER 5

MODELING JOINT ACTIVITY PARTICIPATION AND

HOUSEHOLD TASK ALLOCATION 93

Abstract 93

5.1 Introduction 94

5.2 The Activity Scheduling Process Model 95

5.3 Models Specification 97

5.3.1 Activity Selection 97

5.3.2 Activity Allocation 98

5.4 Data 98

5.5 Analyses 98




5.5.4 Results: Activity Participation Tree 104

5.5.5 Results: Task Allocation Tree 107


References 111

CHAPTER 6

CONTINUOUS CHOICE MODEL OF TIMING

AND DURATION OF JOINT ACTIVITY 112

Abstract 112

6.1 Introduction 113

6.2 Overview of ALBATROSS Model 114

6.3 Data Description 117

6.4 Variable Specification 119

6.5 Methods 123



6.6 Results 124


References 129

CHAPTER 7

HOUSEHOLD LOCATION CHOICE MODELS FOR

INDEPENDENT AND JOINT NON-WORK ACTIVITY 131

Abstract 131


vi

7.2 Location Decisions in the Existing Model 133

7.3 Household Location Decisions (Joint Activity) 136

7.4 Data 138

7.5 Overview of Condition and Action Variables 138

7.6 Decision Tree Induction and Impact Table Methods 142

7.7 Descriptive Analysis 143

7.8 Results 145

7.8.1 Independent Activity 146

7.8.2 Joint Activity 148

7.9 Conclusions 149

References 151

CHAPTER 8

CAR ALLOCATION DECISIONS IN CAR-DEFICIENT

HOUSEHOLDS: THE CASE OF NON-WORK TOURS 153

Abstract 153


8.2 Data Description 155

8.3 Methodology 155

8.3.1 Car Allocation Decisions 155


8.3.3 Impact Tables 160


8.4 Descriptive Analysis 163

8.5 Results 165

8.6 Conclusions 168

References 170

CHAPTER 9

THE INTEGRATION MODEL 171


9.2 Test of Validity Using MON Data 172

9.2.1 Frequencies 172

9.2.2 Indicators 176

9.3 Test of Sensitivity 178

9.3.1 Synthetic Populations 178

9.3.2 Scenario 178


vii

CHAPTER 10

CONCLUSIONS AND DISCUSSION 190

SUMMARY 194

Appendix 199

Author Index 247

Subject Index 250

List of Publications 255

Curriculum Vitae 257

viii

List of Tables

TABLE 3.1 Classification of Activities in a Household in ALBATROSS 49

TABLE 3.2 Socio-Economic and Situational Attributes used in

ALBATROSS 49

TABLE 3.3 Accessibility Measures used in ALBATROSS 50

TABLE 4.1 Defining Car Allocation Decisions in Households 75

TABLE 4.2 Distributions of Households across Household Composition

and SEC (%) 77

TABLE 4.3 Distributions of Household Heads across Household

Composition and Work Status of Household Heads by

Gender (%) 78

TABLE 4.4 Work Duration Statistics by Work Status and Gender 78

TABLE 4.5 Work Duration Statistics by Day of the Week and Gender 78

TABLE 4.6 Condition Variables for Car Allocation Model 83

TABLE 4.7 Frequency Distribution of Work Status across the Action

Variables 85

TABLE 4.8 Confusion Matrix for the Training and Validation Sets 89

TABLE 4.9 Impact Tables of Condition Variables of Car Allocation

Model 89

TABLE 5.1 Activity Classifications in a Household 96

TABLE 5.2 Condition Variables for Decision Tree Models 102

TABLE 5.3 Impact of Condition Variables of HH Activity Participation

Model 105

TABLE 5.4 Impact of Condition Variables of Task-Activity Allocation

Model 108

TABLE 6.1 Independent and Joint Activity Frequency (percentage) 117

TABLE 6.2 Average Duration (minutes) 118

TABLE 6.3 Definitions of Condition Variables 121

TABLE 6.4 Duration Tree Model 126

TABLE 6.5 Start-Time Tree Model 128

TABLE 7.1 Condition Variables of Independent and Joint Activity 140

TABLE 7.2 The Percentage of Performing Independent Activity at the

Same Location as Previous and/or Next Activity 143

TABLE 7.3 The Percentage of Performing Joint Activity at the Same

Location as Previous and/or Next Activity 144

ix

TABLE 7.4 The Percentage of Performing Independent and Joint Activity

by Available Distance and Location Size Band in Prisms 144

TABLE 7.5 Results of Location Decision Tree Models 145

TABLE 7.6 Impact Table for Independent Activity 147

TABLE 7.7 Impact Table for Joint Activity 149

TABLE 8.1 Itinerary of Male-Female Heads in a Particular Household 159

TABLE 8.2 Condition Variables for Car Allocation Model 161

TABLE 8.3 Primary Activity of a Tour of Male – Female 164

TABLE 8.4 Percentage of Getting a Car by Male/Female across Work

Status 164

TABLE 8.5 Average Duration of Non-work Tour(s) across Work Status

(in minute) 164

TABLE 8.6 Results of the Car Allocation Model to Non-Work Tours 166

TABLE 8.7 Impact Table of Car Allocation Decision to Non-Work Tour

Model 166

TABLE 9.1 Some Relevant Variables at the Aggregate Level 174

TABLE 9.2 Observed and Predicted of the Old and New Versions 180

TABLE 9.3 Comparison between Base-line and Scenario on

Socio-Demographic Characteristics 181

TABLE 9.4 Predicted Scenario Effects on Some Variables/Indicators:

Old Model Version 182

TABLE 9.5 Predicted Scenario Effects on Some Variables/Indicators:

New Model Version 184

x

List of Figures

FIGURE 3.1 Main Steps in the Scheduling Process of Current

ALBATROSS 53

FIGURE 3.2 Generation Modules in ALBATROSS 53

FIGURE 3.3 The Process Model for Mandatory Activities 54

FIGURE 3.4 The Process Model for Predicting Locations of Work

Activities 55

FIGURE 3.5 The Process Model for Predicting Locations of Work-Related

and Non-Work Activities 55

FIGURE 3.6 The Process Model for Non-Work Activities 56

FIGURE 4.1 Schematic Representation of Main Steps of the ALBATROSS

Process Model 72

FIGURE 4.2 The Process of Car Allocation Model for Work Tours 75

FIGURE 4.3 Examples of Distinguished Cases 76

FIGURE 4.4 Car Allocation Tree Model with 5 Major Branches 87

FIGURE 6.1 Household Activity-Travel Scheduling Process of

ALBATROSS 116

FIGURE 6.2 Start-Time Profiles every 30 minutes for each Activity 118

FIGURE 7.1 The Process Model for Predicting Location of Non-Work

Activities 135

FIGURE 8.1 The Process of Car Allocation Decisions for Non-Work

Tour 157

FIGURE 8.2 An Example of Defining Car Allocation Decision Cases in

Household Schedules 159

1

Chapter 1

INTRODUCTION

1.1 SHIFTING PARADIGMS IN TRAVEL DEMAND MODELING

Travel demand modeling has been considered a fundamental area in transportation

research for decades. It has been customarily used in urban planning and transportation

engineering to predict transport demand and evaluate the possible consequences of

spatial, infrastructure, and socio-economic policies. The traditional paradigm, still

dominant in planning practice, is the trip-based, four-step modeling approach. The

four-step model is a primary tool for forecasting future demand and performance of a

transportation system. In order to assess the impact of infrastructure investments and

other policies, models that predict long term travel demand were deemed critical in

evaluating alternative investment and other policies. The four-step model is achieving

this goal by breaking down the decisions that ultimately lead to traffic flows into trip

generation, destination choice, choice of transport mode and route choice. These four

subsequent decisions are modeled separately and independently. Originally, traffic

zones served as the unit of aggregation; later travel behavior of individuals was

simulated. Predicted flows are then used to determine future road capacity needs. For

more details see Ortuzar and Willumsen (1994) and McNally (2007).

In the 1970s, increasing concerns were raised about these four-step models, which

were criticized for their lack of theoretical appeal and lack of modeling many

2

interdependencies that may exist among the various choice facets. The model forecasts

turned out to be very unreliable and failed to assess especially the secondary effects of

policy measures correctly. The models lacked any explanation in terms of human

decision making. Furthermore, the models disregarded constraints such as intra-

household constraints, situational constraints, space-time constraints, and institutional

constraints. They also neglected the dependencies between travel mode, departure time

and destination choice.

When in the 1990s, policy shifted from long-term investment strategies to short-term

market-oriented solutions, the need to develop transport demand models that could

predict behavioral responses to policy measures was expressed in the academic

research community and was somewhat echoed in policy agendas. It led to the

development of activity-based models, which view travel as the result of people

organizing their activities in time and space. Activity-based models are founded in

behavioral theory and focus on the interdependencies between activity generation,

transport mode choice, destination, stop pattern and route choice, in the context of

multiple constraints that limit the choices of individuals and households. Moreover,

temporal dimensions were added to increase the sensitivity of the model. Activity-

based models also predict the timing and duration of activities.

While the vast majority of planning organizations continue to rely on traditional

models, academic research suggests that activity-based approaches promise greater

predictive capability, more accurate forecasts, and especially more realistic sensitivity

to policy changes (McWethy, et al., 2002). Recognition of the various

interdependencies in activity timing and other travel attributes allow greater realism in

models of travel demand. Moreover, activity-based modeling is better suited to current

transportation planning interests. In general, activity-based models focus on activities

as the unit of analysis as opposed to trips as the unit of analysis in trip-based models.

This shift has enabled the models to address issues related to substitution of non-travel

alternatives. Focusing on activity episodes also permits the incorporation of constraints

such as time constraints related to opening hours, work schedules, expected activity

duration, and multi-day scheduling of activities.

Different modeling approaches have been suggested in the literature, and each of these

has led to operational models. The dominant approach is based on the principle of

utility-maximization and corresponding discrete choice models. Observed activity-

travel patterns are viewed as the results of individuals maximizing their utility. The

potential of discrete choice models has been recognized from the mid 1970s onwards.

The shift from trip-based via tour-based to activity-based models simply meant

3

increasing complexity. Typically some nested structure is assumed and the parameters

of the model are estimated using the principle of utility-maximizing behavior.

A second approach is focusing primarily on time use. An example is AMOS/PCATS

(Kitamura and Fujii, 1998), which was later operationalized for the State of Florida

(FAMOS). CEMDAP, developed by Bhat and his co-workers (2004) is an example of

a hybrid system. It consists of a series of separate submodels, which are linked in a

micro-simulation system. Each submodel applies advanced econometrics.

Finally, rule-based models have been developed. These models assume that choices are

context-dependent. Logical rules are extracted from empirical observations for each

stage of an assumed process model. An example is ALBATROSS (Arentze and

Timmermans, 2000, 2004, 2005) which has been developed for the Dutch Ministry of

Transportation, Public Works and Water Management.

1.2 HOUSEHOLD DECISION MAKING

The aim of introducing more interdependencies in the models is not only concerned

with interdependencies in choice facets, but also with interdependencies between the

decisions of individuals. It was realized that in many cases, it is not the individual, but

rather the household that makes decisions. Households are relevant in at least three

situations. First, the activity-travel patterns of household members need to be

synchronized in time and space for joint activities, such as dinner or a family outing.

Second, resources may need to be allocated to individual household members. In turn,

resource allocation decisions may limit subsequent choices of individual members. For

example, if one member uses the car in a single-car household, other household

members cannot use the car at the same time, implying their action space may be

limited. Thirdly, some activities are household activities, implying that only one

household member has to conduct that activity. In turn, such task allocation decisions

influence other aspects of activity-travel programs.

An examination of the literature shows that although the topic of household decision

making has been high on the research agenda for many years, most activity-based

models of transport demand are still based on individual travel patterns.

1.3 AIMS AND OUTLINE OF THE THESIS

An examination of the literature shows that although the topic of household decision

making has been high on the research agenda for many years, most activity-based

4

models of transport demand are still based on individual travel patterns. The goal of

this thesis therefore is to systematically model household decision making processes in

an activity-based framework with a special focus on resource and task allocation and

joint activity participation. More specifically, this study represents an attempt to

improve ALBATROSS (Arentze and Timmermans, 2004). This model was one of the

few of its generation that did include aspects of household decision making by

simulating the decisions of one household member, and then based on the outcome of

this, model the decision process of another household member.

The aim of this thesis is primarily to refine the ALBATROSS model to represent

household-level decision making more explicitly so that the interaction between

persons can be captured well. In particular, the following elements will be further

elaborated:

1. Joint activity participation choice was not modeled in the sense that the required

synchronization in case of joint activities was not imposed as a constraint. We

will attempt to model joint activity participation in a more consistent manner.

2. Activity allocation to each household head was an implicit decision step. In this

study, we will model this step explicitly.

3. Car allocation to each male and female head was also an implicit decision, in

particular for those households with more drivers than cars. The car allocation

problem will now also be modeled explicitly in this study.

The aim of refining ALBATROSS is to make it more comprehensive and applicable

for household-level decision making. Household heads need to trade-off activity needs

and mobility in the context of joint activity participation, household task allocation and

mobility. Joint activity participation needs compromise when the activity is done by

male and female jointly.

In order to achieve these goals, some component of the process model underlying

ALBATROSS is elaborated or re-designed in terms of household decision making.

Moreover, a new much larger data set (the MON-data) is used, implying that the

decision rules that are derived are based on more (household) data. The structure of this

thesis follows the process model underlying ALBATROSS.

Before going through the chapters that make up this thesis, it should first be noted here

that most chapters are based on previously published conference papers or journal

articles. Intrinsically, this format leads to some overlap between some of the chapters,

especially regarding parts of introduction, methods used and descriptions of data

5

collection efforts, although every attempt has been made to write each paper in such a

way that transitions from one chapter to another are as smooth as possible.

Chapter 2 starts by discussing the literature review in the area of household decision

making in urban travel demand modeling. Chapter 3 further continues to discuss the

general framework underlying the model system. It motivates the potential advantages

of a rule-based system. In terms of modeling, the challenge is to extract decision rules

from observed activity-travel patterns. Throughout the thesis, a CHAID-based decision

tree induction method is used as in the basic model. Chapter 4 describes the results of

the first household decision. It is concerned with the problem of car allocation decision

to work tours focusing on car-deficient households, i.e. households where the number

of drivers is higher than the number of cars, the decision who or no-one at all uses the

car to go to work is modeled.

Chapter 5 discusses the results of the model for generating non-work activities. Two

different models are developed, one for joint activity participation and one for

household task allocation, focusing on two-heads households. A classification of

activities is developed and activity types that likely relate to the needs at the household

level are identified. .

Chapter 6 specifies two subsequent models: duration and start-time models for non-

work activities conducted jointly by the male and female head of a household.

Specifically, the study investigates the timing of non-work activities related to

household and family activities, such as household tasks (e.g., escorting persons,

grocery shopping) and non-task activities (i.e., social and leisure activities). The

analysis focuses on two-heads households (with or without children) and the joint

activities in their schedules.

Chapter 7 re-estimates location choice models in the context of household decision

making. It consists of two primary models for respectively independent and joint non-

work activities. As in the basic model, the concept of detour time is used. This concept

considers relative locations to the previous and next activity as the unit of analysis for

defining location choice. By applying that concept, distances between locations that

may be combined in a single trip-chain can be captured well.

Chapter 8 models the car allocation decision to non-work tours. This chapter is similar

to chapter 4 that is concerned with the car allocation decision to work tours.

Nevertheless, the way of defining the car allocation decision is quite different given the

sequential process.

6

Chapter 9 discusses the integration of these sub-models into the integral ALBATROSS

model. It intends to prove that the simulation of sequential choice facets and

observations are little different.

Finally, Chapter 10 summarizes this study, reflects on the results and identifies some

avenues of future research.

REFERENCES

Arentze, T.A. and Timmermans, H.J.P. (2000), ALBATROSS: A Learning-based Transportation Oriented Simulation System. EIRASS, Eindhoven University of

Technology, The Netherlands.

Arentze, T.A. and Timmermans, H.J.P. (2004), A Learning-Based Transportation

Oriented Simulation System, Transportation Research Part B, 38, 613-633.

Arentze, T.A. and Timmermans, H.J.P. (2005), ALBATROSS 2.0: A Learning-based Transportation Oriented Simulation System. EIRASS, Eindhoven University of


Bhat, C.R., Guo, J.Y., Srinivasan, S., and Sivakumar, A. (2004), A Comprehensive

Econometric Microsimulator for Daily Activity-Travel Patterns, Transportation Research Record, 1894, 57-66.

Kitamura, R. and Fujii, S. (1998), Two Computational Process Models of Activity-

Travel Behavior. In: T. Gärling, T. Laitila and K. Westin (eds.), Theoretical Foundations of Travel Choice Modeling, Elsevier, New York, pp. 251-279.

McNally, M.G. (2007), The Four Step Model. In: Hensher, D.A. and K. Button (eds.)

Transport Modeling, 2nd Edition, Pergamon, Oxford, pp. 55-73.

McWethy, B.L., Lemp, D.J., and Kockelman, M.K. (2002), From Aggregate Methods

to Microsimulation: Assessing the Benefits of Microscopic Activity-Based Models

of Travel Demand. In: Proceedings of the 86th Annual Meeting of the Transportation Research Board, Transportation Research Board, National Research

Council, Washington D.C.

Ortuzar, J.deD. and Willumsen, L.G. (1994), Modelling Transport (second edition),

Wiley, Chichester.

7

Chapter 2

LITERATURE REVIEW

2.1 INTRODUCTION

As briefly discussed in the introduction, the activity-based approach to travel demand

forecasting represents an attempt of improving the integrity of demand forecasting

models by explicitly modeling various dependencies. These dependencies are not only

concerned with the various choice facets (generation, destination, transport mode, etc),

but also with dependencies between members of the household. The focus on the

household as opposed to the traditional focus on the individual is especially important

in the context of task and resource allocation and joint activities. Although the

importance of the household level has been recognized in seminal work, except for

some analytical studies, only recently there have been attempts of modeling these

phenomena (Timmermans, 2006).

The purpose of this chapter is to give an overview of this line of research. First, we will

summarize empirical work, followed by recent modeling attempts. Finally, we will

discuss how household decision making is treated in existing comprehensive activity-

based models of transport demand.

8

2.2 ANALYTICAL STUDIES ON HOUSEHOLD DECISION MAKING

2.2.1 Car Allocation and Usage Decisions

Activity participation and destination choice often depend on the transport modes that

are available to individual household members. Car use often means that more

destinations can be visited during a single trip or that destinations further away from

home can be reached within a given time budget. Especially in car-deficient

households in which the number of cars is less than the number of drivers, car

allocation and usage is a household decision which impacts many other choice facets of

individual activity-travel patterns. Golob, Kim, & Ren (1996) analyzed how drivers are

allocated to vehicles in multi-driver/multi-vehicle households. They found that gender,

income, work status, age and the presence of small children influenced the number of

vehicle miles traveled with the various vehicles. Hunt & Petersen (2005) also found

evidence of gender differences.

Almost similar, Vance and Iovanna (2007) also found that gender play a role in

determining the probability of car use and the distance driven. Drawing on a panel data

collected between 1996-2003 in car-owning households in Germany, the results

indicated that although women, on average, perform more non-work travel than men,

they were more reliance on other modes than car. Another interesting study is done by

Vovsha and Petersen (2007). A model system structure is proposed that can fully

address all needs associated with car allocation and use. The core short-term module

includes two long-term sub-models: 1) household car ownership, 2) main driver

assignment for each car, and four short-term sub-models: 3) individual and joint travel

generation, 4) schedule adjustments, 5) mode choice, and 6) car allocation and type

choice. However, neither of the existing model system has yet included a full set in a

consistent way.

2.2.2 Task and Time Allocation Decisions

One would expect that household characteristics, such as the structure and the number

of persons in a household, influence the number and type of activities conducted in the

household and therefore task allocation and travel decisions. Household structure also

influences where (in-home vs. out-of-home) activities are conducted (Gronau, 1977;

Lawson, 1999). A major factor that influences the decision to travel relates to the role

of paid work within a household. The amount of time spent on paid work strongly

influences the budget available for household consumption, and the total amount of

time and the time of day available for other activities. Lee & Hickman (2004), looking

into this, examined time allocation of households within trip chains using simultaneous

9

doubly-censored tobit models. In particular, they compared trip chaining behavior,

among five types of households: single non-worker households, single worker

households, couple non-worker households, couple one-worker households, and couple

two-worker households. They found that household types, defined by the number of

household heads and work status, strongly influence activity time allocation in trip

chains. The presence of children in the household has a positive effect on the duration

of all out-of-home activities in household trip chaining, except for the duration of out-

of-home discretionary activities of households having children under 5 years old. This

suggests that the presence of children induces more chaining of trips and more time

allocated to these trip chains. Households having more children of 16 years of age and

over are more likely to spend time in trip chaining for out-of-home subsistence

activities. Finally, they found that flexible work arrangements tend to be correlated

with less trip chaining for the work trip.

Another consistent finding in the literature is that the work commute of women is

shorter (e.g., Hanson & Hanson, 1980; Hanson & Johnston, 1985; Singell & Lillydahl,

1986; White, 1986; Fagnani, 1987; Gordon, et al., 1989; Hanson & Pratt, 1990; Turner

& Niemeier, 1997). It reflects the fact that on average working women are less flexible

because they need to combine paid work and household activities. Women are able to

combine work and domestic duties primarily by working closer to home, more trip-

chaining and relying on social networks (Hanson & Pratt, 1995: Kwan, 1999; Dowling,

2000). Consequently, accessibility considerations are more important to them, both in

terms of accepting a job, but also because they need to take care of many other non-

work activities. Stopher & Metcalf (1999, 2000) concluded for several cities in the

United States that beyond the effects of lifecycle, both gender and working status

influence the amount of time allocated to household activities (see also Vadarevu &

Stopher, 1996, 1999). Likewise, Schwanen, Ettema & Timmermans (2006) argued that

if a spouse works longer hours, s/he has less time for domestic tasks. Relegating

household activities to one’s partner may then be a reasonable strategy to cope with

this situation. Alternatively, households may consider an overall reduction of

household tasks at the household level (Morris, 1990; Presser, 1994). However, such

effects are gender-specific in the sense that male’s household tasks do not change much

if women work longer hours and women, irrespective of their employment status,

continue to carry prime responsibility for these tasks (Morris, 1990; Pinch & Storey,

1992; Hanson & Pratt, 1995; Presser, 2003). There are nonetheless variations. In

addition to the impact of class, occupation and lifecycle, gender roles and power

differentials among spouses matter (Morris, 1990). Men tend to conduct more

household tasks if spouses’ roles orientations are more egalitarian (Huber & Spitze,

1983; Presser, 1994), and women’s have more resources relative to men (Antill &

Cotton, 1988; Presser, 1994). Yet another strategy may be task specialization. Men

10

may conduct more tasks in larger households with young children to increase the

efficiency of the household or to comply with the prevailing moral climate and gender

ideology (Knijn, 2004).

There is also some evidence that good accessibility stimulates out-of-home activity

participation and trip making (Boarnet & Crane, 2001; Ettema, Schwanen &

Timmermans, 2006). In contrast, poor accessibility, either as the result of the non-

availability of a car or as the result of the spatial distribution of facilities relative to

home may lead households to assign out-of-home household tasks to one spouse –

usually the female – who can combine several tasks in multi-stop activity chains. For

example, Strathman, et al. (1994) concluded that the likelihood of forming complex

commuting chains is higher for women and high-income households, both of which

tend to be “time challenged” groups. If, however, accessibility is better, men may take

on more household tasks, because accommodating such activities in their activity

schedules is easier (Hanson and Hanson, 1980; Ettema, et al., 2006). Kwan (1999,

2000) found that women’s household activities tend to be more fixed in space and time

than those by men, suggesting that such tasks are a structural component of their daily

schedules, while men conduct such activities on a “standby/basis”. This interpretation

is corroborated by Aitken (2000), who concluded from interviews that fathers

responsible for childcare felt they were merely ‘helping out’ their spouses.

Household tasks also have an effect on other in-home and out-of-home activities. For

example, Gronau (1977) looked at the effects of an increase in the number of children

and the age cohorts of the children. He found that as the number of children in a

household increases, the additional time devoted to children is not spent on work at

home and leisure. Similarly, Redman (1980) found that family size had a negative

effect on meals being eaten outside the home. Golob & McNally (1997) used a

structural equation model to investigate activity participation and travel of couples.

Activities were classified into three categories: work, maintenance, and discretionary.

The total out-of-home duration for these categories was calculated as was total travel

time. A series of household and personal characteristics was used as the exogenous

variables of the model. They studied four types of direct effects: travel requirements of

out-of-home activities, within-person activity interactions, within-person travel

interactions, and cross-person interactions. One of the interesting results was that if the

male increases his participation in work activities, the female’s travel for maintenance

activities increases more than proportionally to the increase in the female’s

participation in maintenance activities.

Borgers, Hofman, Timmermans & Ponjé (2001) used a stated choice approach for

estimating the probability of certain task allocation profiles. The main reason for

11

collecting stated choice data was that revealed daily task allocation patterns may be

influenced too much by unique factors. Examining task allocation behavior under

laboratory conditions as a joint decision making process likely generates more valid

data. The problem addressed in this paper was that stated choice experiments typically

involve a choice between single alternatives and not between portfolios (a specific

combination of tasks). The authors therefore explored alternative approaches of how to

measure the influence of experimentally varied factors on task allocation. In a sequel,

Borgers, Hofman & Timmermans (2002) estimated a slightly simpler model. They

assumed that the presence of children of various ages in the household, the socio-

economic status of the household, age, car availability and work status of the spouses

influence time allocation decisions. Multinomial logit models, including these

variables as contextual effects, were used to predict time allocation of two spouses to a

set of activities. First, a multinomial logit model was estimated to predict the amount of

time spent together. Next, a conditional choice model was estimated to predict the

proportion of time spent by each spouse on conducting a particular activity. Because

the total amount of time is known, these proportions can be translated into the number

of hours spent on particular activities. Respondents were requested to jointly express

the amount of time they typically spend alone and together on 27 different activities,

which were later grouped into activity classes. The following activities were

distinguished: (1) sleep, eat, drink and personal care; (2) work out of home, including

travel time; (3) shopping, services, including travel time; (4) in-home non-leisure; (5)

in-home leisure; (6) out-of-home leisure; (7) bring/get activities, and (8) other. Results

indicated that if an older child is present in the household, the amount of time spent

together significantly increases. The amount of time spent together is less if either

spouse works. Time spent on sleeping, eating, drinking and personal care is

significantly less when older children belong to the household. The amount of time

spent on sleeping, eating, drinking and personal care by men is less if their spouses

have a part-time or full-time job. If men have a part-time job, their time allocation to

sleeping, eating, drinking and personal care are higher. In contrast, it is significantly

less if they have a full time job. Men working part-time spend more hours on shopping,

while men working full-time spend less time on shopping. If spouses work, men tend

to shop more, but this effect was less significant.

The effects of the work status variables were interesting. If men work part-time, they

tend to spend more time on in-home work activities, although the effect was not

significant. If they work full-time, they allocate significantly less time to in-home work.

If their spouses work, men also tend to spend more time on in-home work activities,

but this effect is only significant if their spouses work part-time. If their spouses have a

full-time or part-time job, women allocate less time to shopping, which is especially

true if their spouse works part-time. This result might reflect a shift in the overall

12

activity pattern in the sense that they may spend more time together on other activities.

The impact of work status of women is such that working women spend less time on

shopping, but this effect is only significant if they work full-time, and then only at the

90% probability level. Time pressure might be the reason for this finding. The pattern

of the signs of the work status variables is interesting. Compared to the reference

household, time allocation of men to shopping tends to increase if their spouse work,

suggesting that men take over some of the shopping responsibilities of their spouses.

Women tend to spend more time on in-home non-leisure activities if they have young

children and less time if they have older children compared to the situation where there

are no children in the household. This seems to indicate that older children help out.

Similar results were obtained for other categories. Overall, the results suggest that task

and time allocation in households depends on household type (age, children, number of

workers), the utility that is derived from joint versus solo activities, the urgency of

conducting particular activities, gender roles and the constraints and possibilities

offered by the environment to conduct these activities efficiently in time and space.

Ettema & Van der Lippe (2006) investigated task allocation patterns on a weekly basis.

The results of their analyses indicated that specialization is a dominant weekly pattern

in dealing with time constraints, i.e. each spouse takes primary responsibility for

different tasks. The presence of young children and a lower accessibility to jobs and

services increases the female's share of household tasks and childcare. This

specialization is strongest on Friday and on Wednesday, reflecting school hours and

part time work arrangements in the Netherlands. Non-traditional roles and a highly

qualified job increase the females' share of paid labor and decrease their share of

household and childcare tasks, however this effect is not observed on Fridays,

suggesting that women still, more than men, work in part time jobs where Friday is the

free day.

Cao and Chai (2007) examined activity time allocation of the male-female household

head between weekday and weekend. Based on observations on Shenzhen residents in

China, they found the gender role in the household. Men are dominant in out-of-home

activities, but women are more dominant in in-home activities. On average, women

carry more maintenance responsibilities than men, but men spend more time on work

and leisure activities than women, especially on the weekend. On the weekend,

Shenzhen’s residents are not as mobile as westerner countries because most people

spend time at home and surrounding neighborhoods, especially as for female. Further,

the influences of household structure on time allocation of both household heads

demonstrated substantial gender-role differences. The results also showed some

interesting interpersonal interactions of time allocation. Specifically, the more women

13

participate in leisure activities, the more men spend time on leisure activities, but not

vice versa. Although not substantial, women’s work activity duration tends to increase

men’s leisure activities. During the weekday, as women spend more time on

maintenance activities, overall, men as well as women participate in fewer leisure

activities. On the weekend, once one household head works, the other tends to carry

more maintenance activities.

2.2.3 Joint Activity Participation

Joint participation in activities represents a substantial portion of non-work activities, is

an important component of travel during certain time periods and affects individual

travel schedules. Joint participation in maintenance and leisure activities and the

provision of rides to family members, constrain individual choice sets and affect the

saliency of attributes that contribute to the generalized cost of travel alternatives.

Therefore, this choice problem has received relatively most interest.

The relative importance of joint activity participation is evident in that joint activities

tend to have a longer duration than non-work independent activities, and persons tend

to stay out later and travel farther from home (Kostyniuk & Kitamura, 1983).

Moreover, Fujii et al. (1999) found that time spent on activities jointly with other

household members, particularly with children, was incremental to individual feelings

of satisfaction and in decisions to allocate time to joint and independent activities.

Several studies have examined the effect of household attributes on joint activity-travel

behavior. Kostyniuk & Kitamura (1983) and Chandraskharan & Goulias (1999) found

that joint activities involving household heads are significantly affected by the presence

of children. Couples without children living at home are more likely to pursue joint

out-of-home non-work activities than couples with children. In households with

children, most joint activities between adults are at home. In addition, the employment

status of the household heads influences whether a joint activity originated from home

or from an out-of-home contact point.

Another interesting study that investigated the effects of children on household travel

behavior was done by Senbil, et al. (2008). They examined the impact of children on

various household non-commute trips for four different types of non-commute trips,

i.e., trips to shopping, restaurant, park and recreation centers, and department store.

These variables were regressed against socio-economic and demographic,

neighborhood and various child variables by using two regression analyses: linear

regression on household non-commute trips, and seemingly unrelated regression on

14

non-commute trips by male and female household heads. Results for linear regression

suggest that children can be grouped under general groups, i.e, pre-school, school, for

accounting child effect on household travel behavior. Besides, number of children

reveals significant results. Also in linear regression analyses, they found that the

differentiation outperforms a classical lifecycle classification. Results for seemingly

unrelated regression suggest that there is a general complementary among household

heads in non-commute trips, except shopping trips which display substitution, albeit

minor. Also, for household heads, pre-school children constitute the child group with

significant effects on non-commute trips.

Srinivasan and Bhat (2008) examined the joint participation with household members

and non-household members along with the generation, location, and scheduling of

joint activity episodes. They found that independent activities are different from joint

activities in systematic ways. Specifically, joint episodes are of longer durations,

significantly likely to take place at the residence of other people, and often confined to

certain time periods of the weekday. In addition, within the class of joint episodes,

important differences are also observed based on activity type, companion type, and

the day of the week. Overall, the empirical results from this study highlight the

important need to accommodate intra-household and inter-household interactions in

activity-travel behavior analysis. Specifically, some of the key implications of their

empirical findings include the following. First, given the sheer magnitudes of joint

activity and travel engagement, their results underscore the need for travel demand

models to recognize these inter-dependencies for accurate travel forecasts and policy

analysis. In particular, inter-personal linkages in activity travel behavior imply that

policy actions can also alter the travel patterns of individuals who are not directly

“exposed” to the action. For example, when a husband’s work timings change because

of work-staggering, the wife’s travel patterns can also change. Second, the timing (i.e., duration and time-of-day) of activity-travel is found to be related to the companion

type. Consequently, accurate assessment of soak time distributions for air quality

models requires information on joint activity-travel engagement patterns. Third, a high

fraction of joint leisure type activities is found to be undertaken at “someone else’s

home”. The implication here is that individuals are perhaps not as flexible in their

choice of destination location for the pursuit of discretionary-type activities as they

have been traditionally assumed in travel-demand modeling. Fourth, the desire to

participate in activities with non household members such as friends also generates

additional travel to pick-up and drop-off the activity companions. Such travel cannot be

realistically captured by individual-level models. Fifth, with the gaining prominence of

the need to model weekend travel behavior, accommodating inter-personal interactions

assumes even greater significance as joint activity and travel participation levels during

weekends are found to be greater than those during weekdays. Finally, to enable the

15

development of empirical models that accommodate inter-personal interdependencies,

future travel surveys should be suitably enhanced to adopt a more disaggregate activity

classification scheme and to collect data on individuals’ activity and travel companions.

2.3 PARTIAL MODELS OF HOUSEHOLD DECISION MAKING

2.3.1 Car Allocation

Petersen & Vovsha (2005, 2006), in addition to car allocation, also modeled car-type

choice. First, they simulated which individual and joint activities are conducted and

where these activities are conducted. Then, accessibility to the most important

activities (work and school) in combination with the household characteristics

determines car ownership by vehicle type. Next, generated activities are scheduled and

out-of-home activities are distributed by travel tours. Travel needs of the household

members are further consolidated through joint travel arrangements. Finally, available

household cars are allocated to these tours. The authors argue that numerous feedbacks

can be implemented within this framework in order to enhance the integrity of the

model system and eliminate possible inconsistencies. Interestingly, they notice that

only some of them can be formalized as log-sums in a nested logit model. Other

feedbacks are more complicated in nature and require rule-based algorithms. For

example, re-scheduling and tour-formation procedures are needed to synchronize tours

and enforce joint travel arrangements. If the total time budget proved to be unrealistic

in terms of the travel time share, adjustment of certain activities and locations is

needed.

The actual model is a multinomial model which predicts the choice of household car. A

maximum of 8 choice alternatives is distinguished, varying in terms of five car types

(small auto with 4 or less cylinders; large auto with 6 or more cylinders; van;

SUV/jeep; truck, and car age in years). If a household has less than 8 cars, unavailable

choice alternatives are blocked out. For each tour, assumed known are tour-related

attributes (purpose, destination, distance, schedule, number of stops, pure auto tour

versus drive-to-transit tour), driver-related attributes (person type, gender, age), joint-

travel-related attributes (party type, party size, fully joint versus partially joint tours),

household-related attributes (income group), and zonal attributes (area type at the

origin, area type at the destination). Purpose, distance, number of stops, driver type,

party type, joint activity participation and socio-demographics were used as

explanatory variables.

As part of the TASHA model system, Miller, Roorda & Carrasco (2003) developed a

tour-based model of travel mode choice, based on the principle of utility-maximization.

16

In particular, cars are allocated to household members such as to maximize household

utility, which is assumed to be the sum of household members’ individual utilities. The

scope of explanatory variables is largely restricted to travel time and costs of

alternative transport modes and trip purpose.

2.3.2 Task Allocation

Wen & Koppelman (1999, 2000) proposed a prototype activity stop generation and

tour scheduling model that includes the daily allocation of household maintenance

tasks and automobile use. Their model focuses on travel that is generated from

participation in activities undertaken to satisfy needs and desires of the household and

its members. The model itself is a nested logit model that differentiates between

household subsistence (work and work-related business) needs and mobility decisions,

the generation of maintenance (grocery shopping, personal and household business)

activities (stops) which serve the household in general and each member of the

household and the allocation of stops and autos among household members exclusively

or jointly. Finally, individual daily travel/activity patterns are derived through the

generation of tours, the assignment of stops to tours, and the selection of locations for

each stop and travel mode(s) for tours. The highest level is the choice of the number of

household maintenance stops. The second level is the allocation of maintenance stops

to individuals. The lowest level of the model concerns the allocation of cars to

individual household members. The second stage choices, for each adult household

member, include the number of tours and the assignment of stops to tours, conditional

on the choices of the number of maintenance stops and the allocation of stops and

autos. A distinction is made between workers and non-workers. Thus, this model

remains restrictive, both in terms of characterization of activity patterns level

characteristics and the limited choice facets that are included in the model. Time

allocation and the utility derived from different types of activities are not included in

the model, although it is an important consideration in leisure and maintenance

activities and an essential criterion for decisions regarding joint activity participation.

Zhang, Timmermans & Borgers (2002) developed a more general model of task

allocation and time use of household members. They assumed that households allocate

their time to activities such that household utility is maximized. In contrast to many

other models, household utility is not assumed to be a simple sum of household

members’ utilities, but also incorporates relative influence and interest. Starting point is

the assumption that every household has to perform a set of activities to survive or to

give some meaning or pleasure to their daily life. The utility of these activities is

assumed to differ between individuals. Role patterns within households and more

17

general lifestyle decisions influence the kind of activities that are conducted, the

household member primarily responsible for the task, and activity participation and

allocation of time across activities (and related travel). Activities are classified into

four types, i.e., in-home activities, out-of-home independent, allocated and shared (or

joint) activities. An independent activity is an activity, not being a household task that

is conducted by an individual household member (e.g., work or attending a football

match). Shared activities are those activities that require the presence of more or all

household members (e.g., dinner or a family outing). An allocated activity is a

household task that is assigned to a specific household member (e.g., daily shopping).

Shared activities may be synchronized or non-synchronized. In the former case,

household members carry out the activity together. In the latter case, household

members share the activity partially. The basic structure of their model was formulated

as:

Maximize )u,...,u,u(GGUF n21= [2.1]

Subject to ij ij Tt =∑ , for i = 1, 2, …, n [2.2]

where,

GUF stands for group (household) utility function,

ijt is the time of individual i performing activity j,

iu is household member i’s utility, and

iT is member i’s available time.

A set of alternative specifications of the group utility function was considered. The

multi-linear group utility function can be specified as follows:

n21n~1

n

1i ii iiii

n

1i ii ...uuuw...)uu(wuwGUF1 12 2121

+++= ∑ ∑∑ = >= [2.3]

where,

iw is member i’s weight parameter, and

n~1ii w,...,w21

are the intra-household interaction parameters.

This model assumes that household utility can be derived by weighting the utilities of

the individual household members, and adding interaction effects. The weight wi can

be interpreted as a measure of a member's power or influence in the group decision-

making. The interaction parameters { n~1ii w,...,w21

} moderate the power effect and

reflect the group members’ concern for achieving equality of utilities. The larger the

18

interaction parameter, the higher the group’s collective desire to choose a time

allocation such that the utilities of all household members are more or less equal.

The GUF in equation (3) finds its theoretical roots in group decision theory (Eliashberg

& Winkler, 1981, 1986; Harsanyi, 1955; Keeney, 1972; Messer & Emery, 1980). It can

include several GUFs as special cases. The additive-type group utility function only

uses the first component of the utility function and can be expressed as:

∑ ==

n

1i iiuwGUF [2.4]

Harsanyi (1955) showed that if the group is to behave in a Bayesian rational manner,

then the group utility function must be additive. This model can be arrived at when

household members first average their separate utility functions and then maximize the

resulting mixture function (Curry, et al., 1991). However, this GUF ignores the

interaction among household members. An alternative is a compromise-type group utility function which can be expressed as:

∑∑ ====

n

1i i

n

1i ii /n)(uuwGUF [2.5]

Equation [2.5] shows that household members have equal weights. Curry & Menasco

(1979) called this the compromise weight. There is some empirical support in other

disciplines for such equal weighting (e.g., Davis & Rigeaux, 1974; Munsinger, Weber

& Hansen, 1975; Krishnamurthi, 1988), but there is also empirical support of non-

equal weights (Molin, Oppewal & Timmermans, 1997, 2000). Hence, it may advisable

not to assume equal weights a priori.

Another special case is the capitulation-type group utility function, which takes group

interaction into account by assuming that each household member uses other members’

weights (utilities) as his or her own weight (utility) for joint decision-making.

∑ ==

n

1i iiuwGUF or ∑ ==

n

1i iiuwGUF [2.6]

where iw represents the average weight of other members relative to member i and is

called capitulation weight, and iu represents the average utility of other members

relative to member i and is called capitulation utility.

An alternative to these linear functions are Nash-type functions of a multiplicative

form. The group utility function can be expressed as:

19

( ) iw

i iuGUF ∏=

[2.7]

Equation [2.7] shows that this utility function is multiplicative without a reference

point. It satisfies Nash’s (1950, 1953) axioms for two-party cooperative games or

variations on those axioms. The Nash model assumes that each household member

identifies his/her most preferred outcome and the household then compromises by

averaging along the resulting negotiation frontier (Curry et al., 1991). Gupta & Livne

(1988), however, pointed out that Nash’s definition was particularly inappropriate for

multiple-issue bargaining and suggested the following definition.

( ) iw

i ii uuGUF ∏ −=

[2.8]

This type of GUF uses other members’ capitulation utility as a reference point. Curry et al. (1991) have experimentally tested the validity of this utility function. The reference

point suggests that during negotiations each member can be expected, explicitly or

implicitly, to compare each possible agreement against the reference point.

Zhang, et al. (2002) decided to use the multi-linear specification because it is easier to

estimate and it is more general. Their model only incorporated binary interaction terms

rather than multiple interaction terms. Thus, the estimated model can be formulated as

follows:

∑ ∑∑ = >=+=

n

1i ii iiii

n

1i ii1 12 2121

)uu(wuwGUF [2.9]

Each member’s utility function is further composed of the utilities from different

activities based on the same type of multi-linear GUF.

∑ ∑∑ >+=

i ii iiiii ii a1 a1a2 a2a1a2a1a aai )uu(ruru [2.10]

where,

iau is household member i’s utility for activity ia ,

iar is member i’s weight (or relative interest) for activity ia , which reflects the relative

importance of each activity making for each member’s utility, and

ii a2a1r is interaction parameter for activities i1a and i2a .

20

In particular, they proposed the following utility maximization framework to model

household time allocation based on the multi-linear household utility function, subject

to each member’s available time constraint.

Maximize ∑ ∑∑ +=+=

i 1ii' i'iii'i ii )uu(w)u(wGUF [2.11]

shri

alci

asi

shri

indi

isi

alci

indi

iai

shri

homi

hsi

alci

homi

hai

indi

homi

hii

shri

shri

alci

alci

indi

indi

homi

homii

uuruuruur

uuruuruur

ururururu

+++

+++

+++=

[2.12]

Subject to

ishri

alci

indi

homi Ttttt =+++ [2.13]

ii',ttt shrshri'

shri ≠∀== [2.14]

where, homit is the time staying at home, indit , alc

it and shrit are the time of member i performing out-of-home independent activity

(ind), allocated activity (alc) and shared activity (shr), respectively,

iT is member i’s available time for performing all these activities, shri

alci

indi

homi uu,u,u and are the utility functions,

shri

alci

indi

homi rr,r,r and are their weight parameters, and hi

ir , hair , hs

ir , iair , is

ir and asir are the interaction parameters.

In order to derive operational models of household time use, the following type of

utility function for each activity was used.

( ) ( ) }alc,ind{hom,j,xexp1tlnu jiq

jiq

jiq

ji

ji =++= ∑ εβ

[2.15]

( ) ( )shriq

shriq

shriq

shrshri xexp1tlnu εβ ++= ∑ [2.16]

where, shri

ji and εε are error terms of utility functions, j

iqx , shriqx are explanatory variables

and, jiqβ and shr

iqβ are the parameters.

21

A La Grange function was used to calculate the maximum of the household utility,

subject to the constraints of equation [2.13] and [2.14]. By calculating the conditions

for the first partial derivatives with respect to the time of each activity ( homit , ind

it , alcit

and shrit ), the models for household time use for respectively in-home activity, out-of-

home independent, allocated and shared activities were derived. A seemingly unrelated

regression (SUR) estimation procedure was applied to estimate the models.

The model was initially estimated for 188 households, who reported their activity-

travel patterns in the South-Rotterdam region, the Netherlands. The following

explanatory variables were used: socio-economic class, age of the oldest household

member, household type (no worker, single worker and double worker), the number of

owned cars and bikes, and official work hours per week) for each household member.

The goodness-of-fit of the model was satisfactory, but should be improved. Another

result of the estimation was that the influence of the male on time allocation was on

average higher than the influence of the females in the sample.

In a sequel, Zhang, et al. (2005a) also included travel time in the model. This improved

model performance significantly. Furthermore, the husband has the highest influence in

the allocation of time in nearly half the households; in one-fifth the wife had more

influence, while the remaining households showed evidence of equal relative influence.

They also compared weekday versus weekend time allocation (Zhang, et al. (2005b),

and concluded that, the wife behaves more rationally than the husband on weekdays,

while the husband does so on weekends; that the influence of intra-household

interaction and interdependency among activities seems invariant across days of the

week; that number of owned passenger cars influences the couples’ task and time

allocation behavior on weekends, not on weekdays, and that shared activities become

much more important on weekends than on weekdays.

Later, Zhang, et al. (2004) extended their basic model to also include dependencies

among activities. The results suggested that decisions regarding household task and

time allocation start with in-home activities of household members and personal and

joint out-of-home activities, after which the allocation of time to allocated activities is

negotiated. Women seem to regard the allocated activities more important and the in-

home activity less important than men do.

Zhang & Fujiwara (2004) also estimated an iso-elastic household utility function of the

following form:

1wand0w,uw1

1HUF

i iii ii1 =≥

−= ∑∑ − α

α [2.17]

22

where α is parameter indicating intra-household interaction.

The iso-elastic function is known from research on social welfare (Atkinson, 1970).

The intra-household interaction parameter α describes how and to what extent the

household positions its members (or considers the existence of its members) in the

decision-making process. Therefore, different values of α and iw , and the sign of α

represent different household decision-making mechanisms. Equation [2.17] includes

various types of household utility functions as special cases. For example, a minimum type household utility function is obtained if α is larger than one. In this case,

increasing the utility of the weaker member in a household leads to an increase of total

household utility. In case α becomes positive infinity, the household regards the

utility of its weakest member as the household utility and maximizes it, as shown

below.

( )n,...,2,1i|uminGUF i == [2.18]

If α approximates one, equation [2.17] becomes a “Nash-type” household utility in

the sense that each member first identifies his/her most preferred outcome of household

decision-making and the household then compromises by averaging along the resulting

negotiation frontier (Curry, et al., 1991).

( ) iw

i iuGUF ∏= [2.19]

If α is equal to 0, equation [2.17] results in a “utilitarianism-type” of household

utility, which assumes that the household first averages its members’ separate utilities

and then maximizes the resulting mixture utility function. Finally, if α is negative, the

household utility increases with the utilities of the strong household members. In case

α reaches negative infinity, equation represents “maximum” type of household utility

function. That is to say, the household regards the utility of its strongest member as the

household utility.

( )n,...,2,1i|umaxGUF i == [2.20]

Zhang, et al. (2005) compared these alternative utility functions and found the multi-

linear household utility function to have a better goodness-of-fit than the iso-elastic

function for data, pertaining to Japan.

23

2.3.3 Joint Activity Participation

Research into joint decision structures is less prevalent, although it should be realized

that some models of time use include joint activity participation (see previous section).

Fujii, et al. (1999) modeled the allocation of an individual decision-maker’s time to in-

home and out-of-home activities with other family members, with non-family members

and alone, using a production function paradigm. Gliebe & Koppelman (2001) argued

that at the time of their writing no researcher has presented a model of household

decision making in which the utility of multiple decision makers is represented in both

an individual and a collective sense for the purpose of explaining joint activity

participation and travel. They assumed that the joint decision is an aggregation of

individually formed preferences and that households make activity decisions to

maximize collective utility, subject to time constraints. Individual utility is weighted by

the importance of that person to the household’s total utility. Individual utility is

assumed to be a monotonically increasing function of four components: (1)

consumption of the products of market work (subsistence activity) and household

maintenance activities; (2) satisfaction derived from participation in market work,

household maintenance and leisure activities; (3) altruism from the utilities of other

household members; and (4) companionship from participation in maintenance and

leisure activities with other household members. Overall, different explanatory

variables play different roles in the utilities of different activities for different members

in the sense that they have different values of estimated parameters and statistical

significance.

Scott & Kanaroglou (2002) developed a trivariate ordered probit model to model the

daily number of non-work, out-of-home activity episodes for household heads,

accounting for two activity settings: independent and joint activities. The differentiated

between different types of households: couple, non-workers; couple, one worker, and

couple, two workers. Significant interactions between household heads were found, the

nature of which varied by household type. Traditional gender roles were found to

persist in couple, one-worker households. In terms of predictive ability, the models

incorporating interactions were found to predict more accurately than models

excluding interaction.

Meka, Pendyala & Kumara (2002) examined interactions between two adult household

members in multi-adult households using a data set derived from a 1999 household

travel survey conducted in Southeast Florida. Daily activity and time allocations

between two household members were examined and potential trade-offs and

complementary effects were modeled simultaneously using a structural equations

modeling methodology. In particular, their focus was on causal relationships among

work and non-work activity and travel durations and frequencies. The model included

24

six endogenous variables with six significant error covariances and captured within-

person trade-offs between work and non-work activity engagement. For each person, as

the amount of work activity or travel increased, the amount of non-work activity or

travel decreased. Between persons, the model captured the complementary and joint

nature of non-work activity engagement where household members tend to pursue non-

work activities together. Thus, when one person’s non-work activity or travel

increases, so does the other person’s non-work activity travel engagement.

Srinivasan & Bhat (2006) simultaneously modeled: (1) the male’s decision to

undertake independent in-home discretionary activities and the corresponding duration,

(2) the female’s decision to undertake independent in-home discretionary activities and

the corresponding duration, (3) the male’s decision to undertake independent out-of-

home discretionary activities and the corresponding duration, (4) the female’s decision

to undertake independent out-of-home discretionary activities and the corresponding

duration, and (5) the household’s decision to undertake joint out-of-home discretionary

activities and the corresponding duration. The discrete components of the choices (i.e., the decision to undertake activity) are each modeled using the binary logit structure.

The continuous components of the choices (i.e., the activity duration) are each modeled

using a linear regression structure with the natural logarithm of the corresponding

activity duration as the choice variable.

Srinivasan & Bhat (2008) also considers joint participation accommodating intra-

household and inter-household interactions in activity-travel behavior analysis, and

examines the generation, location, and scheduling of joint activity episodes. The results

of this analysis highlight the high levels of joint activity- travel participation by

individuals. Further, independent activities are found to be different from joint

activities in systematic ways. Specifically, joint episodes are of longer durations,

significantly likely to take place at the residence of other people, and often confined to

certain time periods of the weekday. In addition, within the class of joint episodes,

important differences are also observed based on activity type, companion type, and

the day of the week.

2.3.4 Travel Arrangements

Vovsha, et al. (2005) conceptualized a ride-sharing for mandatory activities as a pure

travel arrangement, where the underlying activity for each participant is assumed to

vary between individuals. Thus, they argued that different from joint activity ride-

sharing modeling for mandatory activities does not require a generation model but

rather a linking and synchronizing model. This is a limiting conceptualization in that it

25

implicitly assumes that activities are fixed. When modeling ride-sharing, the authors

assumed that for each household member the number and purpose of mandatory tours

and their location zone, preferred departure from home, and preferred arrival back

home are known for each tour. They differentiate between outbound and inbound ride-

sharing. Their model involves two stages (i) linkage and synchronization of outbound

and inbound half-tours by means of a partition-choice model that considers all possible

partitions of mandatory half-tours into rides (alone and shared); (ii) Ordered

participation choice model that essentially considers a role of each participant (driver,

passenger) and route along which activity locations of all ride participants are visited.

To restrict the number of possible linkages thresholds, including maximum allowable

differences in departure/arrival times and maximum deviation from the shortest path to

or from the location of activity for the driver are assumed. In addition, the maximum

size of travel party was limited to 3 participants.

The person participation role model considers sequences of persons within the ride in

such a way that the first person plays the driver role, the second person corresponds to

the passenger with the longest route, and so forth. The last person is the first passenger

dropped off on the outbound half-tour or the last person picked-up on the inbound half-

tour. The last person does not experience any route deviation. The order of persons

from the driver to the shortest-leg passenger corresponds to the magnitude of potential

deviations from the shortest route.

Another interesting and in some respects more elaborate model was suggested by

Roorda, Miller & Kruchten (2006). The differentiate between joint trips, serve

passenger trip (see also next section), pure joint tours, partial joint tours, pure serve

passenger tours, and en-route serve passenger tour. The model incorporates individual

tour mode choice, vehicle allocation, a serve passenger matching procedure, and pure

serve passenger tours, and optimizes a utility function. In their application, the number

of explanatory variables was rather limited, but in principle this could be extended to

encompass a wider selection of personal, household, transportation, and especially

activity-travel pattern characteristics.

Escorting is a joint travel arrangement that is characterized by different roles of

participants. There is always an escorting adult driver and one or several escorted

children. Vovsha, et al. (2005) argue that the important characteristic that distinguishes

escorting from all other joint activity and travel arrangements is that only the escorted

persons have a purposed activity to participate while the driver does not participate in

any activity and implement a pure chauffeuring function. A dominant share of

escorting involves children as passengers. For each tour of a child that demands

escorting they distinguish five possible alternatives: 1) no escort; 2) escort in outbound

26

direction only (from home to activity), 3) escort in inbound direction only (from

activity back home), 4) escort in both direction by means of two separate tours of the

same driver or by different drivers without waiting, and 5) escort in both direction by

means of a single tour of the same driver with waiting. The set of children’s tours with

all pertinent characteristics of the person tour purpose/activity type, departure-from-

home time for outbound half-tour, arrival-back-home time for inbound half-tour, and

location is assumed known and fixed. The set of adult chauffeurs with all pertinent

characteristics of the person and availability to serve child tours within the time

window left after scheduling the chauffeur’s mandatory and joint activities (they are

considered of higher scheduling priority) is also assumed known and fixed. Escorting

tours for each chauffer are listed in a chronological order. The first escort tour can take

any outbound or inbound child half-tours that fall into the available time window of the

chauffeur, while each subsequent escorting tour of the same chauffeur has a narrower

window available since the previous tour(s) blocked out some time. Three feasible

conditions are adhered to: The bundle of outbound half tours of children served by the

tour should have close departure-from-home times and locations. A threshold was used

for bundling outbound half-tours. The bundle of inbound half tours of children served

by the tour should have close arrival-back-home times and locations, and all outbound

half tours start earlier than inbound half-tours served by the same escorting tour. These

constraints normally reduce the choice set size significantly. However, further

decomposition may be required, for example by ordering household chauffeurs. Then,

the choice model is developed for a single person and includes only residual

chauffeuring alternatives left after the choices actually made by the previously modeled

chauffeurs. Alternatively, Vovsha & Petersen (2005) suggest an ordering of child tours

demanding escort rather than an ordering of chauffeurs. The utility function then

consists of some combination of escorting utility for each child half tour (no

escort has zero utility), additional child utility of escorting in both directions, chauffeur

suitability and availability for each child half-tour, chauffeur workload saturation effect,

and chauffeur tour disutility.

2.4 HOUSEHOLD DECISION MAKING IN COMPREHENSIVE ACTIVITY-

BASED MODELS

In this section, the existing comprehensive activity-based models will be reviewed in

terms of their inclusion and treatment of household decisions. Comprehensive in this

context means that the model allows predicting a combination of choice facets, at least

compatible with those underlying traditional four-step models: i.e. activity generation,

destination and transport mode choice. Also, we restrict our discussion to fully

operational models.

27

Over the years, many activity-based models have been suggested in the literature,

including constraints-based models, micro-simulation models, (nested logit) utility-

maximizing models, suites of advanced statistical models and rule-based models (see

Timmermans, Arentze & Joh, 2002 for a recent more detailed overview).

2.4.1 Constraints-based Models

These models have primarily been developed to assess whether a planned activity

schedule is feasible, given a set of institutional and space-time constraints. These

models have a long history in activity-based analysis from the early work of

Hägerstrand and his co-workers (PESASP, Lenntrop, 1976) to more recent models

such as MASTIC (Dijst & Vidacovic, 1997) and GISICAS (Kwan, 1994, 1997).

Although there are subtle differences between these models, all have individual activity

schedules as input. Moreover, their purpose is primarily to assess accessibility

conditions and the feasibility of activity schedules as opposed to predicting activity-

travel patterns. Hence, to the best of our knowledge, these constraints-based models

have not dealt with household decision making. However, at least theoretically, it

seems straightforward to extent these models to the household level and assess the

feasibility of household activity schedules, incorporating synchronizing constraints,

possible task allocation and resource allocation. A computational problem is the

explosion of possible combinations of patterns that need to be evaluated. If the purpose

of the model is to identify the number of feasible household activity-travel patterns, a

sophisticated algorithm will be required. If the purpose is to generate a single feasible

activity-travel pattern, even a simple genetic algorithm will be sufficient (e.g.,

Charypar & Nagel, 2004; Meister, Frick & Axhausen, 2005, although it did not (yet)

account for all types of constraints typically incorporated in constraints-based models).

2.4.2 Simulation Models

Examples of these models are McNally (1997) and Ramblas (Veldhuisen, Timmermans

& Kapoen, 2000, 2005). Pribyl & Goulias (2005) are the only ones to my knowledge

who suggested an approach to simulate activity patterns that take interactions within

the family into account. Their approach consists of 6 steps. The objective of step 1 is to

find groups of households with similar activity patterns. To that effect, a K-medoid clustering is applied to the activity patterns of the household, combining the patterns of

the adults. Next, in Step 2 the probability that an individual starts a particular activity at

a particular time and its duration are derived. For every cluster and every time step, the

relative frequencies of leaving home to conduct a particular activity are derived. In

addition, for each time step, activity type and means of travel, the average duration and

28

standard deviation of duration are computed. Then, in Step 3, the identified clusters are

linked to the persons in the data set, based on their personal socio-demographic

characteristics as well as characteristics of their entire household, using a CHAID-

based decision-tree algorithm. In Step 4, the decision trees are used to assigns

households to a cluster. Following Arentze & Timmermans (2003), a probabilistic

action-assignment rule is used. Once households have been assigned to clusters, step 5

simulates daily activity patterns. The activity patterns consist of the sequence of

activities, each with their start time, duration, and the within household interactions.

The model is constructed for each time step from the proportion of cluster members

that start each particular activity within half hour on either side of the time step in

question. Travel is not treated as a separate activity, but rather as an indivisible part of

each activity. A normally distributed random number with the mean value and standard

deviation obtained from the sample for a particular activity type is used at every time

instant.

Another important issue that needs to be addressed is the simulation of alone or joint

activity participation in multi-adult households. The patterns of all adult household

members are simulated sequentially. First, the pattern of the first person in the

household (head of the household) is simulated. In case an activity is a joint activity

with the spouse, the schedule of the spouse defines an exact part of her/his schedule.

The remainder of the schedule is simulated, conditional on the derived probabilities

and the joint parts of the schedule. The probability of an activity to start at the end of

the joint activity is used.

This approach can be viewed as an effective and straightforward extension of the

simulation models that we have in mind. A potential problem of these approaches,

however, is that the simulated schedules may be infeasible within a specific spatial-

temporal context, and this problem may be exaggerated for household activity

schedules. Because these simulation models use observed data, they lack the behavioral

mechanisms of how individuals and household adjust their preferred schedules in time

and space to cope with the various types of constraints they face.

These problems may be more profound from repeated sampling from distribution, one

of the reasons why Veldhuisen, et al. (2000) sampled complete activity-travel patterns.

Alternatively, one can extract skeleton (e.g., Janssens, Wets, Brijs & Vanhoof, 2005),

but such an approach should be extended to cope with possible inconsistencies between

simulated patterns and space-time constraints. This may not be a major issue in similar

(planned) spatial contexts, but too strong an assumption if the space-time

characteristics dramatically differ. Under such circumstances, one cannot reasonably

expect that similar activity-travel patterns will or can be implemented.

29

2.4.3 Utility-Maximizing Models

Over the years, several activity-based models, founded on the principle of utility

maximization, have been suggested in the literature. Most of these have relied on

nested or GEV logit models (e.g., Kawakami and Isobe, 1982, 1989; Bowman, 1998;

Fosgerau, 1998; Wen & Koppelman, 1999; Bowman & Ben-Akiva, 2001).This

approach has been further developed for applications in the United States. An overview

of this work is given in Vovsha, Bradley & Bowman (2005). This work, however,

constitutes an exception in the sense that all other models are based on individual

patterns.

In particular, they consider three principal layers of intra-household interactions: (i)

Coordinated principal activity pattern (DAP) types at the entire-day level; (ii) Episodic

joint activity and travel and (iii) Intra-household allocation of maintenance activities.

DAP’s are coordinated to make sure that a particular household member can, for

example, take care of the children at home. They distinguish between a mandatory

pattern, further subdivided into work, university and school, and the frequency of tours,

a non-mandatory pattern and a stay home pattern. Vovsha, Peterson & Donnelly

(2004a) and Bradley & Vovsha (2005) showed a strong correlation between DAP types

of different household members. Joint activity and travel is distinguished between fully

joint travel tours for shared activities and partially joint tours, in which household

members share transportation without participation in the same activity. Finally, intra-

household allocation of maintenance activities is relevant because the allocation of

such activities to a particular household member is a function of a household decision-

making process. The following categories of out-of-home episodic joint activity and

travel are distinguished: Joint travel generated by the shared activity; Joint travel to

synchronize mandatory activities, and Escorting. This leads to a sequence of five

models: 1-coordinated DAP, 2-joint travel for shared non-mandatory activity, 3-joint

travel (ride-sharing) for mandatory activities, 4-escorting children, and 5-allocation of

maintenance tasks. Alternative DAP types are broken down into a group, containing

individual mandatory activities and a group containing non-mandatory and staying at

home patterns that potentially can be conducted jointly by several household members.

From a modeling perspective, the authors have attempted different structures. First,

they adopted a sequential processing of persons according to an intra-household

hierarchy in several regional travel models in US (Vovsha, et al., 2004a). Second,

simultaneous modeling of potentially joint alternatives for all household members with

subsequent modeling of individual alternatives can be attempted. This involves for

each household member a trinary choice (M, NM, H) and modelling sub-choice of the

mandatory alternative by a separate choice model conditional upon the choice of

mandatory alternative in the trinary choice (Bradley & Vovsha, 2005). Finally, a

30

parallel choice structure that considers combinations of main trinary choices at the

upper level and individual sub-choices simultaneously in one choice structure can be

applied (Gliebe, 2004; Gliebe & Koppelman, 2004, 2005). These nests correspond to

the combination of activities where joint participation is essential. The structure of

these nests captures different levels of intra-household interaction. Under each nest, the

correspondent individual choices of mandatory alternatives are considered for each

person individually.

Episodic joint non-mandatory activities are associated with fully-joint travel tours. A

frequency-choice model is used to predict the number of joint tours by purpose /

activity type at the household level. Subsequently, a Person participation choice model

predicts probability of having a certain participation matrix conditional upon the

chosen set of joint tours (Vovsha, Person & Donelly, 2003, 2004). The various

structures are modeled by a simple nested logit model or generalized nested structures

of the GEV class.

Allocation of maintenance tasks to individual household members is modeled as a two

step process. First, a frequency choice model predicts the number of maintenance tasks.

Next, each task is assigned to a particular household member. Vovsha, Peterson &

Donnelly (2004b) applied a task allocation choice model that is applied for each task

independently and returns a choice probability of a person the most suitable for the task

as a function of the activity type and person characteristics (person type, residual time

window left after mandatory activities, the number of joint and escorting tours in which

the person participates, etc). Next the resulting fractional matrix of allocation

probabilities obtained at the previous stage is discretionized instead of a simple Monte-

Carlo pick for each row because independent picks for each task may result in illogical

allocations with one person overloaded while others may have no tasks.

2.4.4 FAMOS

FAMOS (Pendyala, et al. 2005), derived from HAGS/PCATS developed by Kitamura

and his colleagues in Japan (e.g., Kitamura & Fujii, 1998), is a micro-simulator of

individual-level activity-travel patterns. The model system does not include explicit

household-level allocation models, but the individual-level models do incorporate

“intra-household interaction” effects. The individual activity type choice models, for

example, incorporate variables reflecting household demographics and associated

activity needs. The individual mode choice models consider household vehicle

availability and the micro-simulator keeps track of the availability or non-availability

of household vehicles at any point in time. There is no explicit model of joint activity

31

engagement; however, household level activity-travel patterns (including joint travel)

could be constructed/deduced from the simulated individual-level activity-travel

patterns.

2.4.5 CEMDAP and CEMDAP-2

This model system, developed by Bhat and his co-workers (2004), can best be viewed

as a suite of separate models, predicting the activity travel patterns of workers and non-

workers, and students and non-students. In turn, for some segments, the patterns are

further broken down into sub-patterns. For example, the daily pattern of workers is

characterized by four different sub-patterns: before-work pattern, commute pattern,

work-based pattern, and after-work pattern. Within each before-work, work-based and

after-work patterns, there might be several tours.

Considering practical implementation constraints, certain restrictions are imposed on

the maximum number of tours and the maximum number of stops in any tour. A set of

22 different advanced econometric models is used to predict different facets of activity-

travel patterns, where the type of model chosen does justice to the statistical properties

of the data. This is a strong feature of this model system. On the other hand, where one

of the major objectives of developing activity-based models was improved integrity

and better capturing the many interdependencies in activity-travel patterns, the

CEMDAP models does not differ that much in a fundamental sense from traditional

four step models. First, activity generation is predicted, much as trip generation used to

be modeled. The explanatory variables are largely socio-demographics data:

constraints and characteristics of the daily patterns do not play a major role: pattern

level characteristics are limited in number of level of detail. In terms of household

decisions, the model system is primarily based on individual choices. Household

characteristics sometimes are used as explanatory variables, but processes such as

coordination, synchronization, etc are not explicitly represented.

The new version of CEMDAP considers joint activities, though data constraints did not

allow considering all possible joint activities among each subset combination of

individuals in the household. This is seen as a separate choice and hence, the utility of

joint activities against individual activities is not addressed in much detail. Car

allocation (as needed) and task allocation are also modeled as part of CEMDAP.

Children's activity-travel behavior is explicitly modeled in CEMDAP now, which is a

distinct feature.

32

2.5 CONCLUSIONS

This chapter described several studies applications in household decision making.

Incorporating household decision-making into activity-based models of transport

demand receives increasing attention in recent times. The out-of-home activity needs

synchronization between persons in particular in multi-person households. Problems

such as activity allocation, joint activity participation and resource allocation have in

the past typically been addressed separately, i.e. not in the context of a scheduling

process or at most ad hoc as part of a more comprehensive activity-based model of

transport demand. Decisions to travel jointly in multi-person households require joint

decision between persons.

The comprehensive operational system of activity-based travel demand modeling is

still few exist until recently. ALBATROSS concerns about joint decision making

matter. Therefore, the development of ALBATROSS could enhance the studies of

activity-scheduling models in activity-based transport demand modeling.

REFERENCES

Aitken, S.C. (2000), Fathering and Faltering: ‘Sorry, But You Don't Have the

Necessary Accoutrements’, Environment and Planning A, 32, 581-598

Antill, J.K. & Cotton, S. (1988), Factors Affecting the Division of Labour in

Households, Sex Roles, 18, 531-553

Arentze, T.A. & Timmermans, H.J.P. (2000), ALBATROSS: A Learning Based Transportation Oriented Simulation System, European Institute of Retailing and

Services Studies, Eindhoven, the Netherlands.

Arentze, T.A. and Timmermans, H.J.P. (2004), A Learning-Based Transportation

Oriented Simulation System, Transportation Research Part B, 38, 613-633.

Arentze, T.A. & Timmermans, H.J.P. (2005), ALBATROSS 2.0: A Learning Based Transportation Oriented Simulation System, European Institute of Retailing and

Services Studies, Eindhoven, the Netherlands.

Arentze, T.A. and Timmermans, H.J.P. (2007), Parametric action decision trees:

Incorporating continuous attribute variables into rule-based models of discrete

choice, Transportation Research B, 41, 772-783.

Atkinson, A.B. (1970), On the Measurement of Inequality, Journal of Economic Theory, 2, 244-263.

Bhat, C.R., Guo, J., Srinivasan, S., Sivakumar, A., Pinjari, A. and Eluru, N. (2004),

Activity-Based Travel-Demand Modeling for Metropolitan Areas in Texas:

33

Representation and Analysis Frameworks for Population Updating and Land-Use

Forecasting, Report 4080-6, prepared for the Texas Department of Transportation.

Boarnet, M.G. and Crane, R. (2001), Travel by Design: The Influence of Urban Form on Travel, Oxford University Press, Oxford.

Borgers, A.W.J., Hofman, F., Ponje, M. and Timmermans, H.J.P. (2001), Towards A

Conjoint-Based, Context-Dependent Model of Task Allocation in Activity Settings:

Some Numerical Experiments, Journal of Geographical Systems, 3, 347-367.

Borgers, A.W.J., Hofman, F. and Timmermans, H.J.P. (2002), Conditional Choice

Modeling of Time Allocation Among Spouses in Transport Settings, European Journal of Transportation and Infrastructure Research, 2, 5-17.

Borgers, A.W.J. and Timmermans, H.J.P. (1993), Transport Facilities and Residential

Choice Behavior: A Model of Multi-person Choice Processes, Papers in Regional Science, 72, 45-61.

Bowman, J.L. (1998), The Daily Activity Schedule Approach to Travel Demand

Analysis. PhD dissertation, MIT.

Bowman, J.L. and Ben-Akiva, M. (2001), Activity-Based Disaggregate Travel Demand

Model System with Activity Schedules, Transportation Research A, 35, 1-28.

Bradley, M. and Vovsha, P. (2005), A Model for Joint Choice of Daily Activity

Pattern Types of Household Members, Paper presented at 84th Annual Meeting

Transportation Research Board, Washington, D.C.

Cao, X. and Chai, Y. (2007). Gender-role based differences in time allocation: A case

study of Shenzhen, China. Paper presented in the 86th Annual Meeting of the

Transportation Research Board, January 2007, Washington, D.C.

Chandraskharan, B. and Goulias, K..G. (1999), Exploratory Longitudinal Analysis of

Solo and Joint Trip Making in the Puget Sound Transportation Panel.

Transportation Research Record ,1676, 77-85.

Charypar, D. and Nagel, K. (2003), Generating Complete All-Day Activity Plans with

Genetic Algorithms. Paper presented at the 10th International Conference on Travel

Behaviour Research, Lucerne, Switzerland.

Curry, D.J. and Menasco, M.B. (1979), Some Effects of Differing Information

Processing Strategies on Husband-Wife Joint Decision, Journal of Consumer Research, 6, 192-203.

Curry, D.J., Menasco, M.B. and van Ark, J.W. (1991), Multiattribute Dyadic Choice:

Models and Tests, Journal of Marketing Research, 28, 259-267.

Davis, H.L. and Rigeaux, B.P. (1974), Perception of Marital Roles in Decision

Processes, Journal of Consumer Research, 1, 51-61.

Dijst, M. and Vidacovic, V. (1997), Individual Action Space in the City. In: D.F.

Ettema & H.J.P. Timmermans (eds.), Activity-Based Approaches To Travel Analysis, Pergamon, Oxford, pp. 117-133.

34

Disanayake, D. and Morikawa, T. (2002), A Nested Logit Model of Vehicle-

Ownership and Household Participation on Trip Chaining for Commuter Travel in

Developing Countries, Paper presented at the 81th

Annual Meeting Transportation

Research Board, Washington, D.C.

Dowling, R. (2000), Cultures of Mothering and Car Use in Suburban Sydney: A

Preliminary Investigation, Geoforum, 31, 345-353.

Eliashberg, J. and Winkler, R.L. (1981), Risk Sharing and Group Decision Making, Management Science, 27, 1221-1235.

Eliashberg, J. and Winkler, R.L. (1986), Assessing the Predictive Accuracy of Two

Utility-based Theories in a Marketing Channel Negotiation Context, Journal of Marketing Research, 23, 101-110.

Ettema, D.F., Schwanen, T., and Timmermans, H.J.P. (2004), The Effect of Locational

Factors on Task and Time Allocation in Households, Proceedings 83rd TRB Conference, Washington, D.C. (CD-Rom).

Ettema, D.F., Schwanen, T., and Timmermans, H.J.P. (2006), The Effect of Location,

Mobility and Socio-Demographic Factors on Task and Time Allocation in

Households, Transportation, 34 (forthcoming).

Ettema, D.F.and Van der Lippe, T. (2006), Weekly Rhythms in Task and Time

Allocation of Household. Paper presented at TRB 85th Annual Meeting. Washington

DC.

Fagnani, J. (1987), Daily Commuting Time: The Stakes for Working Mothers in

France. Transportation Research Record, 1135, 26-30.

Fosgerau, M. (1998), PETRA: An Activity Based Approach to Travel Demand

Analysis, Paper presented at the 8th

WCTR Conference, Antwerp.

Fujii, S., Kitamura, R., and Kishizawa, K. (1999), Analysis of Individuals’ Joint

Activity Engagement Using a Model System of Activity-travel Behavior and Time

Use, Transportation Research Record, 1676, 11-19.

Gliebe, J.P. and Koppelman, F.S. (2000), A Model of Joint Activity Participation,

Paper Presenting at the 9th International Association for Travel Behavior

Conference, Gold Coast, Queensland, Australia, 2-7 July.

Gliebe, J.P. and Koppelman, F.S. (2001), A Model of Joint Activity Participation,

Paper Presented at the 80th Annual Meeting Transportation Research Board,

Washington, D.C.

Gliebe, J.P. and Koppelman, F.S. (2002), A Model of Joint Activity Participations

between Household Members, Transportation, 29, 49-72.

Gliebe, J.P. and Koppelman, F.S. (2005a), An Integrated Model of Individual and Joint

Daily Activity-travel Patterns Incorporating Inter-agent Choice Set Constraints,

Transportation Research B, forthcoming.

35

Gliebe, J.P. and Koppelman, F.S. (2005b), Modeling Household Activity-Travel

Interactions as Parallel Constrained Choices, Paper Presented at 84th Annual

Meeting Transportation Research Board, Washington, D.C.

Golob, T.F., Kim, S., and Ren, W. (1996), How Households Use Different Types of

Vehicles, Transportation Research A, 30, 103-118.

Golob, T.F. and McNally, M.G. (1997), A Model of Household Interactions in Activity

Participation and the Derived Demand for Travel, Transportation Research, 31B,

177-194.

Goulias, K. (2002), Multilevel Analysis of Daily Time Use and Time Allocation to

Activity Types Accounting for Complex Covariance Structures Using Correlated

Random Effects, Transportation, 29, 31-48

Gronau, R. (1977), Leisure, Home Production, and Work - the Theory of the

Allocation of Time Revisited, Journal of Political Economy, 85, 1099-1123.

Gupta, S., and Livne, Z.A. (1988), Resolving a Conflict Situation with a Reference

Outcome: An Axiomatic Model, Management Science, 34, 1303-1314.

Hanson, S. and Hanson, P. (1980), Gender and Urban Activity Patterns in Uppsala,

Sweden, Geographical Review, 70, 291-299.

Hanson, S. and Hanson, P. (1981), The Impact of Married Women's Employment on

Household Travel Patterns: A Swedish Example, Transportation, 10, 165-183.

Hanson, S. and Johnston, I. (1985), Gender Differences in Work-trip Length:

Explanations and Implications, Urban Geography, 6, 193-219.

Hanson, S. and Pratt, G. (1990), Geographic Perspectives on the Occupational

Segregation of Women, National Geographic Research, 6, 376-399.

Harsanyi, J.C. (1955), Cardinal Welfare, Individualistic Ethics, and Interpersonal

Comparisons of Utility, Journal of Political Economy, 63, 309-321.

Huber, J. and Spitze, G. (1983), Sex Stratification: Children, Housework and Jobs.

Academic Press, New York.

Hunt, K. and Petersen, E. (2005), The Role of Gender, Work Status and Income in

Auto Allocation Decisions, Proceedings of the Annual TRB Meeting, Washington,

D.C.

Janssens, D., Wets, G, Brijs, T., and Vanhoof, K. (2005), Simulating Daily Activity

patterns Through The Identificatoin of Sequential Dependencies, In Timmermans,

H.J.P. (ed.), Progress in Activity-Based Analysis, Pergamon Press, Oxford, pp. 67-

89.

Kawakami, S. and Isobe, T. (1982), Development of a one day travel-activity

scheduling model for workers. In S. Carpenter & P. Jones (eds.), Recent Advances in Travel Demand Analysis, Gower, Aldershot, pp. 314-344.

Kawakami, S. and Isobe, T. (1989), Development of a scheduling model considering

time constraint and temporal transferability test of the model, Selected Proceedings of the 5th WCTR Conference, Vol. IV, pp. 221-233.

36

Keeney, R.L. (1972), Utility Functions for Multi-attributed Consequences,

Management Science, 18, 209-222.

Knijn, T. (2004), Challenges and Risks of Individualisation in the Netherlands, Social Policy & Society, 3, 57-65

Kitamura, R. and Fujii, S. (1998), Two Computational Process Models of Activity-

Travel Choice, In: T. Garling, T. Laitila & K. Westin (eds.): Theoretical Foundations of Travel Choice Modeling, Elsevier, Oxford, 251-279.

Kostyniuk, L.P. and Kitamura, R. (1983), An Empirical Investigation of Household

Time Space Paths. In: S. Carpenter & P. Jones (eds.), Recent Advances in Travel Demand Analysis, Gower Publishing Co., Ltd., Aldershot, U.K.

Krishnamurthi, L. (1989), Conjoint Models of Family Decision Making, International Journal of Research in Marketing, 5, 185-198.

Kwan, M-P. (1994), GISICAS: A GIS-Interfaced CPM for Activity Scheduling in ATIS,

unpublished dissertation, University of California at Santa Barbara, USA.

Kwan, M-P. (1999a), GISICAS: An Activity-Based Travel Decision Support System

Using A GIS-Interfaced Computational Process Model. In: D.F. Ettema & H.J.P.

Timmermans (eds.), Activity-Based Approaches To Travel Analysis, Pergamon,

Oxford, pp. 263-282.

Kwan, M-P. (1999b), Gender, the Home-Work Link and Space-Time Patterns of Non-

Employment Activities, Economic Geography, 76, 370-394

Kwan, M-P. (2000), Gender Differences in Space-Time Constraints, Area, 32, 145-

156

Lawson, C.T. (1999), Household Travel/Activity Decision: Who Wants to Travel?

Paper Presented at the 78th Annual Meeting Transportation Research Board,

Washington, D.C.

Lee, Y. and Hickman, M. (2004), Household Type and Structure, Time Use Pattern,

and Trip Chaining Behavior, Paper Presented at the 83th

Annual Meeting


Lenntrop, B. (1976), Paths in Space-Time Environment. Lund Studies in Geography,

No. 44, Royal University of Lund, Lund, Sweden.

Leuthold, J. (1968), An Empirical Study of Formula Income Transfers and the Work

Decisions of the Poor, Journal of Human Resources, 3, 312-323.

Louviere, J.J. (1984), Hierarchical Information Integration: A New Method for the

Design and Analysis of Complex Multiattribute Judgement Problems, Advances in Consumer Research, 11, Association for Consumer Research, Provo, UT

Louviere, J.J. and Timmermans, H.J.P. (1990), Hierarchical Information Integration

Applied To Residential Choice Behaviour, Geographical Analysis, 22, 127-145

Lu, X. and Pas, E.I. (1997), A Structural Equation Model of the Relationships among

Socio-demographics, Activity Participation and Travel Behavior, Paper Presented at

the 76th

Annual Meeting of Transportation Research Board, Washington, D.C.

37

Lundberg, S. and Pollak, R.A. (1993), Separate Spheres Bargaining and The Marriage

Market, Journal of Political Economy, 101, 988-1010.

Madden, J. F. (1981), Why Women Work Closer to Home, Urban Studies, 8, 181-194.

McLafferty, S. and Preston, V. (1991), Gender, Race, and Commuting Among Service

Sector Workers, Professional Geographer, 43, 1-15.

Meister, K., Frick, M. and Axhausen, K.W. (2005), A GA-Based Household

Scheduler, Proceedings of the TRB Annual Conference, Washington, D.C. (Cd-

Rom)

Meka, S., R. Pendyala R., and Kumara, M. (2002), A Structural Equations Analysis of

Within-household Activity and Time Allocation between Two Adults, Proceedings

81st TRB Conference, Washington, D.C. (CD-Rom).

Messer, W.F. and Emery, D.R. (1980), Some Cautionary Notes on the Use of Conjoint

Measurement for Human Judgement Modelling, Decision Sciences, 11, 678-690.

Miller, E.J. (2004), An Integrated Framework for Modelling Short- and Long-Run

Household Decision-Making, Paper Presented at the EIRASS Conference on

Progress in Activity-Based Analysis, Maastricht, the Netherlands, May 28-31.

Miller, E.J. (2005), Propositions for Modelling Household Decision-Making, In: M.

Lee-Gosselin and S.T. Doherty (Eds.), Integrated Land-Use and Transportation Models: Behavioural Foundations, Elsevier, Amsterdam.

Miller, E.J., Nurul, H., and Kandker, M. (2006), Modeling Individuals’ Frequency and

Time Allocation Behavior for Shopping Activities Considering Household Level

Random Effects, Proceedings TRB Annual Meeting, Washington, D.C. (Cd-Rom).

Miller, E.J. and Roorda, M.J. (2003), A Prototype Model of Household Activity-Travel

Scheduling, Paper Presented at the 82th Annual Meeting Transportation Research

Board, Washington, D.C.

Miller, E., Roorda, M.J., and Carrasco, J. (2003), A Tour-based Model of Travel

Choice. Proceedings 10th IATBR Conference, Lucerne, Switzerland (CD-Rom).

Mohammadian, A. and Miller, E. (2003), Dynamic Modeling of Household

Automobile Transactions, Transportation Research Record, 1831, 98-105.

Molin, E.J.E., Oppewal, H., and Timmermans, H.J.P. (1996), Predicting Consumer

Response to New Housing: A Stated Choice Experiment, Netherlands Journal of Housing and the Built Environment, 11, pp. 297-312.

Molin, E.J.E., Oppewal, H., and Timmermans, H.J.P. (1997), Modeling Group

Preferences Using a Decompositional Preference Approach, Group Decision and Negotiation, 6, 339-350.

Molin, E.J.E., Oppewal, H., and Timmermans, H.J.P. (1999), Group-Based versus

Individual-Based Conjoint Preference Models of Residential Preferences: a

Comparative Test, Environment and Planning A, 31, pp. 1935-1947.

38

Molin, E.J.E., Oppewal, H., and Timmermans, H.J.P. (2000), A Comparison of Full

Profile and Hierarchical Information Conjoint Methods in Modeling Group

Preferences, Marketing Letters, 11, 165-172.

Molin, E.J.E., Oppewal, H., and Timmermans, H.J.P. (2001), Group-based versus

individual-based conjoint preference models of residential preferences: a

comparative test. Environment and Planning A, 31, 1935-1947.

Molin, E.J.E., Oppewal, H., and Timmermans, H.J.P. (2002), Conjoint Modelling of

Residential Group Preferences: A Comparison of the Internal Validity of

Hierarchical Information Integration Approaches. Journal of Geographical Systems,

4, pp. 343-358.

Molin, E.J.E. and Timmermans, H.J.P. (2003), Testing Hierarchical Information

Integration Theory: the Causal Structure of Household Residential Satisfaction.

Environment and Planning A, 35, pp. 43-58.

Morris, L. (1990), The Workings of the Household, Polity Press, Cambridge.

Munsinger, G.M., Weber, J.E. & Hansen, R.W. (1975), Joint Home Purchasing

Decisions by Husbands and Wives, Journal of Consumer Research, 1, 60-66.

Nash, J.F. (1950), The Bargaining Problem, Econometrica, 18, 155-162.

Nash, J.F. (1953), Two-Person Cooperative Games, Econometrica, 21, 128-140.

McNally, M.G. (1997), An Activity-Based Micro-Simulation Model for Travel

Demand Forecasting. In: D.F. Ettema & H.J.P. Timmermans (eds.), Activity-Based Approaches to Travel Analysis, Pergamon, Oxford, pp. 37-54.

Oppewal, H, Timmermans, H.J.P., and Louviere, J.J. (1994), Modelling Hierarchical

Conjoint Processes With Integrated Choice Experiments, Journal of Marketing Research, 31, 92-105

Pendyala, R., Kitamura, R., Kikuchi, A., Yamamoto, T., and Fujii, S. (2005), Florida

Activity Mobility Simulator, Transportation Research Record, 1921, 123-130.

Petersen, P.E. and Donnelly, R. (2003), Explicit Modeling of Joint Travel by

Household Members: Statistical Evidence and Applied Approach, Paper Presented

at the 82th Annual Meeting Transportation Research Board, Washington, D.C.

Petersen, E. and Vovsha, P. (2005), Auto Allocation Modeling in Activity-Based

Models, Presented at the 10th Transportation Planning Application Conference,

TRB, Portland, Oregon.

Petersen, E. and Vovsha, P. (2006), Intrahousehold Car Type Choice for Different

Travel Needs, Paper Presented at 85th Annual Meeting Transportation Research

Board, Washington, D.C.

Pinch, S. and Storey, A. (1992), Who does what, where?: A Household Survey of

Domestic Labour in Southampton, Area, 24, 5-12.

Presser, H.B. (1994), Employment Schedules Among Dual-Earner Spouses and the

Division of Household Labor by Gender, American Sociological Review, 94, 348-

364.

39

Presser, H.B. (2003), Working in a 24/7 Economy: Challenges for American Families, Russell Sage Foundation, New York.

Pribyl. O. and Goulias, K.G. (2005), Simulation of Daily Activity Patterns. In: H.J.P.

Timmermans (ed.), Progress in Activity-Based Analysis, Pergamon, Oxford, 43-66.

Recker, W.W. (1995), The Household Activity Pattern Problem: General Formulation

and Solution, Transportation Research, 29B, 61-77.

Redman, B.J. (1980), The Impact of Women's Time Allocation on Expenditure for

Meals Away from Home and Prepared Foods, American Journal of Agricultural Economics, 62, 234-237.

Rich, J.H. (2000), Modeling Two-worker Households, Paper Presenting at the 9th

International Association for Travel Behavior Conference, Gold Coast, Queensland,

Australia, 2-7 July.

Roorda, M., Miller, E.S., and Kruchten, N. (2006), Incorporating Within-Houshold

Interactions into a Mode Choice Model Using a Genetic Algorithm for Parameter

Estimation, Proceedings of the Annual TRB Meeting, Washington, D.C.

Rosenbloom, S. (1987), The Impact of Growing Children on Their Parents' Travel

Behavior: A Comparative Analysis, Transportation Research Record 1135, 17-25.

Schwanen, T, Ettema, D.F., and Timmermans, H.J.P. (2006), If You Pick up the

Children, I’ll do the Groceries: Spatial Differences in Between-partner Interactions

in Out-of-home Household Activities, To appear in Environment and Planning A. Scott, D.M. (2001), A New Approach to Modeling Household Activity-Episode

Generation that Captures Interactions Between Household Heads, Model

Development and Empirical Analysis, Paper Presented at the 80th Annual Meeting


Scott, D.M. and Kanaroglou, P.S. (2002), An Activity-Episode Generation Model That

Captures Interactions Between Household Heads: Development and Empirical

Analysis, Transportation Research B, 36, 875-896.

Senbil, M., Zhang, J., and Fujiwara, A. (2008). Child Effect on Household Non-

Commute Trips. Paper presented in the 87th Annual Meeting of the Transportation

Research Board, January 2008, Washington, D.C.

Simma, A. and Axhausen, K.W. (2001), Within-household Allocation of Travel – The

Case of Upper Austria, Transportation Research Record, 1752, 69-75.

Singell, L.D. and Lillydahl, J.H. (1986), An Empirical Analysis of the Commute to

Work Patterns of Males and Females in Two-Earner Households, Urban Studies, 2,

119-129.

Srinivasan, K.K. and Athuru, S.R. (2005), Analysis of Within-Household Effects and

Between Household Differences in Maintenance Activity Allocation.

Transportation, 32, 495-521.

40

Srinivasan, S. and Bhat, C.R. (2004), Modeling the Generation and Allocation of

Shopping Activities in a Household, Proceedings of the TRB Conference,

Washington (CD-Rom).

Srinivasan, S. and Bhat, C.R. (2005), Modeling Household Interactions in Daily In-

Home and Out-of-Home Maintenance Activity Participation. Transportation, 32,

523-544.

Srinivasan, S. and Bhat, C.R. (2006), A Multiple Discrete-Continuous Model for

Independent- and Joint- Discretionary-Activity Participation Decisions,

Proceedings of the TRB Conference, Washington (CD-Rom).

Srinivasan, S. and Bhat, C. (2008). An Exploratory Analysis of Joint-Activity

Participation Characteristics Using the American Time Use Survey. Paper presented

in the 87th Annual Meeting of the Transportation Research Board, January 2008,

Washington, D.C.

Stopher, P.R. and Metcalf, H.M.A. (1999), Household Activities, Lifecycle, and Role

Allocation: Tests on Datasets from Boston and Salt Lake City, Transportation Research Record, 1676, 95-102.

Stopher, P.R. and Metcalf, H.M.A. (2000), Allocation Time to Activities and

Household Members: Results from Four Urban Areas, Proceedings IATBR Conference, Gold Coast, Australia (CD-Rom).

Strathman, J.G., Dueker, K.J., and Davis, J.S. (1994), Effects of Household Structure

and Selected Travel Characteristics on Trip Chaining, Transportation, 24, 23-45.

Timmermans, H.J.P., Borgers, A. Van Dijk, J., and Oppewal, H. (1992), Residential

Choice Behaviour Of Dual Earner Households: A Decompositional Joint Choice

Model, Environment and Planning A, 24, 517-533.

Timmermans, H.J.P., Arentze, T.A. and Joh, C.-H. (2002), Analyzing Space-time

Behavior: New Approaches to Old Problems, Progress in Human Geography, 26,

175-190.

Timmermans, H.J.P. (2006), Analyses and Models of Household Decision Making

Processes. In: Proceedings of the IATBR Conference, Kyoto (CD-ROM, 34 p.p).

Turner, T. and Niemeier, D. (1997), Travel to Work and Household Responsibility:

New Evidence, Transportation, 24, 397-419.

Vadarevu, R. and Stopher, P.R. (1996), Household Activities, Lifecycle and Role

Allocation, Transportation Research Record, 1556, 77-85.

Vance, C. and Iovanna, R. (2007). Gender and the Automobile: An Analysis of Non-

work Service Trips. Paper presented in the 86th Annual Meeting of the

Transportation Research Board, January 2007, Washington, D.C.

Veldhuisen, K., Timmermans, H.J.P., and Kapoen, L.L. (2000), Ramblas: A Regional

Planning Model Based on the Micro-Simulation of Daily Activity Travel Patterns,

Environment and Planning A, 32, 427-443.

41

Veldhuisen, K.J., Timmermans, H.J.P., and Kapoen, L.L. (2005), Simulating the

effects of urban development on activity-travel patterns: an application of Ramblas

to the Randstad North Wing, Planning and Design, 32, 567-580.

Vovsha, P., Bradley, M., and Bowman, J.L. (2005), Activity-Based Travel Forecasting

Models in the United States: Progress Since 1995 and Prospects for the Future. In:

H.J.P. Timmermans (ed.), Progress in Activity-Based Analysis, Pergamon, Oxford,

389-414.

Vovsha, P., Gliebe, J., Peterson, E., and Koppelman, F.S. (2005), Sequential and

Simultaneous Choice Structures for Modelling Intra-Household Interactions in

Regional Travel Models. In: H.J.P. Timmermans (ed.), Progress in Activity-Based Analysis, Pergamon, Oxford, 223-258.

Vovsha, P. and Petersen, E. (2005), Escorting Children to School: Statistical Analysis

and Applied Approach, Proceedings 84th TRB Conference, Washington, D.C. (CD-

Rom).

Vovsha, P., Petersen, E. and Donnelly, R. (2003), Explicit Modeling of Joint Travel by

Household Members: Statistical Evidence and Applied Approach, Transportation Research Record, 1831, 1-10.

Vovsha, P., Petersen, E. and Donnelly, R. (2004a), A Model for Allocation of

Maintenance Activities to the Household Members, Paper Presented at the 83th

Annual Meeting Transportation Research Board, Washington, D.C.

Vovsha, P., Petersen, E. and Donnelly, R. (2004b), Impact of Intra-Household

Interactions on Individual Daily Activity-Travel Patterns, Proceedings 83rd TRB Conference, Washington, D.C. (CD-Rom).

Vovsha, P. and Petersen, E. (2007), Incorporating Car-Type Preferences and

Intrahousehold Car Allocation into Travel Demand Models. Paper presented in the

86th Annual Meeting of the Transportation Research Board, January 2007,

Washington, D.C.

Wen, C-H. and Koppelman, F.S. (1999), An Integrated System of Stop Generation and

Tour Formation for the Analysis of Activity and Travel Patterns, Transportation Research Record, 1676, 136-144.

Wen, C. and Koppelman, F.S. (2000). A Conceptual and Methodological Framework

for the Generation of Activity-travel Patterns, Transportation, 27, 5-23.

White, M.J. (1986), Sex Differences in Urban Commuting Patterns, The American Economic Review, 76, 368-332.

Zhang, J. and Fujiwara, A. (2004), Representing Heterogeneous Intra-household

Interaction in the Context of Time Allocation, Paper Presented at the 83th

Annual

Meeting of Transportation Research Board, Washington D.C.

Zhang, J., A. Fujiwara, Timmermans, H.J.P., and Borgers, A.W.J. (2005), An

Empirical Comparison of Alternative Models of Household Time Allocation. In:

42

Timmermans, H.J.P. (ed.) Progress in Activity-Based Analysis, Elsevier, Oxford,

259-284.

Zhang, J., Timmermans, H.J.P., and Borgers, A.W.J. (2002), A Utility-Maximizing

Model of Household Time Use for Independent, Shared and Allocated Activities

Incorporating Group Decision Mechanism, Transportation Research Record, 1807,

1-8.

Zhang, J., Timmermans, H.J.P., and Borgers, A.W.J. (2004), Model Structure Kernel

for Household Task and Time Allocation Incorporating Household Interaction and

Inter-activity Dependency, Proceedings 83rd TRB Conference, Washington, D.C.

(CD-Rom).

Zhang, J., Timmermans, H.J.P., and Borgers, A.W.J. (2005a), Comparative Analysis of

Married Couples’ Task and Time Allocation Behavior on Weekdays vs. Weekends,

Paper presented at 84th Annual Meeting Transportation Research Board,

Washington, D.C.

Zhang, J., Timmermans, H.J.P., and Borgers, A.W.J. (2005b), A Model of Household

Task Allocation and Time Use, Transportation Research B, 39, 81-95.

43

Chapter 3

FRAMEWORK

3.1 INTRODUCTION

Over the last decades, it has become increasingly more recognized that travel choices

are extensively reliant on choices to participate in activities. Travel is a demand

derived from individual’s needs to perform out-of-home activities. Focusing on

activities allows one to take into account the interactions between persons in the

households and constraints imposed on activity schedules that emerge, for example,

from limited opening hours of facilities, working times and household needs (e.g.,

escorting a child to school). Therefore, there is generally a case of inter-dependency in

the travel choices and activity-agenda choices between persons within the same

households. Individuals as households’ members, particularly heads of household,

should play an important role in modeling activity and travel decision making. Their

decisions to engage in out-of-home activities, for example, often depend on household

needs and should take into account the presence of children, if any, accessibility, car

availability, etc. An out-of-home maintenance activity may be conducted

independently or jointly and household tasks often need to be allocated to household

members.

Although the need to incorporate household decision making has been acknowledged

from the beginning, this topic has only recently received little attention, and a

44

comprehensive model system at this level is still missing (Zhang et al., 2005, for

examples and Timmermans, 2006, for a review paper). Several works have been made

on the interactions of individuals within households (Gliebe and Koppelman, 2002,

2005; Scot and Kanaroglou, 2002; Srinivasan and Bhat, 2004), nevertheless, fewer

attempts to integrate the interactions in activity-scheduling models. Activity-scheduling

models share an objective to predict the sequence of decisions that leads to an observed

activity pattern of a household/individual. Activity-based models aim at predicting on a

daily basis and for a household which activities are conducted, with whom, for how

long, at what time, the location, and which transport mode is used when traveling is

involved (Arentze and Timmermans, 2000, 2005, and Miller and Roorda 2003).

ALBATROSS is one of the few operational activity-based models incorporating

household decision making (Arentze and Timmermans, 2000, 2004, 2005). It is a rule-

based computational process model developed for The Dutch Ministry of

Transportation, Public Works and Water Management. ALBATROSS differs from

other models, which use utility maximization as a framework for modeling activity

scheduling decisions. In contrast, ALBATROSS uses IF-THEN rules as a formalism to

represent and predict activity-travel choices of individuals and households. The

decision rules are extracted from activity diary data in the form of a decision tree by

using a CHAID-based decision tree induction method.

The objective of this thesis is principally to improve the ALBATROSS model to

elaborate household decision making more explicitly so that the interaction between

persons in the same household can be captured well. To accomplish these goals, some

components are elaborated, in particular in term of resource allocation, task allocation

and joint activity participation and car allocation. The structure of this thesis follows

the process model underlying ALBATROSS.

The purpose of this chapter is to give an overview of the research framework. Firstly,

we will review the ALBATROSS model and household decision making. It is followed

by introducing the new version of ALBATROSS that is developed in this PhD thesis.

In this section, how household decisions were modeled in the original ALBATROSS

model is also discussed. This is the state of the art against which to position this PhD.

Subsequently, the decision tree induction method is explained and finally, it ends up

with the conclusion.

3.2 ALBATROSS AND HOUSEHOLD DECISION MAKING

An important dissimilarity between the utility-maximizing models and a rule-based

model, such as ALBATROSS (Arentze & Timmermans, 2000, 2004), is that the former

45

models predict the choice probability of multi-faceted choice alternatives defined by

the modeler. In contrast, rule-based models do not a priori assume certain multi-faceted

choice alternatives but induce choice rules, based on a process model, for specific

choice facets. Activity-travel patterns that emerge are not a priori assumed and

classified.

Keeping this in mind, ALBATROSS simultaneously generates activity schedules for

individual household members, in which activity selection of one household member

depends on the activity schedule of the other adult, if any, in the household. In case the

number of cars is less than the number of drivers, a decision tree, representing choice

heuristics is used to assign the car to particular activities/trips of household members.

The result serves as one of the condition variables for other choices and car use is

systematically traced throughout the prediction of activity-travel schedules to create

dynamic choice sets/action spaces, which are used to check for any violations of space-

time constraints.

Joint activity participation are not separately and explicitly modeled. However, bring

and get activities (and sub-classifications if so desired) constitute one class of activities,

while travel party is a structural choice facet of the model, implying that ridesharing,

escorting and chauffeuring are endogenously generated by the model. In addition, it

means that these household activities are also predicted in terms of all other choice

facets included in the same (timing, duration, location, trip chaining, and transport

mode). Having said that, several aspects can be further improved.

Another partly rule-based model is Tasha (Miller & Roorda, 2003). Although not fully

operational yet, a prototype has some unique features to warrant discussion. The model

uses a set of rules to generate schedules. Unlike ALBATROSS, these rules are not

derived from observations, but primarily based on expert decisions and involve

concepts such as priority and flexibility. In addition, an ad hoc fine-tuning algorithm is

used. Activity-travel patterns of household members are generated simultaneously to

allow for possible interaction between members. These joint activities require that the

activity has the same start time, duration and location for each household member

participating in that activity. Thus, a window of opportunity must exist or be created in

the schedules of all of household members taking part in that activity for it to be a

feasible joint activity. The authors acknowledge that several other types of intra-

household interaction exist, but leave that for future research.

This brief characterization of fully operational, comprehensive activity-based models

suggests that at best most of these models have only started to look at household

decision making and household-level activity-travel patterns. In earlier versions and to

46

some extent also in the latest versions, household characteristics have been

incorporated only as explanatory variables in individual-level models. Of course, this is

quite remote from a model of household decision making. Some improvement in

statistical analysis can and should be made by realizing the multi-level nature of this

problem (e.g., Goulias, 2002; Miller, Nurul & Kandker, 2006) but fundamentally this

only implies a marginal adjustment.

Incorporating mechanisms of household decision making should substantially improve

the consistency and interdependencies in activity-travel decisions as an alternative to

the more or less arbitrary breakdown of the multi-faceted decision problem, typical of

the four-step models. However, although the degree of complexity and the

sophistication of the econometric analysis have been substantially enhanced, at a more

fundamental theoretical level considerably less has happened. Separating out the

generation of activities and classifying certain patterns will at best allow us to capture

only some aspects of how households cope with the constraints of their physical and

social environments and organize their activities in time and space in an inherently

dynamic context. In other words, a better understanding of this process and the

underlying mechanisms and determinants is required. Fortunately, the amount of

analytical research and modeling of specific sub-problems has increased rapidly over

the last couple of years. This research will be summarized in the following sections.

3.3 THE NEW VERSION OF ALBATROSS

ALBATROSS has been developed in consecutive phases for couple of years. In every

phase, particular elements of the model system were improved and more than a few

empirical tests were conducted. There have been 4 versions developed so far, and the

newest one, Version 5, is the one that will be developed in this PhD thesis. Version 1

(Arentze and Timmermans, 2000) was based on a limited data set, involving

approximately 3,000 person-days, collected in the South Rotterdam region. This study

was primarily designed to assess the potential of an activity-based approach and the

ALBATROSS framework in particular. The essential nature of the approach however

never changed and a number of key components were developed during this foremost

project. In this version, similar to most activity-based models, fixed (work) activities

were taken as observed and used as anchor point.

The system is acceptable if the main objective is to evaluate the viability of the

modeling approach, however there is a drawback in prediction. Therefore, in Version 2,

the activity skeletons (mandatory activities, such as work, business and other

mandatory activities) were generated. Moreover, the decision rules of the model were

47

re-induced using a national level, pooled data set of approximately 10,000 person-days.

The shift from a regional to a national level also implied that a population synthesizer

was developed. Still in the first application, congestion pricing was developed. Because

such scenarios imply traveler response to external policy, an approach that linked

stated response data to changes in the activity skeleton was developed. It serves as a

general approach to address such problems and use the ALBATROSS system for

policies for which historical data do not exist. These extensions are described in

Arentze and Timmermans (2005).

The application to congestion pricing revealed that rule-based models lack the detail

especially for continuous variables and cannot produce price and travel-time elasticity

of travel-demands satisfactory. Therefore, in Version 3, the principle of what is called

Parametric Action Decision Trees was developed (Arentze and Timmermans, 2007). It

was intended so that ALBATROSS can compete with discrete choice models in this

respect and can represent time and price elasticity and provide utility-based welfare

measures.

The enhancement of the model system went hand-in-hand with several studies and

applications that allow us to better judge the (relative) performance of the model. It

turned out that the activity-scheduling models outperformed the competing model but

the ranking between them is unclear or dependent on the criterion considered. It also

turned out that fair comparisons of completely different models are quite difficult.

They also conducted a spatial transferability study, and found that the model is

sufficiently robust to transfer decision rules derived from one region in the Netherlands

to other regions.

For the time being, the new travel survey in the Netherlands became available (MON).

Although this is not an activity diary, but rather a conventional travel survey, the data

could be transformed into an activity diary format and thus be used as input for

ALBATROSS. The advantages of this dataset compared to a particular purpose activity

diary data collection are clear. The data are collected on a continuous basis, includes a

larger sample of households which covers all of The Netherlands and all seasons of the

year and is less costly (as it is used for multiple purposes). Version 5, which is the

current version, has been developed based on this dataset.

ALBATROSS is one of the few of its generation that incorporate household decision

making aspects, by simulating the decisions of one household member, and then based

on the outcome of this, modeling the decision process of another household member.

ALBATROSS predicts for each household of a studied population the schedule of

activities and trips of each household head for a particular day. However, there are

48

some shortcomings in the model system. Although the approach involves household

decision making, individual schedules are generated. Consequently, the interaction

between persons in multi-person households in particular, was not captured explicitly.

Therefore the goal of this thesis, which is the refinement of ALBATROSS to Version 5,

is to systematically modeling household decision making processes in activity-based

modeling with a special focus on resource and task allocation and joint activity

participation.

3.3.1 Activity-Travel Diary Data

The empirical derivation of the model is based on activity diary data that is derived

from the Dutch National Travel Survey (MON=Mobiliteit Onderzoek Netherlands).

The data used was collected in 2004 covering all of the Netherlands. The survey is

conducted on a regular basis to obtain travel and activity information of residents in the

Netherlands. Although it is a one-day travel diary, the collected data is more complete

regarding activities at trip destinations than its predecessor travel survey called OVG. It

is a household survey where data is collected of all household members on the diary

day as well as general information about household and individual attributes such as,

gender, age, vehicle ownership and driving license ownership, home location,

individual income, occupation, number of working hours per week, etc.

Additionally, respondents are invited to give information about all trips made on a

designated day as also reported. All in all, this survey provides an exclusive data

source to analyze activity-travel behavior of Dutch residents. In the data collection,

29,221 households filled out a one-day travel/activity diary and 28600 of these

households fit the criteria for being considered in ALBATROSS. The data were

transformed to an activity-diary data format for the current estimation purpose.

In ALBATROSS, the classification of activity type is shown in Table 3.1. Ordinarily,

the activity types are grouped into fixed activities and flexible activities. Mandatory

activities are considered fixed activities, while non-mandatory activities are termed

flexible activities. Given the purpose of modeling household decision making in an

activity scheduling process, we distinguish the category of flexible activities as

household task activities and non-household task activities. Mandatory activities

include work, business and other mandatory activities (e.g. school, etc). A household-

task activity refers to an activity that can be allocated to different household members.

A non-household-task activity is a discretionary activity that can be conducted anytime

by any person in the household.

49

TABLE 3.1 Classifications of Activities in a Household in ALBATROSS

No Activity Types Group of

Activity

Person (P)/

Household

(HH) Level

Activity Representation

1 Work P Full-time and part-time

2 Business Mandatory

P Work-related

3 Bring/get person P/HH Drop-off/pick-up children or spouse

4 Shop-1-store P/HH Shopping 1-store

5 Shop-n-store P/HH Shopping multiple stores

6 Service-related

Household-

Task

P/HH Renting movie, getting (fast) food, institutional

purposes (bank, post office, etc)

7 Social P/HH Meeting friends/relatives, religions, social

activities, etc

8 Leisure P/HH Sports, café/bar, eating out, recreational activities

with children, movie, museum, etc

9 Touring

Non-

Household-

Task

P/HH Making a tour by car, bike, or foot (e.g. letting out

the dog)

10 Other Mandatory P Other mandatory activities (school, etc)

TABLE 3.2 Socio-Economic and Situational Attributes used in ALBATROSS

Label Definition Levels

Urb Urban density 0: most densely, 4: least densely

Day Day of the week 0: Monday, 6: Sunday

Comp Household composition 0: single 0 workers, 1: single 1 worker, 2: double 1 worker,

3: double 2 workers, 4: double 0 workers

Child Age of youngest child 0: no children, 1: <6, 2: 6 – 11, 3: 12 – 17 yrs

Age Age category of person 0: <35, 1: 35 – 54, 2: 55 – 64, 3: 65 – 74, 4: 75+ yrs

SEC Socio-economic class (in €) 0: <16,250 (low), 1: 16,251 – 23,750 (low - mid),

2: 23,751 – 38,750 (mid – high), 3: 38,750+ (high)

Ncars # of cars in household 0: no cars, 1: 1 car, 2: 2 or more cars

Driver Person has driving license 0: is not a driver, 1: is a driver

Gend Gender of person 0: female, 1: male

Wstat Work status of person 0: no work, 1: <32 hours week, 2: 32+ hours week

Pwstat Work status of person’s partner 0: no work, 1: <32 hours week, 2: 32+ hours week

50

TABLE 3.3 Accessibility Measures used in ALBATROSS

Label Definition Levels

nEmp1 Daily goods sector: # employees 0: <=115, 1: <=253, 2: <=307, 3: <=507, 4: <=675, 5: >675

within 3.1 km

nEmp2 Non-daily goods sector: # employees 0: <=395, 1: <=635, 2: <=762, 3: <=938, 4: <=2525, 5: >2525

within 4.4 km

nEmp3 All sectors: number of employees 0: <=8785,1:<=12995, 2: <=16120, 3: <=20199, 4: <=70314,

within 4.4 km 5: >70314

SizePop Size of population within 3.1 km 0: <=5050, 1: <=8845, 2: <=13217, 3: <=16833, 4: <=22884, 5: >22884

Dist1 Daily goods sector: distance within 0: <=71, 1: <=127, 2: <=165, 3: <=202, 4: <=346, 5: >346

which 160 employees work

Dist2 Non-daily goods sector: distance 0: <=92, 1: <=145, 2: <=176, 3: <=258, 4: <=334, 5: >334

within which 260 employees work

Dist3 All sectors: distance within which 0: <=92, 1: <=128, 2: <=201, 3: <=274, 4: <=360, 5: >360

4500 employees work

Dist4 Distance within which 5200 people 0: <=0, 1: <=105, 2: <=126, 3: <=163, 4: <=278, 5: >278

live

Household-tasks include the activity types in order of priority: (1) bring/get person, (2)

shopping (one-store), (3) shopping (multiple-stores), and (4) service-related activities.

Non-household-tasks include the following activity types also in order of priority: (1)

social visits, (2) leisure activities (other than touring), and (3) touring (by car, bike or

on foot).

Table 3.2 shows the situational and socio-demographic variables that are used as

prediction variables in ALBATROSS. These variables are the major variables that are

mostly used in the prediction of every choice facet in person-level and household-level

decision making. Obviously, gender is not used as variable in a model that needs joint

decision making. The variables that relate to household-level attributes are urban

density, day of the week, household composition, the presence of young children in the

household, socio-economic class, and car ownership. The remaining variables are

person-level attributes. In addition to that, a set of household-level variables relates to

measures of accessibility of locations given the home location of the household. These

are shown in Table 3.3.

51

3.3.2 The ALBATROSS Process Model and Extension to Include Household

Decision Making

ALBATROSS is an operational system of travel demand modeling underlying activity-

based approach. The model fits into the activity-based approach which is aimed at

predicting which activities are conducted, where, when, for how long, with whom, and

the transport mode involved. Although it does consider the activity-travel interaction

between persons in multi-person households, however, the decision mechanisms are

defined at the individual level. Scheduling steps are made alternately between the

household heads whereby the condition of the schedule after each decision step of one

person is used as condition information in the next decision step of the other person,

and vice versa. As a result, it only has implicit explanation about travel party as well as

joint activity. Additionally, car allocation between household members, in a case of

car-deficient household, is imperfectly treated. Furthermore, activity allocation, in

terms of household task activities, and joint activity participation are not addressed as a

household decision. Taken as a whole, the existing ALBATROSS does not cover the

household decision making explicitly. We therefore intend to accomplish the

shortcomings by improving some aspects in household decision making more

explicitly.

The recent ALBATROSS consists of two major components that together define a

schedule for each individual and each day. The first component generates an activity

skeleton consisting of fixed activities and their exact start-time and duration. The

second component determines the part of the schedule relating to flexible activities that

are conducted that day, their travel party, duration, time-of-day and travel

characteristics. Both components use the same location model. All components assume

a sequential decision process in which pre-defined rules operate to define choice sets

and implement choices in the current schedule.

In order to better capture within-household interactions, we identify the facets of

activity-travel behavior of the two household heads that require household decision

making and expand the household activity-travel scheduling process regarding each

component. The three major components are expanded to be five major components as

can be seen in Figure 3.1. In addition, in some parts of generation modules,

supplementary choice facets covering joint decision making are inserted. The five

major components and each component consisting information of the preceding

component, activity-level and schedule-level information as well as individual and

household attributes.

Figure 3.2 is a description of the generation module in Figure 3.1. It consists of three

components, i.e., generating the work activity, generating the work-related activity, and

52

generating the non-work activity. Each component is considered as person-level

decision making when the activity is conducted independently, and otherwise as

household-level decision making when the activity is conducted jointly. Initially, it

begins from the generation of work activities consisting of maximally two work

episodes of each person (male and female) in the same household. In this particular

component, the model is done exclusively from other models, considering a priority-

based activity scheduling process in that work activity lies in the top most priority in

the hierarchy. Both decisions that are done either at person-level or at household-level

are taken into account in this regard. At the person-level, it involves the choice of

number of work episodes, start-time and duration, and the location of each work

episode (tour). The decision of allocating the car to the work tour, in particular in car-

deficient households, on the other hand, lies at the household-level. Having allocated

the car, if necessary, the household heads face the decision of choosing the transport

mode to the work place which lies at the person-level.

The next stage of the first component defines the cohort of work-related (business and

other mandatory) activities starting from the generation activity concerned and its

number of episodes. Further, the duration and start-time of work-related activities are

determined. Given that work-related activities are sometimes linked to work activities,

the decision on whether or not there is link to work activities is defined afterward. The

last part in this stage is defining location choice. All facets in this step are determined

at the person-level.

The final stage in the first component deals with the facet of non-work activities.

Selecting the non-work activity is performed initially; yes or no a particular activity is

conducted by the two household heads, each for independent and joint participation

cases. In case of household task activities, it is followed by the allocation decision in

the subsequent step, whether a particular activity is done by the male, female, or both.

Further, the duration and start-time of each independent and joint activity is determined.

Having defined the first component, trip-chaining choices are made, as shown in

Figure 3.1. It defines the duration and start time of the activity concerned both at the

household-level and person-level. Further, the non-work tour is accomplished that

consists of car allocation and transport mode. The car allocation decision to non-work

tour is done at the household-level while the transport mode choice to non-work tour is

done at the person-level. It is essential to note that in all decision tree models, the

results of earlier decisions are used as condition variables for each next decision and

the process results is a complete schedule of each person.

53

Transport Mode to Non-Work Tours

Car Allocation to Non-Work Tours

Location choice

Trip-chaining choice

Generation modules

Generating Work Activity

• Person-Level : # episodes, Duration, Start time, Location

• Household-Level : Car allocation to work tour

• Person-Level : Transport mode to work tour

Generating Business & Other Mandatory Activity

• Person-Level : # episodes, Duration, Start time, Link-Work, Location

Generating Task Activities and Non-Task Activities

• Household-Level:

- Activity selection of joint activity categories

- Activity allocation (for task activities)

- Duration and Start time

• Person-Level:

- Activity selection of independent activities

- Duration and Start time

FIGURE 3.2 Generation Modules in ALBATROSS

Integration

FIGURE 3.1 Main Steps in the Scheduling Process of Current

ALBATROSS

54

# episodes J

No Yes

Duration ratio j = 1,2

Start time of episode j = 1

l = 1

Car allocation to Work tour l

l = l + 1

l < L

l = L

STOP

k = 1

Transport mode to Work tour k

k = k + 1

k < K

k = K

STOP

START

START

STOP

1

2

3

5

6

Include work

Work duration

J = 2

Duration of break

J = 1

STOP

4

Go to Location to work place module in Figure 4

14

15

Include business and other mandatory act i

# episode J

Duration of ep. j act. i

Position of j

Link ep. j to work

Start time ep. j act i

Yes

Yes

No

No

i = 1 No i = i + 1

j = j + 1

j < J j = J i = I

START

STOP

STOP

j = J, i < I

16

17

18

19a

19b

START

FIGURE 3.3 The Process Model for Mandatory Activities

Modified after: Arentze and Timmermans (2005)

20

55

START

Relative location of

episode i, j

i = i + 1

21

STOP

i = 1

Size by distance band

of episode i, j

22

Select location from band

OTHER

Same as PREVIOUS

Same as NEXT

j = 1 j = j + 1

i < I, j = J

i = I, j = J j < J

j < J

Same as previous location

In home municipality

Order of municipality

Distance band municipality

Order of zone in municipality

Distance band of zone in mun.

Nearest mun. of chosen order

i = 1

j = 1

j = j + 1

i = i + 1

j < J

i < I, j = J

i = I, j = J

i = I, j = J i < I, j = J

j < J

Yes

Yes

Yes

No

No

No

7

8

9

10

11 12

13

Go to Car allocation to work tour module in #14

START

STOP

STOP

Go to Car allocation to work tour module in #14

FIGURE 3.4 The Process Model for Predicting Locations of Work Activities

Source: Arentze and Timmermans (2005)

FIGURE 3.5 The Process Model for Predicting Locations of

Work-Related and Non-Work Activities

56

START

i = 1

j = 1

Trip-chain of

episode i, j

j = J, i = I

j < J

j = j + 1

i = i + 1

Car allocation to

Non-Work tour l

l < L

l = L

STOP

k = 1

Transport mode to

Non-Work tour k

k = ki + 1

k<K k=K

START

STOP

START

29

34

35

STOP

l = 1

l = l + 1

No Include non-work activities

Independent/Joint

Duration Joint

Start time Joint

If Joint, m = 1

Duration Independent

Start time Independent

START

No

If Independent, n = 1

Yes, i = 1

STOP Next activity

Yes, if task activity

Yes, if non-task activity

23

24

25

26

27

28

Go to Figure 5

Include a next episode of the current activity

Yes/No

j = 1 i = i + 1

j = j + 1 j = j + 1

i = i + 1

i < I, j < J, m = M

i < I, j = J, m = M

i = I

m = m + 1 If m = M, n = 1 n = n + 1

FIGURE 3.6 The Process Model for Non-Work Activities

57

After completing the whole process from generation module to transport mode to non-

work tour model, the process is finished. Nevertheless, in order to define the accuracy

of model, we compare the integration model that simulates the comprehensive model

between old version and new version. It is expected that the prediction result of new

version is higher than the old one.

Figures 3.1 – 3.6 schematically present the structure of each of the main components of

the model in a more detail. Each numbered rectangle corresponds to a decision tree to

be derived from activity diary data. The indices used in the figures are defined as:

i index of activity in order of priority, i = 1…..I j index of episode of activity I in order of start time, j = 1…J

k index of tour in order of start time, k = 1…K (person-level)

l index of tour in order of start time, l = 1….L (household-level) m index of joint activity in order of priority, m = 1….M

The mandatory component comprises decisions 1-9, where the work activity deals with

decision 1-6 and the work-related activity cope with decision 7-9. Non-work activity

that consists of household-task and non-household-task activity copes with decision

10-18. The activity scheduling process is done sequentially from the first component to

the subsequent component. All components in mandatory activity is done at the

individual-level decision making, only car allocation decision to work tour deals with

joint decision making. Given our purpose on household decision making, the

component of non-work activities is not only dealing with person decision making, but

also with household decision making. The location decision for work tour is done

separately with non-work tour.

In the diary data used for estimation, a joint (non-work) activity in a household is

identified as a particular non-work activity that occurs in the diary of both the male and

female head and takes place at the same duration (+/- 15 minutes). Joint activities have

priority over independent activities and hence are scheduled first.

3.3.2.1 The Mandatory Activity Module

As an important feature, the mandatory model component (Figure 3.3) determines

activity patterns that consist of several sub processes including:

1. Determining the pattern of work activity

2. Determining the location of work activity

3. Determining the car allocation decision to work tour

58

4. Determining the transport mode to work tour

5. Determining the pattern of work-related and other fixed activities

6. Determining the location of work-related activity

The work activity has maximally two episodes. The pattern is defined by decisions

about the number of episodes, duration and start-time. The duration and start time is

done as a continuous variable. The location component developed a location choice

model in which location choice decisions are made in a priority order of activities and

within activities in the order in which the activities occur in the schedule. To better

understand activity choice location in the context of a complete activity schedule for a

day, the concept of detour time is applied. This concept is used in ALBATROSS

(Arentze and Timmermans, 2007). Different from any other concept that consider the

travel distance from home or non-home to a particular location, detour time considers

relative location to the previous and next activity. The detour time of a candidate

location for an activity is defined as the extra travel time required that implement the

activity in the context of the current activity schedule. This concept is very useful to

build trip chains and to simulate the emergence of feasible activity-travel schedules

that take space-time constraints into account.

In case of work or work-related (business and other mandatory) activities, the non-

work activities are not yet scheduled and the location of the next (work or work-

related) activity, if it is other than home, is still unknown. In such cases the model

assumes that the next location is the home location. On the other hand, in case of a

non-work activity, the location of the next activity is unknown if it is a non-work

activity of a same or lower priority for the same reason. Again, in such cases the model

assumes that the next location is the home, work location or the location of a higher

priority activity (what comes first in the schedule). Although these assumptions are

simplifications of reality, it is to be expected that they will not seriously affect the

performance of the model. At least, the model is able to take into account the location

relative to home and to a previous/next location in every location decision of a

sequential priority-based scheduling process and consequently it should better cover

interdependencies in these choices. A space-time prism is calculated for each location

decision defining the set of locations that are within reach given the space-time

constraints imposed by the interaction between the environment and the schedule.

Having identified the origin and destination for the activity considered the model

determines the locations, based on postcode areas, which are within reach, i.e. within

the prism. Since transport mode is unknown yet, the model calculates a preliminary

prism based on the fastest transport mode available in the time slot under concern. For

instance, if the person is a driver and the household owns car(s), and the car is not used

59

for a work activity of another household member in the same time slot, then the fastest

travel mode is the car. In case there is no car in the household, the fastest travel mode

is public transport in most cases.

Having identified the fastest transport mode, the shortest travel time across the road

network is determined. Furthermore, the (minimum) duration of the activity, the time

window and opening hours of required facilities at destination, are taken into

consideration. Time window is defined by the earliest possible departure time from the

origin and the latest possible arrival time at the destination. All in all, the resulting set

of locations meet an exhaustive set of space-time and resource availability constraints.

Note that, the location choice for work activity is done differently from other activities.

This conceptualization is similar to the current version of ALBATROSS. However, the

existing model only applies to the context of person-level decision making

(independent activities) while in this study we expand it to cover as well household-

level decision making (joint activities).

3.3.2.2 The Non-Work Activity Module

Figure 4 represents the last part of the models that deals with:

1. Determining the pattern of non-work activity

2. Determining the trip-chaining

3. Determining the location of work-related and non-work activities

4. Determining the car allocation decision to non-work tour

5. Determining the transport mode to non-work tour

Joint decision making takes part in almost component in this module, unless in trip-

chaining and transport mode. The pattern of non-work activity is classified by

decisions about the activity selection, which activity to be done in the context of

household-task and non-household-task activities. In case of household-task activities,

the process continues to a decision of task allocation on who should perform the

household task, either the male, female, or both male-female. Further, participation in

activity is taken into account. If the activity is done alone it is phrased as independent activity and other wise as joint activity when the activity is done together. After

defining those components, the duration and start time is done as a continuous variable.

Further, trip-chaining choice is taken into account. This process is done exactly the

same as in a previous version. As done in mandatory activity module, the location

decision is taken into account in the subsequent component. Note that, in case of

independent activity, the work-related activities (i.e. business and other mandatory) are

included. The process of doing it is also similar with what is done in work activity.

60

However, the special conditions are taken into account for joint participation. For more

detail, please refer to chapter 6 of this thesis.

Car allocation decisions are considered as an element of a more encompassing activity

scheduling process. A car allocation model is applied for two-heads households with

one car, or so called car-deficient households. If one car is available in the household

and both household heads are drivers, then the decision which person is going to use

the car involves a household-level decision. For instance, in case of work tour, if the

two persons undertake a work activity during the same time slot, a decision needs to be

made who can use the car for the trip to work. A car-allocation decision is needed not

only if the two heads in a household both have a work activity. Also, if only one of

them performs a work activity, it is still necessary to identify whether the worker uses

the car or not. The model also includes the option that none of the household heads

uses the car, but some other means of transport. Hence, the decision options are male,

female, or none. The process of allocating the car to the two-heads household is

basically the same for work tour and non-work tour. Note that, work tour is defined

when at least one work episode includes in a trip-chaining. On the contrary, non-work

tour is identified when work episodes is not appeared on a trip-chaining of male or

female. The work-related activity is also accounted for non-work tour. The last part of

this module, which is also the last component in the modeling, is transport mode to

non-work tour. This is also done similar to what is done on work tour and based on the

underlying concept of car allocation decision. The transport mode has 4 choices: car driver, car passenger, public transport, and slow mode (bike and walk).

3.4 DERIVATION OF DECISIONS FROM DECISION TREE

Every decision step in the process model is managed by decision tree, as represented

by a numbered rectangle in Figure 3.3 to Figure 3.6. Each decision tree is derived from

corresponding observations in the activity diary data set using a CHAID based

induction method. This section considers the decision tree induction method used to

determine decisions in the prediction stage, as explained in Arentze and Timmermans

(2005). Discrete and continuous choices are separately discussed.

3.4.1 Discrete Choices

The different levels at which decisions are to be made include the schedule, tour and

activity level. Accordingly, the definition of a case differs between decision trees. As

for example, the abstract illustration is assumed that at the given moment in the

decision process, a decision is derived for N cases. A decision tree defines a

classification function.

61

Pr (k | Xj) = f(Xj) [3.1]

where k is an index of leaf nodes of the tree and Xj is a vector of attribute levels for

given case j. Since the type of decision trees used here is crisp (as opposed to fuzzy

trees) and deterministic (as opposed to co-evolutionary trees), the probability of

assigning case j to node k is one or zero. The action-assignment rule comes into

operation after [3.1] and determines:

Pr (i | k) = f(qk , δj) [3.2]

where i is an index of discrete choice alternatives considered in the given decision tree,

qk is the choice probability distribution across alternatives at the k-th node and δj is a

zero-one vector indicating the availability of each choice alternative in case j. Note that,

where qk is a characteristic of the decision tree, δj is to determined for each case in the

prediction stage. The probability of selecting alternative i in case j is:

Prj (i) = )|Pr()|(Pr kiXkk j∑ [3.3]

Further, the probabilistic action-assignment rule f(qk , δj) used in ALBATROSS is

specified. To simplify notations, the subset of cases assigned to leaf node k is

considered and the subscript k in the symbols is dropped. The rule can be written as:

=

∑i iij

iijij

q

qp

δδ ji,∀ [3.4]

where pij is the probability of selecting choice alternative i in case j (at leaf node k), δij

is a zero-one variable indicating the availability of i in case j, and qi is the choice

probability of alternative i dictated by the decision tree (at leaf node k) and estimated

on the training set. As implied by this equation, probability pij is zero if i is not

available and equals the second term on the RHS of the equation otherwise.

3.4.2 Continuous Choices

In the process model, continuous decision trees describe duration and start time choices.

Rather than a choice probability distribution across discrete choice alternatives, these

trees describe a specific distribution of the continuous duration or start time variable at

each leaf node (Arentze and Timmermans, 2005). Thus, the continuous action-

assignment equivalent of equation [3.2] defines the function:

62

Pr (y | k) = f(Rk , Bj) y = 0, 1, 2, ..., 1440 [3.5]

where Pr (y | k) is the probability of selecting value y at leaf node k, Rk is a vector of

parameters defining the distribution at leaf node k and Bj is a set of tuples (b1, b2)

defining unavailable or blocked ranges [b1, b2)] on dimension y in case j due to

temporal constraints. Since times are measured in minutes and the schedule has a fixed

time window (of 24 hours), y has a pre-defined minimum and maximum. Furthermore,

we assume natural numbers for y.

Continuous decision tree used in ALBATROSS define distributions at each leaf node

in terms of m – 1 cut-off points and the minimum and maximum of the range. The cut-

off points divide the range into m intervals in such a way that an equal number of

training cases at the leaf nodes is observed in each interval. As a consequence of this

method, Rk specifies m+1 parameters. The number of elements of set Bj in a specific

case is zero if the complete range is available and bigger than zero if parts of the range

are blocked by constraints.

To define the probabilistic continuous action-assignment rule used in ALBATROSS,

we the following symbols are used. Let Pj (y) denote the probability of selecting y = 1,

…, 1440 in case j, m denote the number of equal-frequency intervals used in

continuous decision trees, di represent the width of equal-frequency intervals used in

continuous decision trees, di represent the width of equal frequency interval i, bij be the

width of the blocked part of equal frequency interval i in case j defined by the

combination of Rk and Bj and Pj (y) =1, if y falls in the unblocked part of the interval i and 0 otherwise.

∑=ij iyiyP )|Pr()Pr()( j∀ [3.6]

where Pr (i) is the probability of selecting EFI (equal frequency interval) i and Pr (y | i) is the probability of selecting y given i. Pr (i) is defined as:

j

i

ijiC

d

bd

mi

−=

1)Pr( i∀ [3.7]

The first term represents the a-priori probability of selecting i. Because EFIs represent

an equal number of cases, an equal number of cases, an equal probability is assumed

for all m EFIs. The second and third terms define a correction this equal probability.

The first correction is equal to the proportion of the available range in the EFI i and the

third factor makes sure that probabilities sum up to one across EFIs.

63

Similar as in the discrete case, we should include availability variables as potential

predictor variables in inducing the tree to reduce the bias to the extent possible.

3.4.3 Goodness-of-Fit Measures

In order to measure the performance of the decision tree, different goodness-of-fit

measures are used for discrete and continuous choice (Arentze and Timmermans,

2005).

3.4.3.1 Discrete Choices

There are two alternative goodness-of-fit measures for discrete choice decision trees.

First, the so called likelihood or probabilistic theta is conceptualized as:

)|(Pr)|Pr()Pr( ' kikikjek i∑ ∑>−= [3.8]

where e is the probability of correctly predicting the choice for any given case j in the

same sample, Pr(j -> k) is the probability that j belongs to leaf node k, Pr(i | k) is the

probability that choice i is observed in cases belonging to leaf node k and Pr’(i | k) is

the probability of predicting I in those cases. The probabilities on the RHS of [3.8] can

be found as:

n

fkj k=>− )Pr( [3.9]

k

ik

f

fkiki == )|(Pr')|Pr( [3.10]

where n is the total number of cases, fk is the number of cases at leaf node k and fik is

the number of cases at leaf node k with observed choice i. The n and f variables all

refer to the sample from which the tree was derived (i.e. the training set) so that the

probabilities Pr(j -> k) and Pr(i | k) are to be interpreted as estimates of true

probabilities for any sample of unseen cases.

The predicted and observed probabilities in [3.8] are the same due to the probabilistic

action-assignment rule used. Substituting [3.9] and [3.10] in [3.8] gives:

2

∑∑

=

ik

ik

k

k

f

f

n

fe [3.11]

and rewriting results in:

64

∑∑

=k

k

i ik

f

f

ne

2)(1 [3.12]

It should be noted that this measure assumes bias-free predictions. In reality, the

action-assignment rule takes the availability of choice alternatives into account and

therefore is more complicated than [3.10]. The actual rule is given by [3.4].

By comparing e with a null model, we can derive a measure of relative performance.

We consider as the null-model a decision tree consisting of the root node only. Then,

the likelihood or probabilistic theta for the null model can be found as:

∑=i

iie )(Pr')Pr(0 [3.13]

Or ∑∑ =

=

i ii

i fnn

fe 2

2

2

)(1

0 [3.14]

where fi is the overall frequency of choice i in the sample.

The quotient:

0

0

1 e

eeeincr

−

−= [3.15]

then indicates that the increase in likelihood as a ratio of the maximum increase that is

possible given the null model. Note that this indicator is comparable to the (log)

likelihood ratio commonly used as a measure of goodness-of-fit for conventional

discrete choice models.

The second measure is directly derived from the Chi-square statistic used as split

criterion. The tree as a whole defines as I x K frequency table, where I is the number of

choice alternatives and K is the number of leaf nodes. The Chi-square of this table can

be taken as a measure of dependence between tree condition (leaf nodes) and choice.

Because the value of Chi-square is dependent on sample size n, we use a

standardization to obtain a measure on a (approximate) 0-1 scale, known as the

contingency coefficient and defined as:

n

c+

=2

2

χ

χ [3.16]

65

where 2χ is Chi-square of the I x K frequency table and n is sample size. A zero value

indicates a zero association between condition and choice and a value of one a

maximum dependence. So, c can be interpreted as the discrete equivalent of the linear

correlation coefficient between predictions and observations.

3.4.3.2 Continuous Choices

For continuous choice decision trees only one goodness-of-fit measure is available.

This measure is directly derived from the F-statistic used as a split criterion. For the

tree as a whole, the F-statistic is calculated as:

Kn

sn

K

MmnF k kkk kk

−−

−=

∑∑ 22 )(/

1

)( [3.17]

where n is sample size, nk is number of cases at leaf node k, K is the number of leaf

nodes, mk and sk are the mean and standard deviation of the distribution at leaf node k

and M is the overall sample mean. Thus, F represents the ratio between between-group

and within-group variance. The higher the value, the stronger the dependence between

condition and choice is.

3.5 CONCLUSIONS AND DISCUSSION

This chapter described several studies applications in household decision making. The

comprehensive operational system of activity-based travel demand modeling is still

few exist until recently. Therefore, the development of ALBATROSS could enhance

the studies of activity-scheduling models in activity-based transport demand modeling.

The framework of ALBATROSS in the context of person-level and household-level

decision making is also explained. Using the new travel survey in the Netherlands

(MON) data, the data is transformed into an activity diary format and used as input for

ALBATROSS. There are some advantages of using this dataset in terms of a particular

purpose of activity diary data collection. The collected data were done on a continuous

basis, includes a larger sample of households that covers all regions in The Netherlands.

Incorporating household decision-making into activity-based models of transport

demand receives increasing attention in recent times. Problems such as activity

allocation, joint activity participation and resource allocation have in the past typically

been addressed separately, i.e. not in the context of a scheduling process or at most ad

hoc as part of a more comprehensive activity-based model of transport demand.

66

Decisions to travel jointly in multi-person households require joint decision between

persons.

ALBATROSS concerns about joint decision making matter. The out-of-home activity

needs synchronization between persons in particular in multi-person households. By

incorporating car allocation decision either to work tour or non-work tour, the model

gives such an improvement in a recent mode choice model. Those households that are

dealing with car allocation decision process (car-deficient households) have significant

influence to the choice of transport mode, in case of one of male/female get the car to a

particular location.

REFERENCES



Arentze, T.A. and Timmermans, H.J.P. (2003), “Measuring Impacts of Condition

Variables in Rule-Based Models of Space-Time Choice Behavior: Method and

Empirical Illustration”. Geographical Analysis, 35, 24-45.

Arentze, T.A. and Timmermans, H.J.P. (2004), “A Learning-based Transportation

Oriented Simulation System”. Transportation Research Part B, 38, pp.613-633.



Gliebe, J.P. and Koppelman, F.S. (2002). ”A Model of Joint Activity Participation

between Household Members”. Transportation, 29, pp.49-72.

Gliebe, J.P. and Koppelman, F.S. (2005), “Modeling Household Activity-Travel

Interactions as Parallel Constrained Choices”. Transportation, 32, pp.449-471.

Goulias, K.G. (2000, “Companionship and Altruism in Daily Activity Time Allocation

and Travel by Men and Women in the Same Households”. In Proceeding of TRB 200, Washington, D.C., US.

Miller, E.J., and Roorda, M.J. (2003), “A Prototype Model of Household

Activity/Travel Scheduling”. Proceedings of the 2003 Transportation Research Board, Washington DC, USA.

Scott, D., and Kanaroglou, P. (2002), “An Activity-Episode Generation Model that

Captures Interaction between Household Heads: Development and Empirical

Analysis”. Transportation Research B, 36B: 875-896.

67

Srinivasan, S. and Bhat, C. (2004), “Modeling the Generation and Allocation of

Shopping Activities in a Household”. Paper presented at the 83rd

Annual Meeting of

the Transportation Research Board, Washington, DC.

Zhang, J., Timmermans, H.J.P., and Borgers, A. (2005), “A Model of Household Task

Allocation and Time Use”. Transportation Research Part B, 39, 81-95.

68

Chapter 4

CAR ALLOCATION BETWEEN HOUSEHOLD

HEADS IN CAR-DEFICIENT HOUSEHOLDS: A

DECISION MODEL

Anggraini, R., Arentze, T.A., and Timmermans, H.J.P., 2008. European Journal of Transportation Infrastructure and Research, 8(4), pp.301-319.

ABSTRACT This paper considers car allocation choice behavior in car-deficient households explicitly in the context of an activity-scheduling process, focusing on work activities. A decision tree induction method is applied to derive a decision tree for the car allocation decision in automobile deficient households using a large travel-and-activity diary data set recently collected in the Netherlands. The results show a satisfactory improvement in goodness of fit of the decision tree model compared to a null model. Overall, the probability of males getting the car for work is considerably higher than that of female in many condition settings. However, activity schedule, spatial and socio-economic variables appear to have an influence as well. An analysis of impacts of condition variables on car allocation decisions reveals that socio-economic variables have only a limited impact, whereas attributes of the transportation and land-use system have a relatively big impact. The propensity of men driving a car to the work place is higher than that of women. However, the relative accessibility of the work location by bike compared to car appears to have a relatively large influence on who gets the car for work. Household income and presence of children also appear to have significant effects.

69

4.1 INTRODUCTION

One of the major indirect factors contributing to increasing traffic congestion in urban

areas and highways is the increase of household automobile ownership. The vast

majority of households own at least one car and an increasing number of households

own more than one car. It is of no surprise therefore that car ownership and vehicle

fleet choice is one of the areas in transportation research that has received much

attention. A complementary active area of research focuses on transport mode choice

analysis and modeling to shed light on preferences of individuals in choosing one

option among several modes available for the trips they make (Xie, et al., 2003; Miller,

et al., 2005). Despite the substantial amount of research on car ownership in general, the specific

question of who is getting the car for which activities in car-deficient households has

received much less attention. In this context, car-deficient households are households

where the number of drivers exceeds the number of cars. Consequently, we know

relatively little about the factors that play a role in this decision and about the decision

process by which household members arrive at a choice on who should use the car

(Hunt and Petersen, 2004; Vovsha and Petersen, 2007). A model of binary car-

allocation choice (to use car or not) made by the household members for each tour in

an integrated framework of intra-household car-use preferences has been proposed and

estimated by Petersen and Vovsha (2005). They clearly showed that car-allocation

decisions are inter-related with mode choice, joint travel arrangements, and schedule

adjustments.

Yet, the outcome of this decision does not only have a direct impact on transport mode

choice, but also has potentially important ramifications for activity-travel schedules of

individual household members. Action spaces allowed by different transport modes

vary substantially and therefore the generation, location and timing of activities and the

organization of trips into tours depends strongly on the transport mode. Critical

questions in better understanding this decision process include: how do households

make trade-offs between mobility needs of drivers and are there differences between

households related to socio-economic and situational variables? Current travel demand

models have paid little attention to address these car allocation decisions.

The decision which person will use the car is a complex decision in car-deficient

households in the sense that many factors may influence this decision. For example,

gender roles may imply that males are more likely to use the car than women are.

However, it may also be that in case the male is going to work for a long period of time

in a day, while the female has many errands to complete, the flexibility of scheduling

and rescheduling activities made possible by the car, may lead the household to decide

70

that the female will use the car. As argued by Bianco and Lawson (1996), women are

more dependent on the car than men because of their traditional responsibilities related

to childcare and household maintenance as well as their concern for safety. On the

other hand, due to a good provision of public transport and more dense cities, in

Western European countries, we often see that women who do not participate in the

labor force tend to use public transport or use slow modes. Apart from socio-

demographic variables, the relative accessibility of locations for activities by car will

have an influence.

Surprisingly, car allocation decisions have also not received much interest in the

activity-based (micro-simulation) modeling literature. To date, fully operational

activity-based micro-simulation systems include ALBATROSS (Arentze and

Timmermans, 2000, 2004, 2005), TASHA (Miller and Roorda, 2003), Florida’s

Activity Mobility Simulator (FAMOS) (Pendyala, 2004), based on the Activity-

Mobility Simulator (AMOS) (Kitamura et al., 1996), and the Prism Constrained

Activity-Travel Simulator (PCATS) (Kitamura and Fujii, 1998), and CEMDAP (Bhat

et al., 2004), and some projects that have been implemented in the US (Bowman, J.L,

2008; Vovsha, P., 2008). One of the reasons for developing activity-based models was

that typical response patterns to transport demand management involved household

decisions. Such responses could not be captured by trip-based models, at least not

explicitly, as they were founded on individual as opposed to household behavior. In

general, only few of the existing operational activity-based models are based on

household decisions, and this statement also applies to the car allocation decision.

In this study, we examine this emerging issue. The study focuses on households which

have fewer cars than drivers. Car allocation decisions are considered as an element of a

more encompassing activity scheduling process. A large number of factors that

potentially influence car allocation decisions in car-deficient households are considered.

These factors relate to variables of the activity schedule and space-time setting as well

as individual and household characteristics. In this study, we use ALBATROSS as a

framework to investigate the car allocation decisions as part of an activity scheduling

process. ALBATROSS is an operational activity-based model developed for the Dutch

Ministry of Transportation, Public Works and Water Management for travel demand

analysis. More specifically, the paper will report the conceptualization of the problem

and present empirical results of a car allocation model for two household heads. It is

assumed that before the car is allocated, participation in activities both at the household

and person level is known. If one car is available in the household and both household

heads are drivers, then the decision which person is going to use the car involves a

household-level decision. For instance, if the two persons undertake a work activity

during the same time slot, a decision needs to be made who can use the car for the trip

71

to work. Note that the outcome could also be that both will use another transport mode

for the work commute.

The paper is structured as follows. First, the next section briefly explains the

ALBATROSS scheduling process model that provides the framework for the car

allocation model. The sections that follow describe the data used for the analysis and

the proposed car allocation model. After this section, the results of empirical analyses

will be considered focusing on some descriptive statistics and the empirical derivation

of the model. The paper is concluded by drawing conclusions and discussing some

possibilities for future research.

4.2 ALBATROSS PROCESS MODEL

ALBATROSS stands for A Learning Based Transportation Oriented Simulation

System. The model considers household and personal activities and travel performed

on a particular day and generates a schedule for each household head. The model takes

into account the presence of children as an independent variable, but their activities are

not explicitly represented. Work activities are presumably primary fixed activities,

whereas several household activities and work-related activities, such as bring/get

person, business, and others are assumed as secondary fixed activities. Shopping,

social and leisure activities are called flexible activities. It should be noted that fixed

activities are also predicted. ALBATROSS consists of four major components that together define a schedule for

each household head for a certain day as displayed in Figure 4.1. It should be noted

that this describes the computational process model underlying the system merely in

main lines. The first component generates a work activity pattern consisting of one or

two work episodes, if any, and the start time, duration and location of each episode. It

also predicts the transport mode(s) used to travel to the work location(s). The second

component determines the part of the schedule related to secondary fixed activities

(bring/get person, business, and others). It determines which types of these activities

are conducted that day and how many episodes and for each episode the start time,

duration and location. Furthermore, it also determines whether particular trip-linkages

are made with the work activity, if any. The following component considers the scheduling of flexible activities. Almost

similar to the previous component, it predicts activity types, number of episodes of

each activity type and the start time, duration and location of each episode. The

sequence of activities and possible trip-chaining links between activities are also

determined in this stage.

72

Generating Work Activity

- number of episodes

- start time

- duration of each episode

- location of each episode

- transport mode to the work activity

START

Generating Secondary Fixed Activities

- which type of activity

(bring/get,business,other)

- how many episodes of each activity

- start time


- linkage to work activity


Generating Flexible Activities

- which type of activity

(shopping,service,leisure,social,touring)

- how many episodes of each activity

- start time


- trip-chaining for all activities


Transport Mode for Each Non-Work

Tour

STOP

FIGURE 4.1 Schematic Representations of Main Steps of the

ALBATROSS Process Model

73

The latter decisions relate to all activities in the schedule, not just the flexible activities.

Finally, the last component predicts the transport mode used for each tour (except for

tours that include a work activity; for the latter tours the transport mode is known as

the outcome of a higher-level decision). The car allocation model developed in this study predicts who of the two household

heads in car deficient households uses the car for a particular activity. As a case, we

focus here on the work activity given that this activity usually is mandatory, conducted

by one spouse individually (as opposed to jointly), tends to occupy a large part of the

day and may serve as a second base location for other activities besides the home

location.

We emphasize, however, that car-allocation decisions are not confined to the work

activity. In the last step of the ALBATROSS scheduling process (Figure 4.1), the trips

required for non-work activities and the way they are organized into tours are known.

In that stage, a mode choice is made for each non-work tour (chain of trips including

one or more activities). These choices are preceded by a car allocation decision as well.

Although we focus here on the work activity, the same methodology developed here is

used to model car-allocation decisions involved for non-work tours. A car-allocation

decision restricts a subsequent mode choice: if the car is allocated to an activity or tour

no further decision is needed and if the car is not allocated, then a choice is confined to

other modes then the car. Note that car sharing is still open as an option if the car has

not been allocated to an activity or a tour. In ALBATROSS, car sharing is represented

as a car-passenger option. In other words, the car allocation decision has implications

for the possibility of choosing the car-driver mode only, but leaves open the car-

passenger mode.

Because ALBATROSS uses a sequential decision process, to generate a schedule for

each household head, the information available for the car allocation model is limited.

At the moment in the process when the car allocation model generates decisions, the

schedules of the household heads regarding the work activity are known; the schedules

regarding other activities then are still unknown. This does not mean, however, that the

decisions cannot take requirements of other activities (which are scheduled in a later

stage) into account. An outcome of the decision may well be that the car is not used for

a work activity considering the household’s needs for other activities. For example,

presence of children is a condition variable the system can use to anticipate a possible

escorting activity for which a car is needed and, hence, may inform the system not to

allocate the car to a work activity (of one of the two partners). Due to the complexity of

the scheduling problem it is inevitable that the decisions are made in a particular

sequence.

74

4.3 DATA

The data used in this study originates from the so-called MON survey (Mobiliteit Onderzoek Nederlands – Mobility Research Netherlands) held in 2004. The MON

survey is conducted on a regular basis to obtain travel and activity information of

residents in the Netherlands, and although it primarily uses a trip-diary it includes

detailed data on activities (at destinations) as well. More specifically, it is a one-day

travel diary of a sample of households that contains information about each household

member. In addition, individual and household socio-demographics such as age,

household composition, education level, income level, vehicle availability, residential

location, and information about all trips made within 24 hours as well as out-of-home

activities at destinations of trips are collected. For each trip, respondents are asked to

report information about several attributes including type and duration of the activity at

the destination, departure time and arrival time, trip purpose, transport mode, and

origin and destination location. Furthermore, trip-chains can be identified. All in all,

this information provides a suitable source to analyze activity-travel behavior of Dutch

residents because activity and travel information are both revealed. In this data

collection, 29,221 households filled out a one-day travel/activity diary and 28,600 of

these households fit the criteria for being considered here (forms of group housing,

such as for example student housing, are excluded). The data were transformed to an

activity-diary data format for the present estimation purpose.

4.4 CAR ALLOCATION MODEL SPECIFICATION

As said, the car allocation model focuses on car deficient households (i.e., more drivers

than cars present) and a joint decision between the two heads (mostly, a female and

male). The total sample extracted from the MON data includes 28600 households.

Given the purpose of this study, only the following households and days are relevant:

(1) there are two heads in the household; (2) there is one car in the household; (3) both

heads are drivers and (4) at least one of the heads has a work activity on the day

considered. As it appears, 3,523 households (and days) fit these criteria.

The car allocation decision model is schematically shown in Figure 4.2. A car-

allocation decision is needed not only if the two heads in a household both have a work

activity. Also, if only one of them performs a work activity, it is still necessary to

identify whether the worker uses the car or not. Furthermore, the model includes the

option that none of the household heads uses the car, but some other means of transport.

Hence, the decision options are the male, female, and none.

75

TABLE 4.1 Defining Car Allocation Decisions in Households

No. Number of male’s

work episodes

Number of female’s

work episodes

Cases Number

of Cases

Number of car

allocation decisions

1 0 1 520

2 1 0 A

1437 1

3 0 2 132

4 2 0 B

520 2

5 1 1 C 1047 1 or 2

6 1 2 144

7 2 1 D

228 1, 2, or 3

8 2 2 E 68 1, 2, 3, or 4

Total Sample 4096

START

STOP Work is in

HH schedule

Both heads have

work activity

Car Allocation

cases: C, D & E

Car Allocation

cases: A & B

Y

Y

N

N

Allocated to Male,

Female or None k ≤ K

N

k = 1

k = k + 1

k-th = index of # car allocation decisions K = # car allocation decisions

k = 1

k = k + 1

Allocated to Male,

Female or None

STOP

k ≤ K

N

STOP

Y

FIGURE 4.2 The Process of Car Allocation Model for Work Tours

76

In order to determine how many times such car allocation decisions should be made in

a household on the day considered, we need to identify the number of work episodes

performed by male and female heads. Table 1 shows the car-allocation cases that can

be distinguished in that respect. Case A represents the situation that only one of the

heads conducts one work episode, leading to only one car allocation decision in the

household. In this case, one head may use the car, but also there is an option that

he/she may not use the car. In Case B, two work episodes are included for only one of

the household heads (for example, he/she returns home for lunch). This situation thus

involves two car allocation decisions when the break is long enough to allow for

traveling back home and back to work again. In case C, both heads have one work

episode, implying that one or two car allocation decisions have to be made by the two

persons. One car allocation decision is to be made if the work episodes of the two

heads overlap in time (taking travel times into account). On the other hand, when there

is no overlap in time, 2 car allocation decisions have to be made.

The same principle of overlapping episodes also applies to Case D and Case E, leading

to maximally 3 and 4 car allocation decisions respectively. For example, in Case D,

when the male worker has 2 episodes and female worker has 1 episode, there are 1, 2,

or 3 car allocation decisions involved. In case both the first and second work episode of

the male are overlapped with the work episode of the female, then there is only 1 car

allocation decision required. If the first episode of the male worker and the episode of

the female worker are overlapped while the second episode of the male worker is not

overlapped with the episode of the female worker, this would imply 2 car allocation

decisions. Furthermore, if none of the two work episodes of the male are overlapped

with the female’s one, then 3 car allocation decisions are needed. The similar reasoning

applies to Case E. In the stage of the activity-scheduling processes where the work-

M

F

M

F

M

F

A B

C

D

M

F

E

M

F

FIGURE 4.3 Examples of Distinguished Cases

77

related car allocation decisions are made, other activities have not yet been scheduled.

Therefore, other activities that, in the end, are possibly attached to the work activity are

not considered in this model.

In determining whether or not there is an overlap in time, the travel time has to be

taken into account as well. The travel time by car mode (across the road network) is

relevant here. First, the timing and duration of work episodes of the household heads

are derived and then the type of overlap is determined. Note that, travel time by car is

used because that is relevant for car allocation decisions. Further details are provided in

Section 4.5.

4.5 EMPIRICAL ANALYSES

In this section we describe the results of deriving a decision tree model for car

allocation choice. Before discussing these results, we will first consider some

descriptive analyses carried out to get a better understanding of the characteristics of

the sample after selecting car deficient households. Next, we briefly discuss CHAID,

which is the decision tree induction method we use to derive decision rules from the

MON data. To facilitate interpretation of decision tree results, we use a post-processing

technique called impact tables. The impact table technique will be briefly discussed in

the section that follows. Finally, in the last section, we discuss the results of the

induction of the car allocation decision tree model and the corresponding impact table.

4.5.1 Descriptive Analyses

As discussed above, only a subset of households is relevant for the car allocation model,

because the problem concerns car allocation to work activities in car deficient

households. A total of 3,523 households were selected from the MON data, yielding

4,096 relevant cases of car allocation decisions. To describe the final sample, some

further descriptive analyses were conducted.

TABLE 4.2 Distributions of Households across Household Composition and SEC

(%)

SEC Household Composition

Low Mid-Low Mid-High High Total

Double, One Worker 3.2 13.3 12.3 11.6 40.3

Double, Two Worker 0.8 11.7 19.7 23.9 56.1

Double, No Worker 0.9 1.2 1.0 0.5 3.6

Total Sample (4096) 4.9 26.2 32.9 35.9 100

78

TABLE 4.3 Distributions of Household Heads across Household Composition and

Work Status of Household Heads by Gender (%)

Work Status, Male Work Status, Female Household

Composition Non-

worker

Part-

time

Full-

time

Total Non-

worker

Part-

time

Full-

time

Total

Double, One Worker 10.8 1.9 27.6 40.3 29.5 3.2 7.6 40.3

Double, Two Worker 0 8.4 47.7 56.1 0 31.2 24.9 56.1

Double, No Worker 3.6 0 0 3.6 3.6 0 0 3.6

Total Sample (4096) 14.4 10.3 75.3 100 33.1 34.4 32.5 100

TABLE 4.4 Work Duration Statistics by Work Status and Gender

Male Female

Working

Status

Average

duration of

work activity

(min)

Standard

Deviation

(min)

Freq. Average

duration of

work activity

(min)

Standard

Deviation

(min)

Freq.

Part-time 293.78 245.20 422 207.09 219.70 1412

Full-time 373.63 235.71 3085 257.88 245.47 1331

Total 364.02 238.26 3507 231.73 233.90 2743

TABLE 4.5 Work Duration Statistics by Day of the Week and Gender

Male Female

Day of the

Week

Average

duration of

work activity

(min)

Standard

Deviation

(min)

Freq. Average

duration of

work activity

(min)

Standard

Deviation

(min)

Freq.

Monday 373.84 237.49 708 239.32 235.01 576

Tuesday 381.85 228.92 663 249.10 238.62 513

Wednesday 384.33 233.22 623 230.87 229.94 485

Thursday 367.19 235.38 683 226.62 231.05 541

Friday 356.01 241.31 635 223.60 236.36 461

Saturday 237.96 242.12 135 211.68 228.38 114

Sunday 172.83 230.95 60 155.15 219.24 53

Total 364.02 238.26 3507 231.73 233.90 2743

79

Table 4.2 displays the frequency distribution of households across household

composition and socio-economic class combinations after selection. High-level income

households are in the majority (35.9%) and consist most frequently of double-two-

worker households. Double means two adults (male-female) household. This is

followed by mid-high income (32.9%), mid-low income (26.2%) and low income

households (4.9%).

The distribution of household heads across household composition and work status by

gender is presented in Table 4.3. Over 75% of males are full-time worker. Females are

approximately equally distributed across the work-status categories (33.1%, 34.4% and

32.5% for no, part-time and full-time worker respectively). This suggests that gender

still plays an important role in work commitments and task allocation.

Table 4.4 shows the distribution of duration across work activities for male and female

heads by work status. Note that persons may conduct more than one work activity a

day; the figures presented refer to durations on a per-activity basis (as opposed to a per-

episode basis). As we can see, males on average work approximately one and a half

times as long hours than females per work activity. Furthermore, in each work status

group, the average duration of males’ work activity is higher than that of female. The

frequency of work activities conducted by full-time male worker is leading among its

class, as a result of the fact that 75% of the males work full-time. This also suggests

that gender still plays a significant role in household task allocation.

Finally, Table 4.5 describes the household heads work activity duration split up by day

of the week. As can be seen, on average, working hours of males is similar from

Monday through Friday, about 6 hours. Meanwhile, working hours of females is on

average about 3-4 hours during working days. Again, this result shows that on average

males work longer hours than females per work activity.

4.5.2 Decision Tree Induction

We applied a CHAID-based tree induction method to identify the decision rules that

can describe car allocation choice behavior. CHAID (Kass, 1980) generates non-binary

trees, i.e., trees where more than two branches can be attached to a single root or node,

based on a relatively simple algorithm that is particularly well suited for the analysis of

large datasets and probabilistic action assignment. Other commonly used decision tree

induction systems are C4.5 (Quinlan, 1993) and CART (Breiman et al, 1984). All these

methods use a recursive process of splitting the sample based on condition variables

into partitions that are as homogeneous as possible regarding the action variable (i.e.,

the car allocation choice in this case). CHAID relies on the Chi-square test to

80

determine the best next split at each step. To determine the best split at any node, it

merges any allowable pair of categories of the condition variable if there is no

statistically significant difference within the pair with respect to the action variable.

This is done for each candidate condition variable. The split having the highest

significance value (after Bonferroni correction for multiple tests) across condition

variables is selected and implemented. The process is repeated until no more

significant splits are found also taking into account a pre-defined minimum number of

cases requirement at leave and parent nodes. This process of extracting rules is the

same as the one used in the ALBATROSS model. In order to develop the decision tree,

75% of the cases were used for training and the remaining cases were used for

validation. Generally, in deriving ALBATROSS decision models, attributes of the

household, person, space-time setting and schedule as far as known in the stage

considered of the assumed decision process are used as condition variables.

Observations of condition variables and action variables (car allocation choice) in each

case are extracted from the diary data.

The CHAID decision tree induction method allows one to define the threshold for

splitting in terms of a significance level for the Chi-square ( 2χ ) measure and a

minimum number of cases at leaf nodes. Alpha was set to 5% and the minimum

number of cases to 50. The number of leaf nodes gives an indication of the complexity

of the resulting tree. As a measure of prediction accuracy, the expected hit ratio is used.

The expected hit-ratio represents the expected proportion of cases predicted correctly

when a probabilistic action assignment rule is used. It is calculated as: ∑kik

ki

N

f

N

2)(1

where fki is the frequency of the ith action at the kth leaf node, N is the total number of

cases and Nk is the number of cases at the k-th leaf node. Note that the expected hit

ratio is comparable to a likelihood measure and, generally, yields lower scores than the

deterministic counterpart of the measure.

4.5.3 Deriving Impact Tables

Decision trees derived from data may become very large and complex and,

consequently, difficult to interpret. This holds true particularly for the present

application where the number of choice observations is very large. Arentze and

Timmermans (2003) developed a method to derive elasticity information from rule-

based models to facilitate interpretation, which we will use here to describe the results

of tree induction. The principle of the proposed method is straightforward. After

having derived a rule-based model from training data, the model is used to predict for

each condition variable a frequency cross table with the levels of the condition

81

variables in rows and the frequency distribution across the levels of the target variable

(i.e., the action variable) in columns. The frequency table for a given condition variable

is generated by applying the model as many times as there are levels of the condition

variable. In each run, each training case is assumed to take on the level considered on

the condition variable. The frequency distribution across actions of the action variable

predicted under that setting is recorded. Repeating this process for each level of the

condition variable yields a frequency cross table of the condition variable against the

action variable. The impact of the condition variable is then measured as the Chi-

square for this frequency table. Formally:

( )s sIS D= F [4.1]

where D is a Chi-square measure of the frequency table generated (Fs) for condition

variable s. This measure can be decomposed into a measure of impact on each level of

the action variable, as follows:

( )si siIS D= F [4.2]

where again D is a chi-square measure and Fsi is the vector of predicted frequencies of

the i-the action under the levels of the s-th condition variable.

Apart from impact size, we also use a measure of the direction of impact proposed by

Arentze and Timmermans (2003) defined as:

∑

∑

=

−

=−

−

−

=J

jjiij

J

jjiij

si

ff

ff

MS

2

1,

2

1,

||

)(

[4.3]

where fij is the predicted frequency of action i under the j-th level of condition variable

s and J is the number of levels. This measure can be interpreted as a measure of

monotonicity. If the condition variable has a monotonically increasing impact on the

frequency of action i across the levels of the condition variable, then MSsi equals 1 and

if it has a monotonically decreasing impact it equals -1. Any value in between these

extremes indicates that the impact is non-monotonous in the direction indicated by the

sign across the range of the condition variable. We emphasize that the monotonicity

measure is meaningful only for variables that are naturally enumerated; it is not

informative for variables that are purely nominal.

82

4.5.4 Condition and Action Variables

Table 4.6 portrays the condition variables that were used as input to the tree-induction

algorithm. The condition variables concern household level (including accessibilities),

individual level, and activity level variables (note that in this stage of the scheduling

process only work activities are known). Continuous condition variables, such as travel

time, duration, and parking price, are discretisized by using an equal-frequency interval

method which divides a continuous variable into n parts, in which each part contains

approximately the same number of cases.

The presence of young children in a household is taken as a condition variable as well

as other household and individual attributes, such as work status, socio-economic class

(in Euro), urban density (number of home addresses per area unit in the zone where the

household lives classified on a 5-point scale) and the day of the week (no. 1-8 in Table

4.6). The number of work activity episodes that is performed by male or female is 0, 1

or 2 episodes (no.9-11). Accessibility variables, such as travel time, train and bus

connections, parking price and free-paid parking place ratio were also used (no.12-29,

except no.18-19). They are calculated based on national datasets of the transport

system (car, bike/walk and public transport), parking facilities and land-use system

(employment data by sector and postcode area). They all relate to the trip to the work

location. If there is no work activity conducted by the person on a particular day, the

variables are set to zero for that person. If a work activity is conducted in the same

postcode area as where the person lives, then travel time is set to zero too. Travel time

by car is included as a direct measure of accessibility. Travel time ratios between

modes are used as indicators of relative accessibility by particular modes. Ratios are

used to allow the algorithm to identify impacts of modes more easily.

Work duration is an attribute of the activity for which a car allocation decision is made

(no.18-19). The definition of this variable takes the overlap pattern into account. To

explain this, consider for example, a case where the male has a work activity of 9 hours

and the female has two work episodes of 4 hours each with a one hour break in

between. In this case, there are two allocation decisions if the overlap concerns only

one of the female’s work episodes. For both decisions the considered work duration for

the male is 8 hours and for the female 4 hours. On the other hand, if the male’s work

activity overlaps with both female’s work episodes, just one allocation decision needs

to be made. For that decision the considered work duration for the male is 9 hours, as

before, but for the female it becomes 8 hours. Note that some of the variables relate to

the schedule level (a day of the household), whereas others are defined at the level of

the activity which involves a car-allocation decision (i.e., a work activity of one or both

of the heads).

83

TABLE 4.6 Condition Variables for Car Allocation Model

No Variable Classification Acronym

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

Urban Density

Household Composition

Presence of the youngest children

Day of the week

Age of person

Socio-economic class

Working status – M

Working status – F

Number of work episodes – M

Number of work episodes – F

Number of work episodes in household

Travel time by car – M (in minute)

Travel time by car – F (in minute)

Travel time ratio between PT and car –

M

Travel time ratio between PT and car –

F

Travel time ratio between car and bike

– M

Travel time ratio between car and bike

– F

Duration of work episode – M (in

minute)

Duration of work episode – F (in

minute)

Train accessibility – M

Train accessibility – F

Bus accessibility – M

Bus accessibility – F

Work conducted by male

Work conducted by female

Ratio # paid parking places to total #

parking places – M

Ratio # paid parking places to total #

parking places – F

Average price of parking – M

Average price of parking – F

Overlapping between two persons’

episodes

Number of car allocation cases in

household

Type of case for allocating the car

0=most densely , 4= least densely

2=DONEWORK, 3=DTWOWORK,

4=DNOWORK

0=no children, 1=<6, 2=6-11, 3=12-17 yrs

0=Monday to 6=Sunday

0=<35, 1=35-<55, 2= 55-<65, 3= 65-<75, 4=

75+ yrs

0=0-16,250 (low), 1=16,251-23,750 (low-mid),

2=23,751-38,750 (mid-high), 3=38,750+ (high)

0= non-worker, 1= part-time, 2= full-time

0= non-worker, 1= part-time, 2= full-time

0, 1, 2

0, 1, 2

1,2,3,4

0=0; 1=≤8; 2=9-14; 3=15-22; 4=>22

0=0; 1=≤6; 2=7-11; 3=12-18; 4=>18

0=0; 1=≤1.00; 2=1.01-1.98; 3=1.99-4.11;

4=>4.11

0=0; 1=≤1.00; 2=1.01-2.14; 3=2.15-4.49;

4=>4.49

0=0; 1=≤0.25; 2=0.26-0.37; 3=0.38-0.81;

4=>0.81

0=0; 1=≤0.30; 2=0.31-0.42; 3=0.43-1.00;

4=>1.00

0=0; 1=≤275; 2=276-520; 3=521-565; 4=>565

0=0; 1=≤240; 2=241-380; 3=381-540; 4=>540

0= no, 1= yes

0= no, 1= yes

0= no, 1= yes

0= no, 1= yes

0= no, 1= yes

0= no, 1= yes

0=0; 1=≤0.09; 2=0.10-0.15; 3=0.16-0.28;

4=>0.28

0=0; 1=≤0.07; 2=0.08-0.14; 3=0.15-0.24;

4=>0.24

0=0; 1=≤9; 2=10-25; 3=26-66; 4=>66

0=0; 1=≤8; 2=9-22; 3=23-44; 4=>44

0=no, 1=yes

1=1, 2=2, 3=3, 4=4

1=Male only, 2=Female only, 3=M&F (each 1

ep), 4=M (2 ep) & F (1 ep), 5=M (1 ep) & F (2

ep), 6=M&F (each 2 ep)

Urban

Comp

Child

Day

Age

SEC

WstatM

WstatF

NworkM

NworkF

NworkH

H

TTcM

TTcF

TTptM

TTptF

TTcbM

TTcbF

DurM

DurF

TrAcM

TrAcF

BusAcM

BusAcF

Mwork

Fwork

RParkM

RParkF

PParkM

PParkF

overlap

NcarAl

cases

Note: M = Male; F = Female; PT = Public Transport; ep = episode

84

The variables that correspond to the schedule level are number of work episodes of

male and female respectively and number of car-allocation-decision cases in a

household (no.31). The number of car allocation cases occurring in a household can be

1, 2, 3, or 4 cases (see Table 4.1 and Figure 4.3). The variables at activity level are the

following. For each car allocation case, the timing of work activities of both persons

have to be considered to determine whether or not there is an overlap in time (no.30).

Obviously, if only one person performs a work activity in a particular time period, then

there is no overlap in time, and otherwise there might be. Variable no.32 indicates the

type of overlap in terms of all possible combinations of number of work activities

(none, one or two) by male and female. In Cases (1) and (2) only the male or female

has a work activity in a particular time period. In contrast, in cases (3) to (6) there is a

time overlap between their work activities.

As a result, a total of 32 condition variables were defined. The action variable, as the

output of the car allocation model, involves assigning the car to male, female, or none

of the two household heads.

4.5.5 Results

For deriving the car allocation model for work activities, a total of 4,096 observations

could be derived from the data set. 75% of these cases (3,114) were used for training

and the remaining cases were used for validation. Of 4,096 cases, the probabilities of

the car being allocated to male and female are 37.28% and 17.77% respectively. In the

remaining cases, 44.95%, male and female heads use other modes to the work place.

Table 4.7 shows the frequency distribution of allocation decision outcomes over

household types in terms of work status of the heads. In households where male is a

non-worker, in about 50% of the cases household heads choose some other mode to

travel to the work place. In households where male is a part-time worker and the

female is a non-worker, the car is allocated to the male in about 43.59% of the cases.

However, if both male and female are part-time workers, about 50.75% of the cases

they use some other mode than car. In households where male is a full-time worker and

female is a non-worker, the car is allocated to the male in 43.67 % of the cases. In sum,

the figures show that the male gets the car more often than the female even in two-

worker households.

Given a minimum group size of n=50 cases at leaf nodes and a 5% alpha level, the tree

generated by CHAID consists of 29 leaf nodes (decision rules). The hit ratio (based on

85

a probabilistic assignment rule and the training set) of the model, compared to a null-

model (a root-only decision tree) indicates a significant improvement achieved by the

tree: the hit-ratio of the null-model of 0.374 is significantly increased to 0.540.

Figure 4.4 shows the resulting car allocation tree model graphically by branch from the

root note. The first split is implemented on travel time ratio between car and bike for

the work activity performed by female (TTcbF). Recall that the variable is set to zero if

the person has no work activity or the work activity takes place in the same post code

area as the home location. This results in five branches from the root. Branch #1

represents the condition where the female has no work activity or a zero travel time to

the work place. Within this node a split is implemented on travel time by car for the

male work activity (TTcM), and so on. The probability distribution across male, female

and none options is shown in italic font at each leaf node.

TABLE 4.7 Frequency Distribution of Work Status across the Action Variables

Work status % of getting the car No

Male Female Male Female None Total

1 Non-worker 32.88 15.75 51.37 146

2 Part-time 9.23 40.00 50.77 130

3

Non-worker

Full-time 36.10 13.74 50.16 313

4 Non-worker 43.59 16.67 39.74 78

5 Part-time 38.81 10.45 50.75 67

6

Part-time

Full-time 25.99 25.99 48.01 277

7 Non-worker 43.67 7.26 49.07 1129

8 Part-time 35.88 20.91 43.21 1215

9

Full-time

Full-time 36.98 27.13 35.90 741

Total 1508 747 1841 4096

Each path from the root to a leaf node represents a decision rule. For example, the path

printed in bold (Figure 4.4) represents the rule:

IF: TTcbF = 0 ∧ TTcM = 1 ∧ TTcbM = 0-2 ∧ DurM = 3-4

THEN: Male = 35.1%, Female = 0%, and None = 64.9%

The rule denotes that IF female either does not have a work activity or the work and

home location are in the same zone AND travel time (by car) of male is 8 minutes or

less AND the travel time ratio between car and bike for the male is less than 0.37

(traveling by car is at most 2.7 times as fast as the bike) AND male’s work duration is

at least 521 minutes (8.68 hours), THEN the probability that the male gets the car is

35.1%. Thus, the propensity of not using the car to work by male is as high as 64.9%

86

under these circumstances (where the male’s work location is relatively well accessible

by bike).

As another example, in branch #2, the rule printed in bold indicates that IF travel time

ratio between car and bike for female is less than 0.30 (relatively good accessibility by

car) AND the travel time ratio between car and bike for male is at least 0.26 (relatively

good accessibility by bike), THEN the probability of female getting the car (26.2%) is

yet lower than that of male (58.4%). This rule indicates that even if the male’s work

place is well accessible by bike, the propensity of the male to use the car is

considerably higher than that of female. Furthermore, in branch #3, as another example,

the rule printed in bold indicates that IF the female and male both have a work activity

and the travel time ratio between car and bike for the female is in between 0.31 and

0.42 (the car is between 3.2 and 2.4 times faster than bike) AND travel time ratio

between public transport and car of male is greater than zero AND there is a train

connection between home and the female’s work location, THEN the female’s

probability of getting the car is substantially higher than the male’s, namely 57.4% and

29.8% respectively.

The results of a performance analysis are shown in Table 4.8 in the form of a confusion

matrix for the training and validation set. A confusion matrix describes the model

performance in terms of a distribution of predicted choices for each observed choice

category in the data set. The confusion matrix shown is based on probabilistic model

predictions. The diagonal will have high numbers in case of good prediction. Off-

diagonal elements of the matrix indicate the probabilities of predicting wrong actions

for each observed choice category.

Table 4.8 shows that the model achieves a substantial improvement compared to a null-

model as diagonal cells have higher percentages. For example, as it appears in the

training set, in 37.7% of the cases we observe males using the car for the work activity.

In 54.3% of these cases the model predicts car allocation correctly, while for the

remaining cases the model predicts incorrectly that the female will use the car (9.2%)

and none of the heads use the car (36.5%). Note also that, due to the probabilistic

assignment rule used, the predicted distribution exactly matches the observed

distribution overall cases. In that sense the predictions are bias free. Comparing the

diagonals of the training and validation set suggests a small decrease in accuracy. As

the bottom-right cell shows, the overall accuracy on the validation set is slightly

decreased from 0.540 to 0.534. We consider the small decrease in accuracy as

acceptable.

87

To evaluate the quantitative impacts of each condition variable on the action variable,

Table 4.9 displays the impact table for the car allocation model. The condition

variables are listed in order of decreasing impact on the action variable overall (the IS

column). Note that ISmale, ISfemale, and ISnone show the size of the impact for each action

separately.

When we look at the differential impacts of types of condition variable, we see that

socio-economic variables have only a limited impact, whereas attributes of the

transportation system have a relatively big impact. Especially, travel time ratios and

TTcbF 4

Mwork 0

0; 0.500; 0.500

Mwork 1

0.474; 0.308; 0.218

Branch #3

Branch #4 Branch #5

Branch #2 TTcbF 1

TTcbM 0 TTcbM 1

0.613; 0.333; 0.053

SEC 0,1,2

0; 0.780; 0.220

TTcbM 2-4

0.262; 0.584; 0.154

TTcbF 2

TTptM 1-4 TTptM 0

TTcF 0-1

0; 0.220; 0.780 TTcF 2-4

0; 0.465; 0.535 TrAcF 0

0.442; 0.299; 0.259 TrAcF 1

0.296; 0.574; 0.130

TTcbF 3

TTcbM 0 TTcbM 1

0.712; 0.106; 0.182 TTcbM 2-4

TTcF 0-1

0; 0.203; 0.797

TTcF 0-1 TTcF 2-4

0.378; 0.378; 0.243

SEC 3

0; 0.574; 0.426

TTcF 2-4

0; 0.430; 0.570

TTcM 0-1

0.272; 0.141; 0.587

TTcM 2-4

0.547; 0.156; 0.297

TTcM 0 TTcM 1

TTcbM 3-4

0.315; 0; 0.685 TTcbM 0-2

TTcM 2 TTcM 3-4

Day 0,3

0.075; 0; 0.925 Day 1,4,2,5,6

0.267; 0; 0.733

DurM 0-2

0.539; 0; 0.461 DurM 3-4

0.351; 0; 0.649

TTcbM 2-4 TTcbM 0-1

0.735; 0; 0.265

SEC 2,3 SEC 0,1

0.614; 0; 0.386

PParkM 0-1 PParkM 2-4

Child 0

0.547; 0; 0.453 Child 1,3,2

0.320; 0; 0.680

TTcbM 0-3

0.728; 0; 0.272 TTcbM 4

0.536; 0; 0.464

TTcbF 0

TTptM 0-1

0.291; 0; 0.709 TTptM 2-4

0.535; 0; 0.465

Branch #1

FIGURE 4.4 Car Allocation Tree Model with 5 Major Branches

88

parking tariffs for the work location emerge with substantial impacts. The variable that

gives by far the biggest impact is the travel time ratio between car and bike for female

(TTcbF). The monotonicity measure (MSfemale = 0.33) clarifies that with increasing ratio

of this variable, the probability that the female gets the car increases. At first sight, this

seems implausible as the ratio indicates the relative accessibility by bike. However,

note that a value of zero of this ratio means that the female has no work activity or a

work activity in the home postcode area. Hence, an increase of the ratio from a zero

value means a change in condition from no travel to positive travel for the female and,

hence, an increase in the probability of getting the car.

The fact that the impact is non monotonous indicates that the probability does not

increase in the higher range, i.e. where an increase indicates an improvement of

relative accessibility by bike. The monotonicity measure for the variable that gives the

second biggest impact, TTcM, indicates that as travel time (by car) of male goes up, the

frequency of allocating the car to the male increases monotonically (MSmale = 1), as

expected. In sum, travel time and parking price variables have a big influence on car

allocation decisions between the two household heads in a car deficient household, as

indicated by the results of the first six variables.

In terms of socio-economic variables, we find that the most influential variable is

socio-economic class (SEC). Interestingly, the probability of getting the car decreases

monotonically for both male and female (MSmale dan MSfemale = -1.00) as income rises.

This result is somewhat counter-intuitive, given that car possession tends to be higher

among high income groups.

It should be noted, however, that since we consider car-deficient households we have

corrected for number of cars available in the household (we consider only double-adult

households having one car). Within this group, third variables such as education level

and availability of public transport at the work place may exert an influence. Income is

correlated with education level and possibly urban density at the location of

employment (larger cities) and the latter variables are correlated with use of public

transport. As a consequence, increasing income may lead to decreasing car allocation

to work activities. The probability of male getting the car increases when the male has

a work activity on the day concerned (Mwork). The presence of young children in the

household (Child) is the last socio-economic variable that has an impact on car

allocation decisions. Again interestingly, the tendency of not using the car by male

increases monotonically (MSnone = 1.00) when the value of this variable increases, i.e.

going from no children to presence of children with increasing age. Since there is at the

same time no influence on the probability that the car is allocated to the female, it

indicates that the car stays at home more often (possibly, for non-work activities of the

female).

89

TABLE 4.8 Confusion Matrix for the Training and Validation Sets

Training set (N=3114) Validation set (N=982)

Predicted Predicted Observed

Male Female None Total Male’ Female’ None’ Total’

Male 0.543 0.092 0.365 0.377 0.524 0.099 0.378 0.360

Female 0.197 0.471 0.332 0.176 0.166 0.478 0.356 0.184

None 0.307 0.130 0.563 0.448 0.321 0.113 0.566 0.455

Total 0.377 0.176 0.448 0.540 0.365 0.175 0.459 0.534

TABLE 4.9 Impact Tables of Condition Variables of Car Allocation Model

No Variables IS ISmale ISfemale ISnone MSmale MSfemale MSnone

1 TTcbF 3719.77 120.71 2376.06 1223.00 -0.16 0.33 -0.38

2 TTcM 948.75 519.71 0.01 429.02 1.00 -1.00 -1.00

3 TTcbM 446.41 257.61 87.46 101.34 0.09 -0.20 -0.03

4 TTcF 58.58 0.14 43.07 15.37 -1.00 1.00 -1.00

5 PParkM 50.8 28.82 0.00 21.99 -1.00 - 1.00

6 TTptM 45.57 25.64 0.20 19.74 1.00 -1.00 -1.00

7 SEC 8.2 2.15 2.02 4.02 -1.00 -1.00 1.00

8 Day 5.69 3.10 0.00 2.59 0.33 - -0.33

9 DurM 5.11 2.79 0.00 2.32 -1.00 - 1.00

10 TrAcF 4.66 0.54 3.79 0.33 -1.00 1.00 -1.00

11 Mwork 3.41 2.02 0.92 0.47 1.00 -1.00 -1.00

12 Child 2.42 1.33 0.00 1.09 -1.00 - 1.00

As for the situational variables, day of the week (Day) is the most influential variable.

There is a non-monotonous tendency (MSmale = 0.33) of increasing probability of

allocating the car to the male as the week proceeds from Monday to Sunday (the lowest

value on this variable is Monday). Day of the week has no influence on the probability

of allocating the car to the female. Another variable that has no influence on the

probability of allocating the car to the female is work duration of the male (DurM). The

probability of the male getting the car decreases monotonically as his work duration

goes up (MSmale = -1.00). The presence of a train connection between home and the

female’s work location (TrAcF) increases the probability of the female getting the car

and decreases the probability that the male gets the car. Probably, the existence of a

train connection acts as a proxy for distance and urban density: a train connection

generally exists only between locations with relatively high density and far enough

apart.

90

4.6 SUMMARY AND CONCLUSIONS

This paper considered car allocation choice behavior in car-deficient households

explicitly in the context of an activity-scheduling process. Focusing on work activities,

a car allocation model based on rules derived from a large travel diary data set using a

CHAID-based induction algorithm was presented. The face-validity of the decision tree

model is good in the sense that the derived rules and impacts of condition variables are

readily interpretable. The overall goodness-of-fit of the model is satisfactory. Although

the performance on a validation set decreased slightly, the set of decision rules seems

stable across training and validation set to a satisfactory extent.

The propensity of men driving a car to the work place is higher than that of women in

car deficient households, particularly, when women have no work activity or women’s

work place is in the same zone as the home location. This finding is consistent with a

common notion that women use a slow or public transport mode more often to travel to

activity locations. Similar to that, women tend to use the car when men have no work

activities or men’s travel time to work place is zero. When the female’s work location

is relatively well accessible by car, women are prevalent in getting the car.

In terms of decision rules results, in 43.1% of the rules men get the highest probability

to use the car while in only 20.7% of the rules women have the highest probability to

use the car. In the remaining of the rules (36.2%) none of the heads using the car gets

the highest probability.

As the impact table analysis showed, travel time variables and, in particular, the

relative accessibility of the work place by car compared to bike by far plays the most

important role in car-allocation decisions in two-driver, single-car households. Work

duration, day of the week and the existence of a train connection between home and

work location also has an impact on the decisions. Although socio-economic variables

appear to have only small effects on the decisions, presence of young children and

household income has an influence too.

As we showed, car allocation decisions can be modeled as an element of a more

encompassing activity scheduling process. ALBATROSS proved to be a suitable

framework for this. This focus of our approach meant at the same time that only a

limited set of explanatory variables at the level of the individual and household was

taken into account. From an analytical perspective, it is interesting to extend the set of

explanatory variables and investigate what the effects are of additional attributes such

as job characteristics and car characteristics on these decisions. Furthermore, given that

attributes of transportation systems appear to be significant, it is worth while to include

even more detailed descriptors of the transportation system, e.g. public transport

91

services and parking facilities. Finally, the present study focused on car allocation

decisions in relation to the work activity. Clearly, car allocation decisions may also

occur at the level of non-work activities in a scheduling process. The same approach

as developed in this study can be applied for that purpose.

REFERENCES



Arentze, T.A. and Timmermans, H.J.P. (2003), “Measuring the Goodness-of-Fit of

Decision-Tree Models of Discrete and Continuous Activity-Travel Choice: Methods

and Empirical Illustrations”. Journal of Geographical Systems 5: 185-206.

Arentze, T.A., and Timmermans, H.J.P. (2003), “Measuring Impacts of Condition


Empirical Illustration”, Geographical Analysis, 35, 24-45.





Bhat, C. R., Guo, J. Y., Srinivasan, S., and Sivakumar, A. (2004), “A Comprehensive

Econometric Microsimulator for Daily Activity-Travel Patterns”, Transportation Research Record No. 1894, Travel Behavior and Values 2004, Transportation

Research Board, National Research Council, Washington, D.C.

Bianco, M., and Lawson, C. (1996), “Trip-Chaining, Childcare, and Personal Safety:

Critical Issues in Women’s Travel Behavior”. Women’s Travel Issues, Proceedings

from Second National Conference, pp. 123-143.

Breiman L., Friedman, J.J., Olshen, R.A., and Stone, C.J. (1984), Classification and Regression Trees. Wadsworth, Belmont [CA].

Bowman, J.L. (2008), “From Theory to Practice: What Can We Learn from Our U.S

Experience?” Paper presented at TRB 2008, Washington D.C, USA.

Golledge, R.G., Kwan, M.P., and Garling, T. (1994), “Computational-Process

Modeling of Household Travel Decisions Using a Geographical Information

System”. Working Paper UCTC No.218, UCLA.

Hunt, K.L., and Petersen, E. (2004), “The Role of Gender, Work Status and Income in

Auto Allocation Decisions”. Paper presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C., USA.

Kass, G.V. (1980), “An Exploratory Technique for Investigating Large Quantities of

Categorical Data”. Applied Statistics 29, 119-27.

92

Kitamura, R., Pas, E. I., Lula, C.V., Lawton, T. K., and Benson, P. E. (1996), “The

Sequenced Activity Mobility Simulator (SAMS): An Integrated Approach to

Modeling Transportation, Land Use and Air Quality”. Transportation, 23, pp. 267-

291.

Kitamura, R., and Fujii, S. (1998), Two Computational Process Models of Activity-Travel Behavior. In T. Garling, T. Laitila, and K. Westin (eds) Theoretical

Foundations of Travel Choice Modeling, Elsevier Science, Oxford, pp. 251-279.

Miller, E. J. and Roorda, M. J. (2003), “A Prototype Model of Household

Activity/Travel Scheduling”. CD-ROM of the 82nd Annual Meeting of the Transportation Research Board, National Research Council, Washington, D.C.

Miller, E.J., Roorda, M.J., and Carrasco, J.A. (2005), “A Tour-based Model of Travel

Mode Choice”. Journal of Transportation Vol. 32, No. 4, pp. 399-422.

Pendyala, R. M. (2004), Phased Implementation of a Multimodal Activity Based

Modeling System for Florida. FAMOS: The Florida Activity Mobility Simulator.

Final Report Submitted to the Florida Department of Transportation, Research

Center. Volume I: Technical Documentation Available at:

http://www.dot.state.fl.us/research-center/Completed_Proj/Summary_PTO/FDOT_

BA496rpt.pdf and Volume II: Users Guide Available at:

http://www.dot.state.fl.us/research-center/Completed_Proj/Summary_PTO/FDOT_

BA496_Manual.pdf

Petersen, E. and Vovsha, P. (2005), “Auto Allocation Modeling in Activity-Based

Travel Models”. Paper presented at the 10th TRB Transportation Planning

Application Conference, Portland.

Quinlan, J.R. (1993), C4.5 Programs for Machine Learning. San Mateo, Calif.:

Morgan Kaufmann Publishers.

Vovsha, P., and Petersen, E. (2007), “Incorporating Car-Type Preferences and Intra-

Household Car Allocation into Travel Demand Models”. Paper presented at TRB 2007, Washington DC, USA.

Vovsha, P. (2008), “State of the Practice Metropolitan Activity Based Models in US”.

Paper presented at TRB 2008, Washington DC, USA.

Wets, G., Vanhoof, K., Arentze, T.A., and Timmermans, H.J.P. (2000), “Identifying

Decision Structures Underlying Activity Patterns: An Exploration of Data Mining

Algorithms”. Transportation Research Record 1718, Paper N. 00-0171.

Xie, C., Lu, J.Y., Parkany, E. (2003), Work Travel Mode Choice Modeling with Data

Mining: Decision Trees and Neural Networks. Transportation Research Record

1854, 50–61.

93

Chapter 5

MODELING JOINT ACTIVITY PARTICIPATION

AND HOUSEHOLD TASK ALLOCATION

Anggraini, R., Arentze, T.A., and Timmermans, H.J.P., 2008. Paper presented at the 10th International Conference of Advanced Application in Transport and Technology,

Athens, Greece

ABSTRACT This paper describes an empirical derivation of a household-level decision model of activity choice taking into account joint participation and task allocation between household heads. These are considered household-level decisions given that they involve commitments of multiple persons, in particular in two-head households. Attributes of households, for example the presence of young children and attributes of the work activities and space-time setting are considered as explanatory variables. To deal with this large set of attributes and account for non-linear relationships between the variables, a decision tree induction method is used to derive a decision tree model for each decision in an activity-scheduling process. Thus, we show how the decision tree models can be used as a component in an activity-scheduling model to predict travel demand in an activity-based micro-simulation system. The model shows a satisfactory performance in terms of goodness-of-fit on a hold-out set and face validity.

94

5.1 INTRODUCTION

Operational models of individual’s activity-scheduling behavior have begun to emerge

recently. Activity-scheduling models share the objective to predict the sequence of

decisions that leads to an observed activity pattern of a household/individual. Activity-

based models aim at predicting on a daily basis and for a household which activities are

conducted, by whom, for how long, at what time, the location, and the transport mode

that is used when traveling is involved (Arentze and Timmermans, 2000, 2005; Miller

and Roorda, 2003). There has been some research on the interactions of individuals

within households (Gliebe and Koppelman, 2002, 2005; Scot and Kanaroglou, 2002;

Srinivasan and Bhat, 2004), but fewer attempts to integrate the interactions in activity-

scheduling models.

The purpose of this study is to develop and test a model of households’ activity

participation decisions explicitly in the context of an activity-scheduling process.

ALBATROSS is one of the few operational activity-based models incorporating

household level decision making (Arentze and Timmermans, 2000, 2004, 2005). It is a

rule-based computational process model developed for the Dutch Ministry of

Transportation, Public Works and Water Management. In that respect, ALBATROSS

differs from other models, which use utility maximization or hybrid forms as a

framework for modeling activity scheduling decisions. In ALBATROSS, decision

rules for making scheduling decisions are extracted from activity diary data in the form

of a decision tree by using a CHAID-based decision tree induction method. The rules

predict actions in a probabilistic manner to reproduce non-systematic variance in

choice behavior.

The study focuses on two-head households and considers the joint decision making of

individuals related to household task allocation and joint participation in activities. We

propose an activity classification and identify the activity types that likely relate to the

needs at a household level and that can be allocated among the members. We use the

term (household) task activities to refer to these activities. In addition to decisions to

conduct and allocate task activities, the proposed model also predicts households’

decisions to conduct joint activities of a non-task nature on a given day.

The remainder of this paper is arranged into several sections. First, the next section

describes the proposed process model of activity-travel scheduling in the

ALBATROSS process model. Next, the CHAID algorithm that is used to induce

decision trees is briefly reviewed to give readers a better perspective of the

computational process model. Furthermore, the impact table that is used to measure the

size and direction of condition variables across action variables is described as well.

The subsequent sections describe the activity-travel data set used to derive the

95

decision-tree models and the results of deriving the decision trees from the data. The

paper concludes with discussing the major conclusions and remaining issues for future

research.

5.2 THE ACTIVITY SCHEDULING PROCESS MODEL

ALBATROSS (A Learning-Based Transportation Oriented Simulation Systems)

predicts for each household in a studied population the schedule of activities and trips

of each household head for a particular day. The activity scheduling process consists of

four major components: (1) work activity generation (including timing, duration,

location and transport mode choice for each work trip), (2) other fixed activity

generation (including timing, duration and location), (3) flexible activity generation

(including timing, duration and location), and (4) trip-chaining decisions and transport

mode choice for each tour. In the existing ALBATROSS model, interactions between

persons are represented only in a limited manner. Scheduling steps are made alternately

between the household heads whereby the condition of the schedule after each decision

step of one person is used as condition information in the next decision step of the

other person, and vice versa. Some aspects, such as activity allocation, car allocation,

and joint participation in activities and traveling, however, require joint decisions of

the two household heads (Anggraini, et al., 2007). In this study, we consider joint

decision making on the level of activity participation and show how this can be

modeled in the context of a scheduling process in a more elaborate way.

As the above-mentioned phases suggests, the activity types distinguished are grouped

into fixed activities and flexible activities. A fixed activity can be considered as an

activity that has to be done within a particular time horizon on a regular basis, due to

longer term commitments made by the individual. A flexible activity is an activity that

can be done freely at any time. Examples of fixed activities are work and escorting a

child to school, while most non-work activities are considered flexible activities. In

order to identify household-level decision making in activity scheduling and taking

into account available activity data, we cluster activities into 10 activity categories as

displayed in Table 5.1. These activities are similar to the classification used in the

current ALBATROSS model. Nevertheless, to distinguish person (P) and household

(HH) level activity-participation decisions, we subdivide each non-task activity

category into independent and joint activities. A task activity refers to a household task.

Bring/get person, shop-1-store, shop-n-store, and service-related activities (see Table

5.1) are considered task activities. A non-task activity just as a task activity can be

conducted anytime by any person in the household either independently or jointly.

96

TABLE 5.1 Activity Classifications in a Household

No Activity Clustered

Activity

Personal (P) or

Household

(HH) Level

Scope of Activities

1 Work Work P Full-time and part-time

2 Business P Work-related

3 Other

Work-

related P Other mandatory activity (school, etc)

4 Bring/get person HH Drop-off/pick-up children/spouse to a certain

location

5 Shop-1-store HH Shopping, 1 store

6 Shop-n-store HH Shopping, multiple stores

7 Service-related

Task

activity

HH Renting movie, getting (fast) food, institutional

purposes (bank, post office, etc)

Social-independent P 8

Social-joint HH Meeting friends, relatives, etc

Leisure-independent P 9

Leisure-joint HH

Sports, café/bar, eating out, movie, museum,

library, etc.

Touring-independent P 10

Touring-joint

Non-task

activity

HH

Making a tour by car, bike, or foot (eg., letting out

the dog, etc)

Social, leisure and touring activities (see Table 5.1) are considered non-task

discretionary activities. As said, also task activities possibly can be done jointly.

However, as will become clear later, joint participation is a choice within a next

allocation decision.

As we are concerned with task activities and non-task activities, the models developed

in this study fit in the stage of the scheduling process when the work activity (if any),

business activities (if any) and transport mode used for the work activity have been

scheduled by both persons. In this stage, household decision making involves the

selection of task activities (bring/get, shop-1-store, shop-n-store, service) and joint non-

task activities (social-joint, leisure-joint, and touring-joint) and, furthermore,

determining which person conducts which task activity whereby conducting the

activity by both partners jointly is one of the options. Thus, note that joint participation

is a possible outcome for both task and non-task activities, but the processes are

different. In case of a task activity it is the result of two decisions, namely to include

the activity and next to conduct the activity jointly. On the other hand, in case of a non-

task activity it is the result of a one-step decision, namely to include a joint activity in

the schedule of both household heads. The selection and allocation decisions involved

in these steps will be the focus of this paper.

97

Timing of task and non-task activities takes place in the next stage. It defines the

duration and start time of activity categories both at the household level and person

level. Having defined the timing, trip-chaining choices are made. The last two

components include the car allocation and transport mode choice, particularly for each

non-work tour. The latter choices are conducted at either household or person level

depending on whether the tour includes a joint activity or not. It is noteworthy that,

each decision in this process model is modeled by a decision tree whereby the results

of earlier decisions are used as conditions for each next decision. Decisions made are

transformed in operations on an evolving schedule. The process results in a complete

schedule for each person.

5.3 MODELS SPECIFICATION

As mentioned earlier, this study focuses on activity-travel decisions between the heads

of household in activity participation and household task allocation. To give a better

interpretation for readers, we explain both models’ structure in this section.

5.3.1 Activity Selection

We propose the following process model for activity selection decisions including both

the task and non-task joint activities. Activities are considered sequentially based on a

pre-defined priority ordering of activity categories. A particular priority order is

assumed which corresponds to the order in which the activities are listed in Table 5.1.

For each activity category in order of priority, a same decision tree is used to decide

whether an activity of that category will be conducted under a set of relevant

conditions. If the answer is yes, then the next decision by the same decision tree

involves whether or not a second activity of the same category is to be selected. This is

repeated until the answer is negative. Then, the next activity category is considered

repeating the same process. The process continues until a negative decision is

generated for the last activity category. The activity selection decisions are handled by

a (single) decision tree, as developed and tested in the sections that follow. It is noted

that the action variable of the decision-tree model is a yes-no decision, namely whether

the activity considered will be added to the (evolving) schedule of the household or not.

Activity type is included in the decision tree as a condition variable given the notion

that the type of activity will have an influence on this decision (e.g., activities with

higher normal frequency will have a higher a-priori probability of being selected).

98

5.3.2 Activity Allocation

Task activities added to the schedule are subject to a next allocation decision

determining who will do the task. The choice options are male, female, or both jointly.

Task activities are processed in the same order as before. In descending order of

priority, the activity categories are arranged as bring/get person, shopping to 1 store

(shop-1-store), shopping to multiple stores (shop-n-store) and service-related activities.

A single decision tree (to be developed below) will be used to make the allocation

decisions. In this decision-tree model, the condition variables used are the same as in

case of activity participation decisions, except that the number of activities in each task

and non-task activity category is used as additional condition variables.

5.4 DATA

The data used for deriving the decision trees originates from the Dutch National Travel

Survey (MON = Mobiliteit Onderzoek Netherlands) collected in 2004 covering all of

the Netherlands. The survey is conducted on a regular basis to obtain travel and activity

information of residents in the Netherlands. It is a household survey where data is

collected of all household members for the diary day as well as general information

about household and individual attributes such as, gender, age, vehicle ownership and

driving license ownership, home location, individual income, occupation, number of

working hours per week, etc. Respondents were also requested to give information

about all trips made on a designated day as well as on the activities conducted on trip

destinations. Information for each trip includes start time, trip purpose, destination,

activity type at the destination, and transport mode. Situational variables are reported

as well. All in all, this survey provides a comprehensive data source to analyze

activity-travel behavior of Dutch residents. In the data collection, 29,221 households

filled out a one-day travel/activity diary and 28,600 of these households fit the criteria

for being considered in ALBATROSS. The data were transformed to an activity-diary

data format for the current estimation purpose. In this study, we focus on two-heads

household, i.e. households consisting of a single head are not included in the analysis.

Then, there are 18,037 households used for deriving the envisioned decision tree.

5.5 ANALYSES


We applied a CHAID-based tree induction method to identify the rules that describe

which choices (i.e., actions) are made under which conditions. CHAID (Chi-square

99

Automatic Interaction Detector) generates non-binary trees, i.e., trees where more than

two branches can be attached to a single root or node, based on a relatively simple

algorithm that is particularly well suited for the analysis of larger datasets. Other

decision tree induction systems are C4.5 (Quinlan, 1993) and CART (Breiman et al,

1984). CHAID relies on the Chi-square test to determine the best next split at each step.

CHAID generates a decision tree by splitting subsets of the space into two or more

nodes repeatedly, beginning with the entire data set (Kass, 1980). The split that

maximizes a significance value of a Chi-square test, after adjustment for multiple tests

(Bonferroni adjustment), across condition variables is used for splitting if the split is

significant. The process is repeated for each newly created group until no more

significant splits are found. This process of extracting the rules is the same as the one

used in the original ALBATROSS model. In order to develop the decision tree, 75% of

the cases are used for training and the remaining cases were used for validation.

Generally, in deriving ALBATROSS decision models, the observed choice and

attributes of the household, person, space-time setting and schedule as far as known in

the stage considered of the assumed decision process are used as conditions and

extracted from the diary data in addition to the choice outcome (the target or action

variable).

The CHAID decision tree induction method allows one to define the threshold for

splitting in terms of a significance level for the Chi-square ( 2χ ) measure and a

minimum number of cases at leaf nodes. Alpha was set to 5% and the minimum

number of cases at leaf nodes to 50 (model 1) and 75 (model 2). As a measure of

prediction accuracy, the expected hit ratio is calculated. ALBATROSS uses a

probabilistic action-assignment rule and therefore the hit-ratio measure used here

represents the expected proportion of cases predicted correctly when a probabilistic

response-assignment rule is used. It is calculated as:2( )1 kq

kqk

f

N N∑ where fkq is the

frequency of the qth action at the kth leaf node, N is the total number of cases and Nk is

the number of cases at the k-th leaf node. Note that the expected hit ratio is comparable

to a likelihood measure and, generally, yields lower scores than the deterministic

counterpart of the measure.



consequently, be difficult to interpret. This holds true particularly for the present



100





variables in rows and the the levels of the target variable (i.e., the action variable) in

columns. The frequency table is generated by applying the model as many times as

there are levels of the condition variable. In each run, each training case is assumed to

take on the level considered on the condition variable. Thus, the results of a run

indicates the predicted frequency distribution for the action variable assuming that each

training case has the same determined level of the condition variable. Then, the impact

of the condition variable is measured as the Chi-square for the frequency table.

Formally:

( )s sIS D= F [5.1]

where D is a Chi-square measure of the frequency table generated (Fs) for condition

variable s. This measure can be decomposed into a measure of impact on each level of

the action variable, as follows:

( )si siIS D= F [5.2]

where again D is a chi-square measure and Fsi is the vector of predicted frequencies of

the i-the action under the levels of the s-th condition variable. Apart from impact size,

we also use a measure of the direction of impact proposed by Arentze and

Timmermans (2003) defined as:

( ), 12

, 12

ij i jj

si

ij i jj

f fMS

f f

−=

−=

−=

−

∑∑

[5.3]

where fij is the predicted frequency of action i under the j-th level of condition variable

s. This measure can be interpreted as a measure of monotonicity. If the condition

variable has a monotonically increasing impact on the frequency of action i across the

levels of the condition variable, then MSsi equals 1 and if it has a monotonically

decreasing impact it equals -1. Any value in between these extremes indicates that the

impact is non-monotonous in the direction indicated by the sign across the range of the

condition variable.

101


Table 5.2 portrays the condition variables that were used as input to the algorithm for

both decision trees. The condition variables concern household level (including

accessibility indicators), individual level, and activity level variables. Note that in this

stage of the scheduling process, work, business, and other mandatory activities are

known. However, only condition information related to the work activity is fully used

for condition variables, such as number of work activities, duration of each work

activity, mode to work place, and total time engaged in work activity. Continuous

condition variables, such as duration, are discretisized by using an equal-frequency

interval method which divides a continuous variable into n parts, in which each part

contains approximately the same number of cases.

It is worth noting that some variables are related to household level and person level, while others are defined at schedule level and activity level. Age of the youngest child

in a household is considered as a condition variable as well as other household

attributes, such as urban density (of the residence location), household composition,

day of the week, and socio-economic class (#1-5 in Table 3). Given the selection of

two-heads households, household composition refers to only three household types:

double-one-worker, double-two-workers and double-no-workers. The number of cars

in the household is also included as a household variable (# 21). A final set of

household-level variables relates to measures of accessibility of locations given the

home location of the household. On this level, 8 variables (#10-17) are included: (1)

daily goods sector: number of employees within 3.1 km, (2) non-daily goods sector:

number of employees within 4.4 km, (3) all sectors: number of employees within 4.4

km, (4) size of population within 3.1 km, (5) daily goods sector: distance within which

160 employees work, (6) non-daily goods sector: distance within which 260 employees

work, (7) all sectors: distance within which 4500 employees work, and (8) distance

within which 5200 people live.

Note that attributes related to individuals can be incorporated only if they are specified

explicitly for the male-female heads. Thus, individual attributes will be tied together

with gender. Individual attributes such as work status and age are explained in #6-9 in

Table 3. The work status attribute of the male-female heads indicates whether the

person has no work, part-time work, or full-time work. Those who work more than 32

hours per week are considered as full-time worker. Age and possession of driving

license by the householders are also one of the individual attributes (#8-9 & #22-23).

The following attributes are defined on the schedule level. The number of work

activities conducted by male and female is known and we limit it to maximally 2 work

activities per person (#18-20) (only very few diaries where more than two work

activities are included).

102

TABLE 5.2 Condition Variables for Decision Tree Models

No Variable Classification Acronym Category

1 Urban Density 0=most densely , 4= least densely1,2 Urban Ordinal

2 Household Composition 2=2 heads, 1 worker, 3=2 heads, 2 workers, 4=2 heads, no

workers1,2

Comp Nominal

3 Youngest children in HH 0=no children, 1=<6, 2=6-11, 3=12-17 yr1,2 Child Nominal

4 Day of the week 0=Monday to 6=Sunday1,2 Day Nominal

5 Socio-economic class 0=low, 1=low-mid, 2=mid-high, 3=high1,2 SEC Ordinal

6 Working status – M 0= non-worker, 1= part-time, 2= full-time1,2 WstatM Nominal

7 Working status – F 0= non-worker, 1= part-time, 2= full-time1,2 WstatF Nominal

8 Age of person – M 0=<35, 1=35-<55, 2= 55-<65, 3= 65-<75, 4= 75+ years1,2 AgeM Ordinal

9 Age of person – F 0=<35, 1=35-<55, 2= 55-<65, 3= 65-<75, 4= 75+ years1,2 AgeF Ordinal

10 Accessibility – 1 0=<=115, 1=<=253, 2=<=307, 3=<=507, 4=<=675, 5=>6751,2 nEmp1 Ordinal

11 Accessibility – 2 0=<=395, 1=<=635, 2=<=762, 3=<=938, 4=<=2525, 5=>25251,2 nEmp2 Ordinal

12 Accessibility – 3 0=<=8785, 1=<=12995, 2=<=16120, 3=<=20199, 4=<=70314,

5=>703141,2

nEmp3

Ordinal

13 Accessibility – 4 0=<=5050, 1=<=8845, 2=<=13217, 3=<=16833, 4=<=22884,

5=>228841,2

SizePop

Ordinal

14 Accessibility – 5 0=<=71, 1=<=127, 2=<=165, 3=<=202, 4=<=346, 5=>3461,2 Dist1 Ordinal



17 Accessibility - 8 0=<=0, 1=<=105, 2=<=126, 3=<=163, 4=<=278, 5=>2781,2 Dist4 Ordinal

18 Number of work episodes – M 0=no work, 1=1 ep, 2=2 ep1,2 NworkM Ordinal

19 Number of work episodes – F 0=no work, 1=1 ep, 2=2 ep1,2 NworkF Ordinal

20 Number of work episodes – HH 0=no work, 1=1 ep, 2=2 ep, 3=3 ep, 4=4 ep1,2 NworkHH Ordinal

21 Number of cars in HH 0, 1, 2+1,2 Ncar Ordinal

22 Driving license possession – M 0= no, 1= yes1,2 DrivM Nominal

23 Driving license possession – F 0= no, 1= yes1,2 DrivF Nominal

0=0; 1=≤435; 2=436-544; 3=545-575; 4=>5751 24 Duration of work act – M (min)

0=0; 1=≤390; 2=391-540; 3=541-570; 4=>5702 DurM Ordinal

0=0; 1=≤297.25; 2=297.26-480; 3=481-555; 4=>5551 25 Duration of work act – F (min)

0=0; 1=≤250; 2=251-385; 3=386-540; 4=>5402 DurF Ordinal

0=0; 1=≤475; 2=476-566; 3=567-820; 4=>8201 26 Duration of work act in HH (min)

0=0; 1=≤415; 2=416-550; 3=551-655; 4=>6552 DurHH Ordinal

27 Mode to work place – M 0= non-work act, 1=PT, 2=CP, 3=CD, 4=S 1,2 ModeM Nominal

28 Mode to work place – F 0= non-work act, 1=PT, 2=CP, 3=CD, 4=S 1,2 ModeF Nominal

29 Mode to work place – HH 0= non-work act, 1=PT, 2=CP, 3=CD, 4=S 1,2 ModeHH Nominal

30 –

35

Time available for non-work act

(1-6) – M (min) 0=0, 1=<=30, 2=30-60, 3=60-90, 4=90-1201,2

Time1M-

Time6M

Ordinal

36 –

41

Time available for non-work act

(1-6) – F (min) 0=0, 1=<=30, 2=30-60, 3=60-90, 4=90-1201,2

Time1F-

Time6F

Ordinal

42 –

47

Time available to use car for non-

work act in HH (1-6) 0=0, 1=<=30, 2=30-60, 3=60-90, 4=90-1201,2

Time1C-

Time6C

Ordinal

48 Given condition of work act – M 0= no, 1= yes1,2 yWorkM Nominal

49 Given condition of business act – M 0= no, 1= yes1,2 yBusiM Nominal

50 Given condition of other act – M 0= no, 1= yes1,2 yOthM Nominal

51 Given condition of work act – F 0= no, 1= yes1,2 yWorkF Nominal

52 Given condition of business act – F 0= no, 1= yes1,2 yBusiF Nominal

53 Given condition of other act – F 0= no, 1= yes1,2 yOthF Nominal

54 # activities 1 currently done in HH 0=0, 1=1, 2=2, 3=3, 4=4, 5=5+ 1,2 nbr Ordinal

55 # activities 2 currently done in HH 0=0, 1=1, 2=2, 3=3, 4=4, 5=5+ 1,2 nsh1 Ordinal

56 # activities 3 currently done in HH 0=0, 1=1, 2=2, 3=3+ 1,2 nshn Ordinal

57 # activities 4 currently done in HH 0=0, 1=1, 2=2, 3=3+ 1,2 nser Ordinal

58 # activities 5 currently done in HH 0=0, 1=1, 2=2+ 1 nsoc Ordinal

59 # activities 6 currently done in HH 0=0, 1=1, 2=2+ 1 nlei Ordinal

60 # activities 7 currently done in HH 0=0, 1=1, 2=2+ 1 ntou Ordinal

1=bring/get, 2=shop-1-store, 3=shop-n-store, 4=service,

5=social-joint, 6=leisure-joint, 7=touring-joint1 61 HH activity type considered 1=bring/get, 2=shop-1-store, 3=shop-n-store, 4=service 2

HHact Nominal

Note: M=Male; F=Female; ep=episode; PT=Public Transport; CP=Car Passenger; CD=Car Driver; S=Slow

1 Alternative classification used for activity participation model (model 1)

2 Alternative classification used for household task allocation model (model 2)

103

Duration of work activity conducted by male or female and the total work duration

across male-female in a household are also used as condition variable. Transport mode

used to the work place by male or female worker is also used as condition variable

(#27 & 28). In case multiple modes are involved, we aggregated the mode by arranging

modes in a hierarchy as follows: (1) Public Transport (2) Car Passenger (3) Car Driver

and (4) Slow (Bike and Walk) as can be seen in #29.

Additionally, available time in the schedule is captured. In order to identify how much

time is available for doing activities other than work activity at different times of the

day for male and female, we segment the period from 8 am to 8 pm into 6 time spans

of 2 hours. For each 2 hour-period we calculated the available time in the schedule and

classified this time into 5 categories, where zero means no time left for doing a non-

work activity and the remaining categories identify how much time is left for doing a

non-work activity in multitudes of half an hour (#30-41). In addition to available time,

the time a car is available for a non-work activity is also considered as a condition

variable. Hereby, the number of cars available and the transport mode(s) used for work

activities, if any, of the two household heads are taken into account. Time is

segmented in the same way as above (Note: zero means that either car is not available

due to work activities or because no car is available in the household).

Variables #48-53 indicate whether work, business and other mandatory activities (such

as go to school) are conducted on the given day or not. Note that, these variables are

also schedule-level variables. The remaining variables (#54-60) are activity level

variables. These variables indicate for each task and joint non-task activity category the

number of activities that, at the moment of the decision, are included in the schedule as

a consequence of previous activity selection decisions. Thus, these variables are

dynamic and depend on the assumed priority order of activities. At the time of the first

decision, no activity is included in the schedule. Therefore, all these variables (#54-60)

are zero in the beginning. For a second decision, the result of the first decision is

known and if this implied the insertion of a (bring-get) activity the corresponding

variable has a value of one, and so on. In sum, this set of variables indicates for each

current decision the current schedule state as a result of previous activity selection

decisions. Note that, for household task allocation model, only number 54-57 are being

concerned. The last variable encodes the activity type that is considered in the current

selection decision. This variable has seven levels corresponding to the seven activity

categories (task and joint non-task activities) considered in the process model.

Consequently, only four levels variables for the allocation model.

As a result, a total of 61 condition variables were defined for the activity participation

model as indicated by superscript 1 in Table 5.2. For household task allocation model,

104

the condition variables are actually almost the same as that we used for activity

participation model. Nevertheless, different classifications were used for some

variables, such as work duration, number of instances of a particular activity at the

moment the decision is made, and obviously the type of the activity considered, as

indicated by superscript 2 in Table 5.2. Hence, 58 condition variables remain for

household task allocation model.

5.5.4 Results: Activity Participation Tree

For deriving the activity participation model, a total of 153,856 observations could be

derived from the data set. 75% of these cases (115,458) were used for training and the

remaining cases were used for validation. Of 153,856 cases, the probability of

observing a yes decision for each activity category is as follows: bring/get person

20.2%, shop-1-store 46.5%, shop-n-store 9.1%, service 10.4%, social-joint 6.4%,

leisure-joint 4.5% and touring-joint 3.0%.

The tree generated by CHAID consists of 386 leaf nodes (decision rules). The hit ratio

(based on a probabilistic assignment rule) of the model compared to a null-model (a

root-only decision tree) indicates a modest but significant improvement: the hit-ratio of

a null-model equals 0.697 and the hit ratio of the tree after splitting equals 0.777. A

Chi-square-based contingency coefficient of 0.455 confirms that there is a moderately

strong impact of the decision tree structure on the action variable. The overall accuracy

on the validation set is almost the same, dropped slightly from 0.777 to 0.772.

Due to limited space and given the large number of decision rules, we cannot display

the entire results of the decision tree. Instead, in order to give a summary view of the

outcome, we will discuss the results of the impact analysis in terms of the IS and MS

measures explained above.

Table 5.3 displays the impact table for the activity participation model. In this case, the

choice variable is a binary variable (yes/no decision), so that the MS measures are

perfectly correlated (MSyes = − MSno). As it appears, activity type (HHact) is by far the

most important variable for the activity selection decision. The monotonicity index MS

in this case is close to zero (+/- 0.21) and negative for the yes decision indicating that

the frequency of adding an activity decreases across the activity categories in the order

they are put, but not monotonically. The second most important variable is day of the

week (Day). There is a tendency (MSyes = -0.23) of decreasing probability of adding an

activity of the activity category concerned to the schedule with increasing values of this

variable (running from Monday to Sunday).

105

TABLE 5.3 Impact of Condition Variables of HH Activity Participation Model

No Variables IS ISyes ISno MSyes MSno

1 HHact 107437 91962.17 15452.71 -0.21 0.21

2 Day 4168.91 3390.36 778.64 -0.23 0.23

3 nbr 2053.11 1616.28 436.82 0.23 -0.23

4 Child 1986.13 1601.17 384.98 0.05 -0.05

5 nsh1 1051.52 864.75 186.81 -0.49 0.49

6 nser 790.11 629.86 160.25 0.52 -0.52

7 nshn 501.63 405.30 96.39 0.29 -0.29

8 DurHH 499.02 406.40 92.64 -1.00 1.00

9 ModeF 273.66 221.14 52.53 -0.07 0.07

10 ModeHH 75.3 61.26 13.91 -0.25 0.24

11 Dist1 54.04 43.99 10.05 0.34 -0.33

12 SEC 48.87 39.61 9.23 1.00 -1.00

13 Time3M 26.23 21.22 5.01 0.81 -0.81

14 Time4F 23.57 19.14 4.38 1.00 -1.00

15 DurF 22.29 18.10 4.13 -1.00 1.00

16 ntou 20.8 16.79 4.00 1.00 -1.00

17 NworkHH 15.22 12.47 2.74 -1.00 1.00

18 AgeF 13.02 10.60 2.44 0.12 -0.12

19 yBusiM 12.69 10.37 2.32 -1.00 1.00

20 DrivF 11.34 9.30 2.06 1.00 -1.00

21 yWorkM 8.73 7.13 1.58 -1.00 1.00

22 NworkF 7.59 6.14 1.44 -0.15 0.14

23 AgeM 7.43 6.05 1.35 -1.00 1.00

24 Ncar 6.00 4.81 1.17 0.56 -0.57

25 yOthM 5.76 4.70 1.08 1.00 -1.00

26 Dist2 5.5 4.41 1.03 -0.04 0.02

27 DrivM 4.99 4.07 0.92 1.00 -1.00

28 ModeM 4.68 3.80 0.86 -1.00 1.00

29 Dist3 4.40 3.50 0.83 0.05 -0.06

30 Urban 4.31 3.54 0.81 0.06 -0.05

31 nlei 3.58 2.88 0.67 1.00 -1.00

32 nEmp2 3.52 2.81 0.64 -0.15 0.15

33 DurM 3.39 2.79 0.64 -1.00 1.00

34 Time1F 2.93 2.38 0.55 -0.25 0.25

35 SizePop 2.64 2.09 0.48 0.85 -0.85

36 Time4M 2.48 2.02 0.47 1.00 -1.00

37 NworkM 2.31 1.86 0.43 0.28 -0.29

38 WstatM 1.76 1.43 0.35 -1.00 1.00

39 Comp 1.65 1.40 0.29 0.38 -0.38

40 Time2F 1.56 1.24 0.27 1.00 -1.00

41 Time3F 1.47 1.20 0.28 0.63 -0.66

42 nEmp1 1.32 1.17 0.26 -0.07 0.07

43 Dist4 1.21 0.94 0.21 -1.00 1.00

44 nsoc 1.11 0.92 0.22 1.00 -1.00

45 Time3C 0.92 0.74 0.18 1.00 -1.00

46 Time1M 0.73 0.59 0.14 -0.02 0.03

47 Time4C 0.46 0.41 0.09 -1.00 1.00

48 Time5C 0.37 0.28 0.06 0.23 -0.23

49 nEmp3 0.33 0.28 0.07 1.00 -1.00

50 WstatF 0.17 0.17 0.04 -0.79 0.75

51 yWorkF 0.11 0.09 0.02 -1.00 1.00

52 Time5M 0.09 0.07 0.01 0.42 -0.39

106

The next most influential variable is the number of bring/get activities already included

in the schedule at the moment of the decision (nbr). Given a positive sign of MS for the

yes decision (MSyes = 0.23), there is a tendency of bring/get activities to generate other

activities. However, given that MS is smaller than 1, it does not monotonically increase,

at some point the probability of adding activities decreases with increasing number of

bring/get activities in the current schedule.

In terms of socio-demographic variables, in order of decreasing importance, we find

that presence/age of young children in the household (Child), income (SEC), age of

female head (ageF), female has a driving license (DrivF), age of male head (ageM),

number of cars (Ncar), male has a driving license (DrivM), work status of male

(WstatM), household composition (Comp), and work status of female (WstatF) all have

an influence.

As for the variable Child the MS index indicates that with increasing level of this

variable, the frequency of household activities is not increasing monotonically (MS =

0.05) across the levels (ordered as no children and children of increasing age group).

Interestingly, the MS measure for the variable that gives the second biggest impact,

SEC, indicates that as income goes up, the frequency of household activities

monotonically increases (MS = 1). Having a driving license also increases the number

of household activities as indicated by the positive sign (MS = 1) for DrivM and DrivF

variables.

In terms of the work-related condition variables, total duration of work activity across

the household heads (DurHH), mode to work by female (ModeF) and mode to work by

the two heads (ModeHH) turn out to be the most significant variables on this level.

With increasing total work duration (DurHH) the frequency of household activities

decreases monotonically, as one would expect (MS = -1). On the other hand, with

increasing values of mode (ordered as Public Transport, Car Passenger, Car Driver and

Slow mode) either by female or aggregated across the two householders the frequency

of household activities does not increase but rather shows a tendency to decrease (MS

= -0.07 and MS = -0.25).

The variables related to available time for non-work activities show influences on

household activity selection choices in expected directions. In particular, time available

during 12 am - 2 pm for male (Time3M) and during 2 - 4 pm for female (Time4F) have

almost monotonically increasing impacts on the activity frequencies (MS = 0.81 and

MS = 1.00). In terms of the accessibility of locations, the number of employees within

3.1 km in the daily good sector (Dist1) turns out to be the most influential variable.

With increasing number of employees the frequency of household activities increases,

107

although not monotonically (MS = 0.34). In summary, almost all of the 61 condition

variables used as input to the induction process recur in the decision tree. Only 9

variables do not affect predictions of activity participation, as indicated by zero value

of the chi-square-based impact measure.

5.5.5 Results: Task Allocation Tree

Having discussed the activity participation tree model, we now turn to the derivation of

the task allocation model. In total 22,512 observations could be derived from the data

set. To develop the decision tree, again 75% of these cases (16,893) were used for

training and the remaining cases were used for validation. Of 16,893 cases, the

probability of observing task allocation decisions for each household task activity

category is as follows: bring/get person 25%, shop-1-store 52.3%, shop-n-store 9.7%,

and service 13%. Overall, the probabilities of observing each person category are as

follows: male 31.8%, female 57%, and both 11.2%.

The tree generated by CHAID consists of 94 leaf nodes (decision rules). The hit ratio

(based on a probabilistic assignment rule) of the model, compared to a null-model (a

root-only decision tree) indicates a modest but significant improvement: the hit-ratio of

a null-model equals 0.439 and the hit ratio of the tree after splitting equals 0.545. A

Chi-square-based contingency coefficient of 0.504 confirms that there is a moderately

strong impact of the decision tree structure on the action variable. The overall accuracy

on the validation set is almost the same, dropped slightly from 0.545 to 0.537. Again,

space limitation does not allow us to discuss the structure of the decision tree. To give

a summary view of the result, we will discuss the results of the impact analysis in

terms of the IS and MS measures.

Table 5.4 shows the impact table for the task-allocation model. In this model, the

choice variable is a non-binary variable. Work duration of male (durM) appears to be

the most important variable for the allocation decision. There is a tendency of

increasing probability of female to do household tasks as male’s work duration

increases, although not monotonically (MSfemale = 0.88).

Over the same range, the probability of both (jointly) tends to decrease (MSboth = -0.78).

The second most important variable is the number of bring/get activities that are

included in the schedule of the household as the result of the previous participation

decisions (nbr). The probability that the female conducts a task activity increases

monotonically with increasing number of bring/get activities in the current schedule

(MSfemale = 1.00).

108

TABLE 5.4 Impact of Condition Variables of Task-Activity Allocation Model

No Variables IS ISmale ISfemale ISboth MSmale MSfemale MSboth

1 durM 9887.64 3615.91 3100.84 3170.67 -0.79 0.88 -0.78

2 nbr 7867.01 1799.28 1630.63 4437.18 -1.00 1.00 -1.00

3 nsh1 6188.85 562.57 1123.68 4502.61 -0.94 1.00 -1.00

4 durF 4324.79 2209.56 1235.52 879.65 1.00 -0.87 -0.52

5 yBusiM 904.81 100.46 250.61 553.74 -1.00 1.00 -1.00

6 nshn 562.63 9.47 97.49 455.70 -1.00 1.00 -1.00

7 nser 238.56 0.14 33.39 205.04 -1.00 1.00 -1.00

8 HHact 187.19 90.37 12.62 84.18 -0.27 0.67 0.05

9 time1F 97.84 57.56 15.58 24.69 -1.00 1.00 0.98

10 wstatF 64.95 44.63 18.04 2.27 -0.92 0.91 1.00

11 time4M 35.73 18.25 15.16 2.32 0.63 -0.62 0.56

12 durHH 32.42 21.78 7.96 2.67 1.00 -1.00 -1.00

13 Child 30.23 6.19 11.55 12.48 -0.34 0.37 -0.40

14 SEC 21.47 8.00 0.99 12.48 0.14 0.28 -0.52

15 Urb 13.11 0.11 1.94 11.07 0.81 1.00 -1.00

16 SizePop 7.08 1.00 0.18 5.91 -0.45 -0.80 1.00

17 AgeM 4.98 0.00 0.79 4.18 -1.00 -1.00 1.00

18 nEmp2 3.59 2.37 1.12 0.09 1.00 -1.00 -0.57

19 wstatM 3.38 1.70 1.43 0.25 1.00 -1.00 1.00

20 Ncar 3.11 0.47 0.50 2.13 -1.00 0.01 0.55

21 yWorkF 2.83 1.85 0.97 0.01 1.00 -1.00 -1.00

22 time5M 2.07 0.90 0.90 0.29 1.00 -1.00 1.00

23 Day 1.67 0.56 0.01 1.10 0.00 0.00 0.00

24 Dist1 1.21 0.49 0.51 0.20 -1.00 1.00 -1.00

25 drivF 0.76 0.02 0.19 0.55 -1.00 1.00 -1.00

26 time1C 0.52 0.30 0.21 0.01 -1.00 1.00 -1.00

27 modeF 0.5 0.32 0.10 0.06 0.00 0.00 0.00

28 time4F 0.4 0.24 0.15 0.00 -1.00 1.00 -1.00

29 AgeF 0.32 0.17 0.13 0.02 -1.00 1.00 -1.00

30 Dist3 0.32 0.21 0.07 0.03 0.61 -0.53 -1.00

31 Comp 0.14 0.05 0.06 0.02 1.00 -1.00 1.00

The same holds for number of 1-store shopping activities (nsh1), number of n-store

shopping activities (nshn) and number of service activities (nser) in the current

schedule. These results suggest that female tends to take on task activities particularly

when multiple tasks are scheduled. The next most influential variable is the work

duration of female (durF). The tendency of female getting a household task decreases

when her work duration increases, whereas the probability of the male doing a

household task increases monotonically (MSmale = 1.00). The last variable in the top 5

most significant variables is presence of a business activity in the male’s schedule

(yBusiM). The probability that the female takes on the household task increases when

the male has a business activity in the schedule. In contrast to the activity participation

tree, activity type (HHAct) is not the most important variable in the allocation tree.

109

Nevertheless, the probability of the female doing the household task increases non-

monotonically (MSfemale = 0.67) when the level of activity category rises in the order of

bring/get, shop-1-store, shop-n-store, service related.

In terms of socio-demographic variables, in order of decreasing importance, we find

that work status of female (WstatF), presence of young children in the household

(Child), income (SEC), age of male head (ageM), work status of male (WstatM),

number of cars (Ncar), female has a driving license (DrivF), age of female head (ageF),

and household composition (Comp) all have an influence. An increase of male’s or

female’s work status (WstatM and WstatF) tends to increase the probability of task

allocation to both of the heads monotonically (MSboth = 1.00). As for the presence of

young children in the household, Child, the MS index indicates that with increasing

level of this variable, the probability of female takes over the household task increases

non-monotonically (MS = 0.34) across the levels. Interestingly, the MS measure for the

variable of income level, SEC, indicates that as income goes up, the probability of male

and female jointly doing the household task decreases non-monotonically (MSboth = -

0.52). In terms of a person’s age, increasing age of the male tends to increase the

probability of both heads to conduct household tasks together (MSboth = 1.00).

However, increasing age of the female only seems to increase the female’s probability

to do the household tasks (MSfemale = 1.00). An increase of number of cars in the

household, Ncar, tends to decrease the male’s probability to conduct the household

task (MSmale = -1.00). Having a driving license increases the female’s probability to

perform the household tasks as indicated by the positive sign (MSfemale = 1.00) for

DrivF variable. Finally, as the level of household composition rises (Double-1-worker,

Double-2-workers, and Double-no-workers), the tendency of male to do household

tasks increases monotonically (MSmale = 1.00).


This study was intended to refine the ALBATROSS model. In the present paper, we

focus on activity participation choice of male-female heads, in particular those

activities that are related to a household task or that are conducted jointly, in order to

capture within-household interactions in a better way.

For household activity selection and task allocation decisions rule-based models using

a CHAID-based algorithm were derived from activity-trip diary data. The activity

participation model, given the large number of observations that could be derived from

the data, included more than 300 condition-action rules. Although slightly smaller, the

household task allocation model also involved an extensive set of decision rules,

110

involving more than 90 condition-action rules. In both cases, the validity of the

decision tree is satisfactory in the sense that the derived rules are readily interpretable

and the overall goodness-of-fit of the model on a validation set is acceptable as well.

Furthermore, in both cases, a substantial improvement in goodness-of-fit relative to a

null model indicates that there is a moderately strong association between condition

variables at household, individual, activity and schedule level, on the one hand, and the

participation and allocation decisions, on the other. Furthermore, the stability of

performance on a validation set suggests that derived rules are generalizable to unseen

cases. These results suggest that the way of structuring the household decisions as we

proposed in this study has merits.

To aid interpretation of the complex decision trees, the method of impact tables

(Arentze and Timmermans 2003) was used to measure the size and sign of the impact

of each condition variable on predictions. As expected, the most important variable in

household activity participation model is activity type, followed by day of the week

and number of bring/get activities already included in the schedule by the time of

making the activity participation decision. The presence of young children in the

household and income are also relatively influential variables among the socio-

demographic variables considered in the model. In terms of time-availability variables,

the total duration of work activity across household heads appears to be the most

influential variable. Accessibility variables only have modest impacts. Accessibility of

facilities in the daily goods sector appears to be the most important variable among the

accessibility variables. However, in general, the accessibility variables only have

modest impact on the frequency of doing household activities. This suggests that

spatial developments or policies will only have a slight influence on this aspect of

household behavior. On the other hand, a significant impact of income level suggests

that economic growth has larger impact on the behavior.

In the household task allocation model, the most important variable is male’s work

duration. Although not as big as the male’s influence, female’s work duration also has

an influence. Work status of female turns out to be the most significant variable among

the socio-demographic variables.

By refining the existing ALBATROSS in this way we expect that the accuracy and

sensitivity of predictions will be improved. The focus of the present study is on the

estimation of the model and the results of the model-based analysis for those particular

choice facets. Since the structure of the activity scheduling process differs, the models

cannot be compared on a single-choice facet basis. Only on the level of activity

patterns that are the result of a full activity scheduling process the two models could be

compared. This comparison is left as a topic of future research.

111

REFERENCES

Anggraini, R., Arentze, T.A., and Timmermans, H.J.P. (2007), “Modeling Car

Allocation Decisions in Automobile Deficient Households”. In: Proceedings ETC 2007 Conference, Noordwijk, The Netherlands.





Empirical Illustration”. Geographical Analysis, 35, 24-45.





Breiman L., Friedman, J.J., Olshen, R.A., and Stone, C.J. (1984), Classification and Regression Trees. Wadsworth, Belmont [CA].

Gliebe, J.P. and Koppelman, F.S. (2002), ”A Model of Joint Activity Participation

between Household Members”. Transportation, 29, pp.49-72.

Gliebe, J.P. and Koppelman, F.S. (2005), “Modeling Household Activity-Travel

Interactions as Parallel Constrained Choices”. Transportation, 32, pp.449-471.

Kass, G.V. (1980). “An Exploratory Technique for Investigating Large Quantities of


Miller, E.J., and M.J. Roorda. (2003), “A Prototype Model of Household

Activity/Travel Scheduling”. Proceedings of the 2003 Transportation Research Board, Washington DC, USA.

Quinlan, J.R. (1993). C4.5 Programs for Machine Learning. San Mateo, Calif.:

Morgan Kaufmann Publishers.

Scott, D. and Kanaroglou, P. (2002), “An Activity-episode Generation Model that

Captures Interaction between Household Heads: Development and Empirical

Analysis”. Transportation Research B, 36B: 875-896. Srinivasan, S. and Bhat, C. (2004), “Modeling the Generation and Allocation of

Shopping Activities in a Household”. Proceedings of the 2004 Transportation Research Board, Washington, DC.

112

Chapter 6

CONTINUOUS CHOICE MODEL OF TIMING AND

DURATION OF JOINT ACTIVITIES

Anggraini, R., Arentze, T.A., and Timmermans, H.J.P., 2009. Paper is accepted for

publication at Transportation Research Record, Journal of Transportation Research Board, USA In press.

ABSTRACT This paper contributes to the recent interest in household decisions in activity-based analysis. It focuses on the joint participation of male-female heads in non-work activities and attempts to model the timing and duration decisions for these activities, using decision tree induction. The data used originate from the 2004 National Travel Survey in the Netherlands. The results show that activity type has the most significant influence in both models. In addition, time availability for non-work activities during morning off-peak periods has a strong influence on start time decisions. The results also suggest that there is a substantial influence of duration decisions on start time decisions. Joint participation of household members in activities tends to lead to longer activity duration and earlier start times. Overall, modeling timing and duration of joint activity participation decisions at the household level proves to have some clear advantages.

113

6.1 INTRODUCTION

Interactions between persons in the household will strongly influence individuals’

activity-travel patterns. Joint activity participation requires the synchronization of the

activity patterns of the persons involved (Golob and McNally, 1997; Gliebe and

Koppelman, 2002; and Bhat and Pendhyala, 2005). In addition, understanding

relationships among different persons and their underlying motivations for activity

participation can also help to understand the potential impacts of policy triggers to

change travel behavior. All this explains a recent surge in research papers and

forecasting models that attempt to explicitly account for within-household interactions

in activity participation and travel (Goulias, 2000).

In the ALBATROSS system (Arentze and Timmermans, 2000, 2004, 2005), on which

we focus in this study, within household interactions are incorporated in the sense that

the process-wise sequential formation of an individual’s activity schedule does not only

take into account the previous schedule decisions of that individual but also those of

the spouse. Hence, agendas are formed sequentially and in parallel between the adults

in a household. Although this approach gives satisfying results, we decided to give the

model system an overhaul by looking more systematically at the modeling of the

various choice facets from the perspective of household decision making.

This paper reports the results of modeling duration and start time decisions for joint

activities. The results of this paper should not only be viewed as a step in refining the

process model of ALBATROSS but are also relevant in their own right. In the

ALBATROSS framework, duration and timing decisions for joint activities are

explicitly considered in the context of a broader activity scheduling decision process.

This means that only those choices that were already made in a previous scheduling

step can be considered to influence the duration and timing decisions for joint activities

at the moment they are made. The decision-process perspective also means that most

previous research on start time and activity duration is not immediately relevant for our

approach.

It may however be of interest to compare the ultimate results of the previous rich

literatures on departure time choice (e.g., Bhat and Steed, 2002, and Steed and Bhat,

2000), activity duration (e..g. Bhat, 2002, and Niemeier and Morita, 1996), and the

combination of timing and duration decisions (Pendhyala and Bhat, 2004, and Vovsha

and Bradley, 2004). Most studies addressed one particular activity purpose, such as

shopping or social activity, or one particular non-work group activity, such as

maintenance or leisure activities. Moreover, existing studies typically focused on

individual decision making and ignored the broader context of a daily activity schedule.

There are a few notable exceptions. Vovsha and Bradley (2004) considered the

114

relationship between departure time and duration in the context of an activity-based

model of activity travel demand. They considered independent and joint activities in

the context of travel tours. Habib et. al. (2008) also acknowledging activity scheduling

behavior emphasized the social context of activity scheduling decisions. They

investigated the influence of the “with whom” dimension of activities to start time and

duration decisions, where ‘with whom’ refers to joint participation by family members,

household members and/or friends. Using a hazard-based modeling approach, these

two choice facets are considered as continuous variables. In an empirical application,

they found that when household members participate jointly in a social activity, the

activity episodes are usually of longer duration and tend to start earlier on the day.

These results indicate that joint participation in activities tends to have significant

influences on duration and timing decisions.

In the present paper, we similarly consider the activity scheduling process context but

adopt a different approach. In line with a common assumption of the computational

process approach, we formulate the continuous timing and activity duration model as a

set of decision rules which are derived from choice observations using a decision-tree

induction method. Thus, the model proposed here differs from hazard-based and utility

maximization methods in the sense that a process perspective is adopted and rules

instead of algebraic equations are used to represent and predict the activity timing and

duration decisions of people/household. In this study, we consider the timing and

duration decisions for non-work activities conducted jointly by the male and female

head of a household. Specifically, this study investigates the timing of non-work

activities related to household and family activities, such as household tasks (e.g.,

escorting persons, grocery shopping) and non-task activities (i.e., social and leisure

activities). Thus, the study focuses on two-heads households (with or without children)

and the joint activities in their schedules. The paper is organized as follows: the next

section gives an overview of the ALBATROSS model. This is followed by an outline

of the approach, a description of data and methods, a discussion of empirical results,

and a summary of major conclusions.

6.2 OVERVIEW OF ALBATROSS MODEL

ALBATROSS is a learning-based transportation oriented simulation systems that is

capable of simulating daily activity schedules and travel patterns of individuals and

households. It predicts for each household of a studied population the schedule of

activities and trips of each household head for a particular day. In the existing

ALBATROSS model, joint decisions involved in joint activity participations are

represented merely implicitly. In order to refine ALBATROSS such that intra-

115

household interactions are better represented, in previous studies we focused on several

aspects that require joint decisions of the two household heads, such as the generation

of joint activities and the generation and allocation of household tasks (Anggraini, et, al., 2008). In addition to that, car allocation decisions, i.e. who uses the car for which

(work) activity, in case of households where the number of drivers exceed the number

of cars, was also taken into consideration (Anggraini, et, al., 2007).

In ALBATROSS, the activity types are grouped into fixed activities and flexible

activities. Mandatory activities are considered fixed activities, while non-mandatory

activities are termed flexible activities. Given the purpose of modeling household-level

decision making in an activity scheduling process, we distinguish within the category

of flexible activities household task activities and non-task activities. Mandatory

activities include work, business and other mandatory activities (e.g., school). A

household task activity refers to an activity that can be allocated to different household

members. A non-task activity is a discretionary activity that can be conducted anytime

by any person in the household. Household-tasks include the following activity types

in order of priority: (1) bring/get person, (2) shopping (one-store), (3) shopping

(multiple stores), and (4) service-related activities. Non-task activities include the

following activity types also in order of priority: (1) social visits, (2) leisure activities

(other than touring), and (3) touring (by car, bike or on foot). Since task as well as non-

task activities can be conducted jointly, we have in total 7 non-work activities to be

considered in the present analysis. In the diary data used for estimation, a joint (non-

work) activity in a household is identified as a particular non-work activity that occurs

in the diary of both the male and female head and takes place at the same location with

approximately the same start time (+/– 15 minutes) and approximately the same

duration (+/– 15 minutes).

In the ALBATROSS model, the timing of task and non-task activities takes place at

some stage in the activity scheduling process (Figure 6.1). Joint activities have priority

over independent activities and, hence, are scheduled first. Each time after having

added a joint activity of a particular type, if any, to the schedules of the two persons, a

duration and timing decision is made before adding a next activity is considered. This

means that at the moment a timing and duration decision is made for an added joint

activity only the mandatory activities (work/school/business) and joint activities of

higher-priority categories, if any, are known. It is noteworthy that each decision in this

process model is modeled by a decision tree whereby the results of earlier decisions are

used as conditions for each next decision. Decisions made are transformed in

operations on an evolving schedule. The process result is a complete schedule for each

person.

116

Generating Work

Activity

• Person-Level: # episodes, Start time, Duration, Location

• Household-Level: Car allocation to work place

• Person-Level: Transport mode to work place

Generating Business

& Other Mandatory

Activity

• Person-Level: # episodes, Duration, Start time, Link-Work,

Location

Generating Task

Activities and

Non-Task Activities

• Household-Level:

- Activity selection of joint activity categories

- Activity Allocation (for allocated activities)

• Person-Level:

- Activity selection of independent non-task activities

Timing of Task

Activities and Non-

Task Activities

Trip-Chaining Choices

Location of Task

Activities and Non-

Task Activities

Transport Mode of

Non-Work Tours

• Household-Level (if Joint): Duration, Start time

• Person-Level (if Independent): Duration, Start time

STOP

START

• Household-Level (if Joint)

• Person-Level (if Independent)

Car Allocation Decisions

for Non-work Tours • Household-Level

• Household-Level (if Joint)

• Person-Level (if Independent)

FIGURE 6.1 Household Activity-Travel Scheduling Process of

ALBATROSS

117

6.3 DATA DESCRIPTION

The data used for deriving the decision trees originates from the Dutch National Travel

Survey (MON = Mobiliteit Onderzoek Netherlands) collected in 2004 covering all of

the Netherlands. The survey is conducted on a regular basis to obtain travel and activity

information of residents in the Netherlands. It is a household survey where data is

collected of all household members for the diary day as well as general information

about household and individual attributes such as, gender, age, vehicle ownership and

driving license ownership, home location, individual income, occupation, number of

working hours per week, etc. Respondents were also requested to give information

about all trips made on a designated day as well as on the activities conducted on trip

destinations. Information for each trip includes start time, trip purpose, destination,

activity type at the destination, and transport mode. Situational variables are reported

as well. All in all, this survey provides a comprehensive data source to analyze

activity-travel behavior of Dutch residents. In the data collection, 29,221 households

filled out a one-day travel/activity diary and 28,600 of these households fit the criteria

for being considered in ALBATROSS. The data were transformed to an activity-diary

data format for the current estimation purpose. Given the present focus on two-heads

households, households consisting of a single head are not included in the analysis here.

This leaves 18,037 households for the present analysis. As the focus of this paper is on

modeling joint activities, a total of 4,515 days of households and 6,526 joint activities

were identified and used for deriving decision tree models.

The description of the MON data employed in this study is as follows. About 44.7% of

these households are zero-workers households, 32.3% are two-worker households, and

the rest is one-worker household.

TABLE 6.1 Independent and Joint Activity Frequency (percentage)

Weekday Saturday Sunday Activity

Indep Joint Indep Joint Indep Joint

Bring/Get 18.65 2.57 5.02 1.86 5.91 2.19

Shop-1-Store 30.34 30.07 39.02 30.18 4.04 3.29

Shop-N-Store 5.01 7.34 7.83 9.61 1.08 1.29

Service 8.23 9.35 3.93 2.44 4.63 1.48

Social 12.37 22.05 15.84 28.89 26.43 42.16

Leisure 15.78 17.03 19.48 18.78 33.02 27.31

Touring 9.62 11.59 8.87 8.24 24.90 22.27

Total 27448 3582 4677 1395 2032 1549

Note: Indep = independent activity

118

TABLE 6.2 Average Duration (minutes)

Weekday Saturday Sunday

Activity Indep

St.

Dev Joint

St.

Dev Indep

St.

Dev Joint

St.

Dev Indep

St.

Dev Joint

St.

Dev

bring/get 21.26 36.45 30.47 40.6 27.3 51.8 34.19 47.22 30.11 41.64 44.26 65.63

shop-1-

store 53.8 53.25 69.29 62.57 57.43 54.98 72.90 63.26 56.27 66.97 102.39 74.22

shop-n-

store 82.06 66.91 100.32 70.59 89.44 75.38 103.47 69.56 134.09 93.83 166.75 105.47

service 63.15 67.52 79.45 73.4 55.05 60.69 62.09 44.51 46.64 70.26 55.78 56.34

social 132.32 112.5 172.12 120.24 158.67 131.13 211.78 139.26 148.25 118.89 183.93 117.90

leisure 128.21 99.53 145.93 129 151.03 134.07 149.51 126.42 142.16 111.58 133.43 110.15

touring 29.47 84.47 78.85 150.65 43.44 103.91 102.81 164.29 48.80 99.21 70.51 118.75

Ave Dur 69.03 85.01 108.36 112.52 91.36 106.77 131.83 126.28 105.93 114.95 137.00 122.49

NOTE: Ave Dur = average duration ; St. Dev = standard deviation

0

200

400

600

800

1000

1200

start time

frequency

BR

SH1

SHN

SER

SOC

LEI

TOU

FIGURE 6.2 Start-Time Profiles every 30 minutes for each Activity

Concerning car ownership, 67.44% of the 4,515 households own 1 car, 27.2% have 2+

cars, and only 5.3% have no car. In terms of household income, the high income

households are in the majority (34.2%). It is followed by low-medium income

households (29.6%) and medium-high income households (25.8%). Only 10.4% of the

household is categorized as low income household. In terms of the presence of children

younger than 18 years old in the household, around 74.6% of these households have no

young children. The rest of the households are split into 11.5%, 7.4%, and 6.6% for

119

having children younger than 6 years old, 6-11 years old and 12-17 years old,

respectively. These results indicate that, although the percentage of two-worker

households is lower than zero-worker household, their average income is relatively

high.

As mentioned, the focus of this study is on joint activities. Nevertheless, the frequency

of activities that are conducted independently, which is termed as “Indep” in short, is

also revealed here, in a broader context involving 18,037 households. For the next two

tables, we will see how each activity is distributed across independent and joint

activities during a weekday, Saturday and Sunday. An independent and joint activity

frequency of each activity type is presented in Table 6.1. The predominant activity

pursued independently is one-store-shopping on Saturday, which accounts for 39.0%.

During weekday, one-store-shopping is the dominant (30.3%) independent activity. As

for joint activities on Sunday, social is the most common purpose (42.2%) and it is

followed by leisure (27.3%).

Table 6.2 shows the mean duration (in minutes) of activities. Regarding the mean

duration, joint-social is the activity that has the longest duration, on all days, which

takes 212 minutes (3 hours 32 minutes) on Saturday, 184 minutes (around 3 hours) on

Sunday, and 172 minutes (2 hours 52 minutes) on weekday. Similar to the joint

category, in the independent category, social also has the longest average duration, on

all days, which account for 159 minutes (2 hours 39 minutes) on Saturday, 148 minutes

(2 hours 28 minutes) on Sunday, and 132 minutes (2 hours 12 minutes) on weekday.

Figure 6.2 shows start time profiles of each activity (joint and independent) for every

30 minutes starting from 5 am to 9 pm. Among other activities, bring/get activities

show the highest frequency during 8.00 – 8.30 am. This activity is also done relatively

frequently jointly during midday (13.00 – 13.30 pm) and afternoon (15.00 – 15.30 pm).

Shopping to 1 store is the second activity in terms of frequency and is most often

conducted during 10.00 – 10.30 am and 11.00 – 11.30 am. Leisure activities have a

clear peak in the late afternoon and evening.

6.4 VARIABLE SPECIFICATION

Table 6.3 portrays the condition variables that were used as input to the algorithm for

induction of both decision trees. It is worth noting that some variables are related to

household level and person level, while others are defined at schedule level and activity level. Note that in this stage of the scheduling process, work, business, and other

mandatory activities are known. However, only condition information related to the

work activity is fully used for condition variables, such as number of work activities,

120

duration of each work activity, mode of traveling to work place, and total time engaged

in a work activity. Continuous condition variables, such as duration, are discretisized

by using an equal-frequency interval method which divides a continuous range into n

parts, in which each part contains approximately the same number of cases.

There are two sets of household-level attributes. The first set includes household

composition, the presence of young children in a household, day of the week, socio-

economic class, and the number of cars in the household. Given the selection of two-

heads households, household composition refers to only three household types:

‘double-head, one-worker’, ‘double-head, two-workers’ and ‘double head, no-workers’.

The presence of young children in a household only considers the children younger

than 18 years old. If children over 18 years old are in a household, they are considered

as adult, not children any longer. Income is classified as follows: low (< €16,250), low-

mid (€ 16,250-€ 23,750), mid-high (> € 23,750 - € 38,750) and high (> € 38,750).

The second set of household-level variables relates to urban density and several aspects

of accessibility of locations given the home location of the household. In terms of

accessibility, 8 variables are included: (1) daily goods sector: number of employees

within 3.1 km, (2) non-daily goods sector: number of employees within 4.4 km, (3) all

sectors: number of employees within 4.4 km, (4) size of population within 3.1 km, (5)

daily goods sector: distance within which 160 employees work, (6) non-daily goods

sector: distance within which 260 employees work, (7) all sectors: distance within

which 4500 employees work, and (8) distance within which 5200 people live.

Note that attributes related to individuals can be incorporated only if they are specified

explicitly for the male and female heads. Thus, individual attributes will be tied

together with gender. They include work status and age. The work status attribute of

the male and female heads indicates whether the person has no work, part-time work,

or full-time work. Those who work more than 32 hours per week are considered as

full-time worker. Age and possession of driving license by the householders are also

defined as person-level attributes.

The following attributes are defined on the schedule level. The number of work

activities conducted by male and female is known and we limit it to maximally 2 work

episodes per person per day. The total number of work episodes across male and

female heads is included as well, which then has a maximum of 4 work episodes.

Duration of work activity conducted by male or female and the total work duration

across male and female heads in a household are also used as condition variables.

Transport mode used for the trip to the work place by male or female worker is also

used as condition variable.

121

TABLE 6.3 Definitions of Condition Variables

Variable Description

Household-level

Household socio-demographics Household composition (Com) double-one-worker, double-two-worker, double-no-worker

1,2

Presence of young children (Child) no children, <6 years, 6-11 years, 12-17 years1,2

Day of the week (Day) Mon, Tue, Wed, Thu, Fri, Sat, Sun1,2

Socio-economic class (SEC) Low, Low-Mid, Mid-High, High1,2

Number of cars in household (ncar) 0, 1, 2+1,2

Land use-Accessibility measures Urban density (Urb) Population density of the residence location

1,2

Accessibility 1 (nEmp1) Daily goods sector: # employees within 3.1 km1,2

Accessibility 2 (nEmp2) Non-daily goods sector: # employees within 4.4 km1,2

Accessibility 3 (nEmp3) All sectors: # employees within 4.4 km1,2

Accessibility 4 (SizePop) Size of population within 3.1 km1,2

Accessibility 5 (Dist1) Daily goods sector: distance within 160 employees work1,2

Accessibility 6 (Dist2) Non-daily-goods sector: distance within 260 employees work1,2

Accessibility 7 (Dist3) All sectors: distance within which 4500 employees work1,2

Accessibility 8 (Dist4) Distance within which 5200 people live1,2

Individual-level Work status (wstatM/F) non-worker, part-time, full-time

1,2

Age (AgeM/F) <35 years, 35-54, 55-64, 65-74, 75+ years1,2

Driving license possession (driveM/F) 1 if male/female has driving license and 0 otherwise1,2

Schedule-level

Number of work episodes (nworkM/F/HH) maximally 2 work episodes of male, female, and household

1,2

Duration of work activity (wdurM/F/HH) total work duration of male, female, and household1,2

Mode to work place (modeM/F/HH) aggregation mode to work by male, female, and household1,2

Available time for non-work activity time available for male-female to do non-work activity1,2

(avaT1-T6)

Available time car for non-work activity car available for male-female to do non-work activity1,2

(time1C-6C)

Work is on schedule (yworkM/F) work is on schedule of male or female worker1,2

Business is on schedule (ybusM/F) business is on schedule of male or female worker1,2

Other mandatory is on schedule (yothM/F) other mandatory (school) is on schedule of male or female1,2

Activity-level

# each activity j (j= 1…7) # each joint activity j currently done in household1,2

(nbr,nsh1,nshn,nser,nsoc,nlei,ntou)

Duration of activity concerned j (j= 1…7) duration of each activity j currently done in household1,2

(dur-br/sh1/shn/ser/soc/lei/tou)

Total duration across the 7 activity types total duration across the 7 activity types currently done in

(durtot) household1,2

Activity type (acty) activity type considered (7 activities) 1,2

Duration of activity (dur) duration of the activity considered2

Note: 1 Alternative classification used for duration model (model 1)

2 Alternative classification used for start time model (model 2)

122

In case multiple work episodes are included in a schedule, we aggregated the mode by

arranging modes in a hierarchy as follows (in order of priority): (1) Public Transport

(2) Car Passenger (3) Car Driver and (4) Slow (Bike and Walk). Additionally, the

available time in the current schedule to do non-work activities by both male and

female is represented. In order to identify how much time is available for these

activities at different times of the day we segmented the period from 8 am to 8 pm into

6 time spans of 2 hours. For each 2 hour-period we calculated the available time in the

schedule and classified this time into 5 categories, where zero means no time left for

doing a non-work activity and the remaining categories identify how much time is left

for doing a non-work activity in multitudes of half an hour. These variables are

computed for male and female together and, thus, indicate the time available to both.

In addition to available time, the time a car is available for a non-work activity is also

considered as a condition variable. Hereby, the number of cars available and the

transport mode(s) used for work activities, if any, of the two household heads are taken

into account. Time is segmented in the same way as before. Note that zero means that

either car is not available due to a work activity or because no car is available in the

household. The next variables indicate whether or not work, business and other

mandatory activities (such as go to school) are conducted on the given day by male or

female.

The subsequent variables are activity-level variables. The first variable at this level

indicates for each joint task and joint non-task activity category the number of

activities that, at the moment of the duration and start-time decisions, are included in

the schedule, as a consequence of previous activity selection decisions for joint

activities. Thus, these variables are dynamic and take into account the assumed priority

order of activities. At the time of decisions for the first joint activity, no other joint

activities are included in the current schedule. Therefore, all these variables are zero in

the beginning. For decisions of a second joint activity, the result of decisions related to

a first optional joint activity (i.e., a bring/get activity) is known and if this implied the

insertion of a (bring-get) activity the corresponding variable has a value of one, and so

on. In sum, this set of variables indicates for each current decision the current schedule

state as a result of previous activity selection decisions. The next series of variables at

this level indicates the total duration of each (joint) activity that, at the moment of the

decision, is included in the schedule as a consequence of previous activity selection

decisions. The way of doing it is the same as for the previous attribute. The following

variable is the total duration of activities across the 7 activity category that is currently

included in the schedule (of both heads) at the moment the decision is made. The for-

last variable encodes the activity type that is considered in the current duration/start

time decision. This variable has seven levels corresponding to the seven activity

123

categories (joint task and joint non-task activities) considered in the model. The last

variable is the duration of the activity for which a decision is made. Note that, in the

sequential process, a duration decision is made before a start time decision. Hence, at

the moment a start time decision is made, the duration is known. The result of the

(previous) duration decision (as observed) is added as conditional information for the

(next) start time decision for the same activity. This means that the duration variable at

activity-level is only included as a condition in the start time model. All other variables

are used for both models.

As a result, a total of 63 condition variables were defined for the duration model as

indicated by superscript 1 in Table 6.1. For the start time model, the condition variables

are actually almost the same. The difference merely concerns the additional duration

variable at the activity-level as indicated by superscript 2 in Table 6.1. Hence, there are

64 condition variables for the start time model.

6.5 METHODS


ALBATROSS is a rule-based multi-agent system of activity-travel behavior. It consists

of a large number of rules (IF…. THEN…) for each choice facet, indicating the

choices made by individuals dependent on conditions in terms of socio-demographic

characteristics and other context variables. These rules are extracted from activity-

travel diary data using a tree induction method. Thus, these induction methods identify

the rules that describe which choices are made under which conditions. The basic

algorithm used in ALBATROSS is a CHAID-based tree induction method which

generates non-binary trees. The basic algorithm is appropriate for a categorical or

nominal action variable (also called response or decision variable), as implied by the

chi-square test statistic that is used as split criterion. However, because in this case, we

are not dealing with a categorical or nominal action variable but rather with continuous

action variables (duration and start time), the chi-square test statistic cannot be used.

Therefore, we use the F-statistic for timing and duration decision trees instead. The tree

induction algorithm, however, apart from that, remains the same. It generates a

decision tree by splitting the condition space on one condition variable at a time into

two or more subsets repeatedly, beginning with the entire data set. The split that

maximizes a significance value of the F-test across condition variables is used for

splitting if the split is significant. The process is repeated for each newly created group

until no more significant splits are found. In order to develop the duration and timing

decision tree, 75% of the cases were used for training and the remaining cases were

used for validation. Generally, in deriving ALBATROSS decision models, the

124

observed choice (in this case, duration or start time) and condition variables (in this

case the ones listed in Table 6.3) are extracted from the diary data for each observation

(in this case, a joint activity).







of tree induction. The method was developed for discrete decision trees, but can be

used with minor adjustment also for a continuous decision tree. The principle of the

proposed method is straightforward. After having derived a rule-based model from

training data, the model is used to predict for each condition variable the impacts on

the action variable. The model is applied to the training set as many times as there are

levels of the condition variable considered. In each run, each training case is assumed

to take on the level of the condition variable considered in that run. The mean and

standard deviation of predictions (start time or duration) across training cases under

that setting are recorded. Repeating this process yields a table with for each level of the

condition variable a distribution of the action variable defined by a mean and standard

deviation. The impact of the condition variable is then measured as the F-statistic for

this table.

6.6 RESULTS

In total 6,526 observations of joint activities can be derived from the data set. Across

these observations, the mean duration is 121 minutes (with a standard deviation of 120

minutes) and the mean start time is 2.08 pm (with standard deviation of 207 minutes).

As said, 75% of the cases were used for training and the remaining cases were used for

validation. Given a minimum group size of 75 cases at leaf nodes and a 5% alpha level,

the trees generated by CHAID consists of 17 and 31 leaf nodes (decision rules) for

duration and start time models, respectively.

The duration model consists of 17 decision rules with an F-statistic value of 74.57. The

S value, which stands for the average number of minutes mispredicted by the model

(i.e., standard error), shows a slight improvement from 120.2 minutes (null model) to

107.9 minutes (decision tree model). The start-time model consists of 31 decision rules

125

with an F-statistic value of 76.55. The S value shows a more substantial improvement

compared to the duration model, namely from 206.9 minutes (null model) to 170.3

minutes (decision tree model). The results of the S value on the validation set indicate

that virtually no overfitting occurred, i.e. S = 104.0 minutes and S = 178.8 minutes for

duration and start time models respectively. In case of the duration model the average

misprediction is even slightly lower.

As explained in Section 6.5.2, the impact table analysis shows which condition

variables have an influence on the decision choice of the individual/household and the

size of the influence. As it appears, of the 63 condition variables, 7 variables have an

influence on the activity duration of joint activities. Activity type is by far the most

important variable for the duration decision, as it is the first splitter and has the highest

F-statistic (F = 1439.8). It is followed by the total duration of joint activities included

in the current schedule of the household (F = 88.8). Specifically the time available

during the afternoon off-peak period (12 – 2 pm) has an impact as well (F = 45.3).

Several individual and household attributes also have an influence. These include age

(male: F = 7.4 and female: F = 5.16) and car possession in household (F = 5.9). Finally,

spatial variables and, specifically, urban density also impacts the duration choice (F =

1.52).

For the start-time model, of the 64 condition variables, 16 variables have an influence

on the start-time choice for joint activities. As in the duration model, activity type also

has a big influence in the start-time model (F = 455). Nevertheless, the available time

to do a non-work activity during morning off-peak period is by far the most influential

variable in this model as indicated by the high F-statistic value (F = 1576.4).

Subsequently, the dynamic variable indicating that a particular joint activity is included

in the current schedule of the household when the decision is made has an influence in

the start-time model (joint-leisure: F = 75.9 and joint-touring: F = 1.1). Duration of

social activities currently done in the household (F = 59.6) also influence the start time

decisions. Day of the week is a next variable that influences the start-time decision for

doing joint activities (F = 37.4). Accordingly, the total duration of joint activities

currently included in the schedule of the household (F = 29.3) and the duration of the

activity currently considered (F = 22.1) have an influence in the model, which suggests

that there is an influence of duration decisions on start time decisions. On the other

hand, individual and household attributes influence the start time decisions as well,

such as household income (F = 4.2), household type (F = 2.6), age of male (F = 1.3),

presence of young children (F = 1.05), and car ownership of the household (F = 0.75)

and accessibility measures (accessibility 7: F = 0.8 and accessibility 1: F = 1.8).

Availability time in the afternoon to do a non-work activity shows an impact as well (F

= 2.4).

126

TABLE 6.4 Duration Tree Model

Rule # Urb AgeM AgeF ncar Ava T3 durtot acty Average Duration St. Dev

1 - - - - - - 1 35,25 50,63

2 - - - - 0-1 0 2,4,7 47,26 43,87

3 - 0-1 - - 2-4 0 2,4,7 82,95 89,34

4 - 2 0-1 - 2-4 0 2,4,7 144,45 216,77

5 - 2 2-4 - 2-4 0 2,4,7 85,12 101,59

6 - 3-4 - - 2-4 0 2,4,7 74,52 92,20

7 - - 0 - - 1-4 2,4,7 88,13 107,09

8 - - 1-4 - - 1-4 2,4,7 56,75 81,44

9 - - - - - - 3 105,23 75,28

10 - - - - 0-3 0-1 5 157,08 80,99

11 - - - - 4 0-1 5 200,49 129,66

12 - - - 0-1 - 2-4 5 144,15 109,77

13 - - - 2 - 2-4 5 174,74 114,72

14 0-2 - - 0-1 - 0-1 6 171,60 155,98

15 3-4 - - 0-1 - 0-1 6 133,99 113,04

16 - - - 2 - 0-1 6 179,47 158,88

17 - - - - - 2-4 6 120,42 86,86

To interpret these results, the underlying decision trees are shown in Table 6.4 and

Table 6.5 below. To explain the table format, we describe one arbitrary decision rule as

an example for each model. For example, Rule #9 of the decision tree for duration says

“IF the activity considered is shopping to multiple stores, THEN the average duration

is 105 minutes (standard deviation is 75 minutes)” and Rule #4 of the start time

decision tree “IF there is no time available in the morning off-peak period AND the

available time at 2 – 4 pm is 1.5 – 2 hours AND the activity considered is a (joint)

social, leisure or touring activity, THEN the average start time is 17:46 hour (1067

minutes) (standard deviation is 152 minutes)”.

Several relationships are revealed. First, for duration choice, the tree reveals that the

influence of activity type is consistent with the patterns that we saw in the descriptive

analysis and, hence, is as expected. When the schedule already includes a joint activity

the duration of a next activity becomes shorter. Furthermore, the duration of a joint

social activity increases with time available for non-work activities. These patterns

indicate that particularly time-budget and travel–time considerations influence duration

decisions in ways as could be expected. Presence of 2 or more cars in the household

tends to increase the duration particularly for social and leisure activities. Leisure

127

activity takes longer duration in high urban density areas than in low urban density

areas.

As for the start-time model, the underlying decision tree (Table 6.5) indicates that joint

activities occur only very rarely in schedules where one of the household heads have a

work activity between 10 am and noon and when this occurs the start time is

substantially earlier. The effects of activity type are as expected and in line with the

patterns that emerged from the descriptive analysis. The presence of a previously

scheduled joint activity leads to a later start time for a next joint activity. Furthermore,

Sunday leads to earlier start times, if no joint activity has been scheduled yet. The

presences of young children in the household influences household heads decision to

start earlier performing a joint bring/get activity. Presence of two or more cars in a

household tends to leads to a later start time. Higher income households tend to start

later performing a joint social activity, especially, on Monday and Sunday when longer

time is available in the morning. In non-worker households, having already scheduled a

joint activity leads to a choice of an earlier start time for a next joint activity. In sum,

these patterns suggest that scheduling considerations play a dominant role in start time

decisions, whereas other factors have relatively small impacts.


This paper presented the development and testing of models for duration and timing

choice for joint household-task and non-task activities as part of a full-fledged activity-

travel scheduling model for a refined ALBATROSS. Focusing on joint participation

between the two household heads in non-work activities, a rule-based duration and start

time model using a CHAID-based algorithm was derived from activity-trip diary data.

Decision tree results indicated that there were 17 and 31 condition-action rules derived

for the duration model and start time model, respectively. The improvement in S-value

(a measure of prediction accuracy) relative to a null model as well as an F-statistic

indicates that there is a moderately strong association between condition variables at

household, individual, activity and schedule level, on the one hand, and the decision,

on the other. The S-value shows a more substantial improvement in the start-time

model compared to the duration model. The results of the S-value on the validation set

indicate that virtually no overfitting occurred, i.e., that the rules are generalizable to

unseen cases. As an impact analysis of the decision-tree models showed, activity type

has a substantial influence in both the duration and start-time model. Within activity

categories, duration decisions seem to be primarily driven by time-budget and travel-

time considerations.

128

TABLE 6.5 Start-Time Tree Model

Rule

# Com Child Day SEC AgeM nEmp1 Dist3 ncar

Ava

T2

Ava

T4 nlei dursoc durtot acty dur

Ave

Start-

Time St. Dev

1 - 0 - - - - - - 0 - - - - 1,4,2,3 - 1005,33 164,27

2 - 1,2,3 - - - - - - 0 - - - - 1,4,2,3 - 890,90 247,09

3 - - - - - - - 0 0-2 - - - 5,6,7 - 1172,34 96,88

4 - - - - - - - 0 3-4 - - - 5,6,7 - 1066,82 152,09

5 2,3 0,2,6 - - - - - 1-4 - 0 - - 1,6 0-2 824,88 229,28

6 4 0,2,6 - - - - - 1-4 - 0 - - 1,6 0-2 737,25 204,25

7 - 0,2,6 - - - - - 1-4 - 0 - - 1,6 3 708,90 190,75

8 - 0,2,6 - - - - - 1-4 - 0 - - 1,6 4 782,70 223,36

9 - 1,3,4,5 - - - - - 1-4 - 0 - - 1,6 - 854,06 219,77

10 - - - - - - - 1-4 - 1-2 - - 1,6 - 989,69 176,59

11 - - - 0-2 - - - 1-4 - - - 0 2,3,4 0 797,87 169,68

12 - - - 3-4 - - - 1-4 - - - 0 2,3,4 0 733,39 161,70

13 - - 0-1 - - - - 1-4 - - - 0 2,3,4 1-4 713,71 142,21

14 - - 2-3 - - 0-2 - 1-4 - - - 0 2,3,4 1-4 757,13 142,76

15 - - 2-3 - - 3-5 0-1 1-4 - - - 0 2,3,4 1-4 710,38 133,48

16 - - 2-3 - - 3-5 2 1-4 - - - 0 2,3,4 1-4 746,14 135,52

17 2,4 - - - - - - 1-4 - - - 1-4 2,3,4 - 783,54 137,51

18 3 - - - - - - 1-4 - - - 1-4 2,3,4 - 841,17 126,03

19 0,6 0-1 - - - - 1-4 - - 0-1 - 5 - 813,79 164,09

20 0,6 2-3 - 0-1 - - 1-4 - - 0-1 - 5 - 884,32 157,52

21 0,6 2-3 - 2-5 - - 1-4 - - 0-1 - 5 - 827,05 171,48

22 1,3,2 - - - - 1-4 - - 0-1 - 5 - 875,26 211,28

23 4,5 - - - - 1-4 - - 0-1 - 5 0-3 892,54 195,80

24 4,5 0-1 - - - 1-4 - - 0-1 - 5 4 1035,17 186,10

25 4,5 2-4 - - - 1-4 - - 0-1 - 5 4 949,53 227,60

26 - - - - 1-4 - - 2-4 - 5 - 1016,75 155,42

27 - - - - 1-4 - - - 0 7 0-1 856,75 151,69

28 - - - - 1-4 - - - 0 7 2-4 750,47 125,54

29 - - - - 1-4 - - - 1-4 7 0-1 900,86 176,61

30 - - - - 1-4 - - - 1-4 7 0-1 964,04 166,45

31 - - - - 1-4 - - - 1-4 7 2-4 840,67 131,46

129

There is, however, substantial heterogeneity across cases, which is only partly related

to day of the week and socio-demographic variables. More than in case of duration

choice, start-time decisions are influenced by scheduling opportunities. In particular,

the available time during the late morning in the schedules of the two heads, given their

work activities (if any) appears to be crucial. Some individual and household

characteristics such as household type, presence of children, income, car availability,

and age of male show a big influence on start time decisions. In both models, car

availability and age of male play also a modest role.

Finally, the results indicate that there is a relatively strong influence of duration choice

on start-time decisions. This means that a state-dependent sequential decision process,

as used in ALBATROSS, works well. Consistently with Habib et. al (2008), we find

that joint participation of household members in activities tends to lead to a longer

activity duration. Furthermore, joint activities by household heads tend to start earlier

in the day especially when this involves a bring/get activity (e.g., bringing a child to

school). These results suggest that the decision tree induction method can also capture

the joint decision making in household.

REFERENCES

Anggraini, R., Arentze, T.A. and Timmermans, H.J.P. (2007), “Modeling Car

Allocation Decisions in Automobile Deficient Households”. In: Proceedings ETC 2007 Conference, Noordwijk, The Netherlands.

Anggraini, R., Arentze, T.A. and Timmermans, H.J.P. (2008), “Modeling Joint

Activity Participation and Household Task Allocation”. In: Proceedings AATT Conference 2008, Athens, Greece.



Arentze, T.A. and Timmermans, H.J.P. (2003), “Measuring Impacts of Condition


Empirical Illustration”, Geographical Analysis, 35, pp.24-45.





130

Bhat, C. (1996), “A Hazard-based Duration Model of Shopping Activity with

NonParametric Baseline Specification and NonParametric Control for Unobserved

Heterogeneity”. Transportation Research Part B, Vol. 30, No. 3, 1996, pp. 189-207.

Bhat, C.R., and Steed, J.L. (2002), “A Continuous-time Model of Departure Time

Choice for Urban Shopping Trips”. Transportation Research Part B, 36, pp.207-

224.

Bhat, C.R. and Pendyala, R.M. (2005), “Modeling Intra-Household Interactions and

Group Decision-Making”. Transportation, Vol. 32, No.5, pp. 443-448.

Gliebe, J.P. and Koppelman, F.S. (2002), “A Model of Joint Activity Participation”.

Transportation, 29, , pp. 49-72.

Golob, T.F. and McNally, M. (1997), “A Model of Household Interactions in Activity

Participation and the Derived Demand for Travel”. Transportation Research B, Vol.

31, pp. 177-194.

Goulias, K.G. (2000), “Companionship and Altruism in Daily Activity Time

Allocation and Ttravel by Men and Women in the Same Households”. In:

Proceeding TRB 2000, Washington, D.C., US.

Habib, K.M.N., Carrasco, J.A., and Miller, E.J. (2008), “The Social Context of

Activity-Scheduling: A Discrete-Continuous Model of the Relationship between

“With Whom”and Episode Start Time and Duration”. In: Proceeding TRB 2008,

Washington D.C, USA.

Niemeier, D.A., and Morita, J.G. (1996), “Duration of Trip-making Activities by Men

and Women: A Survival Analysis”. Transportation, 23, pp. 353-371.

Pendyala, R.M., and Bhat, C.R. (2004), “An Exploration of the Relationship between

Timing and Duration of Maintenance Activities”. Transportation, 31, pp. 429-456.

Steed, J.L., and Bhat, C.R. (2000), “On Modeling Departure Time Choice for Home-

based Social/Recreational and Shopping Trips”. Transportation Research Record, 1706, pp.152-159.

Vovsha, P., and Bradley, M. (2004), “A Hybrid Discrete Choice Departure Time and

Duration for Scheduling Travel Tours”. In: Proceeding TRB 2004, Washington

D.C., USA.

131

Chapter 7

HOUSEHOLD LOCATION CHOICE MODELS FOR

INDEPENDENT AND JOINT NON-WORK

ACTIVITIES

Anggraini, R., Arentze, T.A., and Timmermans, H.J.P., 2009. Paper presented at XIII Euro Working Group on Transportation, Padua, Italy.

ABSTRACT Modeling location (destination) choice has a long history in transportation research and in disciplines such as urban planning, geography and regional science. The vast majority of this literature is concerned with the problem of how individuals choose a destination. However, trips may involve household and hence it could be argued that conceptualizing destination choice as a problem of individual choice behavior may be inadequate, especially in case of joint household activities. Therefore, we develop the household location choice model taking into account the heads of household (male-female) independent and joint activity, in particular in non-work activities. The model is incorporated in the activity-scheduling model of ALBATROSS – an activity-based model of travel demand, predicting travel demands in an activity-based micro-simulation system. In this paper, we examine the location choice model for independent and joint activity participation of the household heads based on the concept of detour time. To deal with a large set of attribute variables and account for non-linear relationships between the variables, a CHAID decision tree induction method is used to derive a decision tree model. There are two models incorporated in this study: (i) determining whether the activity is conducted on the same location as the previous activity, the same location as the next activity, or at some other location and, if at some other location, (ii) determining the location in terms of a combination of size class and distance class of the postcode area. The performance of the location choice model for joint activity turns out to be superior to that of the independent activity, in particular for the first model. The tendency of male and female performing multiple activities at the same location is higher when travelling alone than travelling together.

132

7.1 INTRODUCTION

Activity-based modeling of travel demand has gradually shifted from academic

exploratory models to full-fledged operational models which are now being applied in

transportation planning practice. Activity-based approaches have augmented our

conceptualization of travel behavior by including mutual dependencies, for example

between the activity-travel schedules of different household members. Although the

necessity to include multiple households members have been identified from the very

beginning of activity-based analysis, the topic of household decisions has still received

relatively scant attention.

According to Gliebe and Koppelman (2002), employment commitments and childcare

responsibilities have significant effects on trade-offs between joint and independent

activities. Chandraskharan and Goulias (1999) found that joint activities are

appreciably affected by household size, and age of the household members. In addition,

car ownership levels have been observed to be positively correlated with individuals

following independent paths over the course of a day. Likewise, Golob and McNally

(1997) also modeled the interactions between household heads. Although not

considering joint activity participation, they found the relationships between time

allocated to work, maintenance, and discretionary activity, and to the travel generated

by each activity.

It has been recognized that joint activity participation modeling is one of the more

complicated behavioral patterns in activity-based travel modeling. In particular in

multi-persons households, definitions of joint activity vary. Gliebe and Koppelman

(2002) identify activities as being joint-in-purpose, joint-in-location, joint-in-time, or

some fuzzy subset along those dimensions. In this study, we focus on household

decision making with regard to activity location choice, in particular for the two heads

of household. We, further, intend to analyze the location decisions for independent and

joint activities.

An understanding of the factors that influence the choice of location can contribute to

more effective land-use and zoning policies. To better understand activity choice

location in the context of a complete activity schedule for a day, the concept of detour time is applied. This concept is used in ALBATROSS (Arentze and Timmermans,

2007). Different from any other concept that considers the travel distance from home

or non-home to a particular location, detour time considers relative location to the

previous and next activity. The detour time of a candidate location for an activity is

defined as the extra travel time required to implement the activity in the context of the

current activity schedule. This concept is very useful to build trip chains and to

133

simulate the emergence of feasible activity-travel schedules that take space-time

constraints into account.

This paper reports the results of modeling activity location decisions for independent

and joint activities from the perspective of household decision making. The results of

this paper should be viewed as a step in refining ALBATROSS (Arentze and

Timmermans, 2000, 2004, and 2005), but the results are also relevant in their own right.

The paper is organized as follows: the subsequent two sections present a synopsis of

the location decision model in current models. This is followed by an outline of the

approach, adopted in this study, a description of the data, a discussion of empirical

results, and a summary of major conclusions.

7.2 LOCATION DECISIONS IN THE EXISTING MODEL

Arentze and Timmermans (2007) developed a location choice model in which location

choice decisions are made in a priority order of activities and within activities in the

order in which the activities occur in the schedule. The model uses the concept of

detour time. The detour time of a candidate location for an activity is defined as the

extra travel time required to implement the activity in the context of the current activity

schedule. Let xi be the origin of the trip to a candidate location i and yi be the

destination location of the trip from i. Thus, the detour time related to location i is

defined as follows:

dti = dt(xi, i) + dt(i, yi) – dt(xi, yi) [7.1]

where, dt(i, j) is the travel time between locations i and j.

In order to estimate the detour time for a particular location, the origin and destination

location should be known. The extent to which this information is available at the

moment of the decision depends on the activity type. ALBATROSS considers several

activity types in general: work, work-related such as business and school, and non-

work activities such as escorting, shopping (daily and non-daily), service-related, social,

leisure, and touring. Location decisions for those activities are made at different

moments in the sequential scheduling process.

As a consequence of the sequential decision process, the available information for

location decisions is limited. At the moment the location decisions are made, the

following information is available. Generating a work activity is the top priority in the

ALBATROSS scheduling process. It includes information about the number of work

134

episodes, start time and duration, and the location and mode choice of the work tour.

Furthermore, it takes work-related activities (business and other mandatory) into

account. The subsequent process generates non-work activities, such as escorting,

shopping, service, social, leisure and touring. The first three activities are considered to

be a household task, while the last three activities are non-household activities (even

though they may be conducted jointly). Household tasks need commitment from

household heads, such as who delivers the children to school. As a result, a task

allocation decision is needed to be made. Joint participation in activities also needs

trade-offs between adult heads of household. This can be applied to either household

task or non-household activities. Having scheduled all activities in a priority order,

duration, timing and trip-chaining decisions are established. Subsequently, location

decisions are simulated.

In case of work and work-related activities, the non-work activities are not yet

scheduled and the location of the next (work or work-related) activity, if it is other than

home, is still unknown. In such cases the model assumes that the next location is the

home location. On the other hand, in case of a non-work activity, the location of the

next activity is unknown if it is a non-work activity of a same or lower priority for the

same reason. Again, in such cases the model assumes that the next location is the home,

work location or the location of a higher priority activity (what comes first in the

schedule). Although these assumptions are simplifications of reality, it is to be

expected that they will not seriously affect the performance of the model. At least, the

model is able to take into account the location relative to home and to a previous/next

location in every location decision of a sequential priority based scheduling process

and consequently it should better cover interdependencies in these choices. A space-

time prism is calculated for each location decision defining the set of locations that are

within reach given the space-time constraints imposed by the interaction between the

environment and the schedule.

Having identified the origin and destination for the activity considered, the model

determines the locations, based on postcode areas, which are within reach, i.e. within

the prism. Since transport mode is unknown yet, the model calculates a preliminary

prism based on the fastest transport mode available in the time slot under concern. For

instance, if the person is a driver and the household owns a car(s), and the car is not

used for a work activity of another household member in the same time slot, then the

fastest travel mode is the car. In case there is no car in the household, the fastest travel

mode is public transport in most cases. Having identified the fastest transport mode,

the shortest travel time across the road network is determined. Furthermore, the

(minimum) duration of the activity, the time window and opening hours of required

facilities at destination, are taken into consideration in the model.

135

Time window is defined by the earliest possible departure time from the origin and the

latest possible arrival time at the destination. All in all, the resulting set of locations

meet an exhaustive set of space-time and resource availability constraints.

This conceptualization is similar to the current version of ALBATROSS. However, the

existing model only applies to the context of person-level decision making

(independent activities) while in this study we expand it to cover as well household-

level decision making (joint activities).

Figure 7.1 shows the procedure assumed in the current model. The model basically has

two sub-models, both represented by decision rules modeled as a Decision Tree (DT).

The first tree (DT1) determines whether the activity is conducted at the same location

as the previous activity (in the schedule), the same location of the next activity or at

some other location. In the second tree (DT2), a choice from the set of locations

available in the prism is made. It determines the location in terms of a combination of

size class and distance class of the postcode area. The size class depends on the

activity type under concern and the size of available facilities at the activity location

(i.e., number of employees in the relevant sector). Size is classified into 5 categories

START

STOP

Same as PREVIOUS Size by distance

band of activity i

i = 1 i > I i = i + 1

Select location

from band

i = index of activity episode in order of occurrence in the schedule (i = 1….I)

DT1

DT2

Relative location of

activity i

Same as NEXT

OTHER

FIGURE 7.1 The Process Model for Predicting Location of Non-

Work Activities

136

based on employment in the relevant sector for the activity considered (e.g., the retail

sector in case of a shopping activity) and distance is classified in terms of a detour

travel time (by car) also into 5 categories. Hence, the choice alternatives for the DT

consist of 25 location classes. Given the prism and the choice of a location class, the

exact postcode area is determined by Monte Carlo simulation, if there are multiple

locations of that class within the prism. Furthermore, a distance decay factor is taken

into account so that locations with longer detour time within the class have smaller

probabilities of being selected.

7.3 HOUSEHOLD LOCATION DECISIONS (JOINT ACTIVITY)

The above conceptualization is implemented in the current version of ALBATROSS.

However, the existing model only applies to the context of person-level decision

making (independent activities) while in this study we expand it to cover as well

household-level decision making (joint activities). This will be the subject of this

section.

Since we consider the activity that is performed independently or jointly by male-

female heads, each DT model is applied to those. Independent activities are treated

practically the same as in the current version of ALBATROSS, where one of the heads

conducts a particular non-work activity singly. In contrast, the joint activity model is

applied when both adult heads of household perform the same activity at the same start

time (+/- 15 minutes), the same duration (+/- 15 minutes) and the same location (based

on postcode area). The space-time prisms are determined for the two persons jointly.

That is to say, a location is within reach only if it is within reach for both persons given

their schedule context settings.

As in the current model, each independent and joint activity model consists of two sub

models. The first model consists of 3 choice alternatives: the activity is conducted at

the same location as the previous location, the activity is conducted at the same

location as the next location or other. In case of the latter, it proceeds to the second

model that consists of 25 choices as explained above. As a result, four DTs are

generated. The DTs concerned with the joint location decisions are defined on

household level and, in that sense, represent rules of joint decisions between the

household heads.

Fundamentally, the attributes (i.e. condition variables) used in independent and joint

activity models are equivalent, i.e. relate to the same dimensions. There are four

137

categories, person-level, household-level, schedule-level, and activity-level attributes.

For more details about the condition variables, see Section 7.5.

Although the dimensions of the variables are the same, some operational decisions

need to be made for joint activities when the schedule contexts differ. In specific,

variables related to a person, schedule or the activity need to be merged to represent the

relevant conditions for the two persons together. Since a location should meet

requirements of both persons, the most restrictive requirements are taken as an

indicator. This means that a minimum (in case of resources) or maximum (in case of

demands) is used for each attribute variable across the two persons to arrive at the

corresponding attribute variable at joint-person level.

To some extent, there are additional remarks for joint activity variables, in particular,

the activity-level attributes. First of all, the start times and durations of the joint activity

considered are not necessarily exactly the same. If the time is different, the model takes

the minimum start time and duration across the persons. The previous or next activity

in the schedule may also be different between the persons. Following the above rule,

the model takes the minimum or maximum of schedule context variables. For instance,

if the distance (measured as travel time by car) of the previous activity location from

home is different the maximum is taken. As another example, if available facilities

(measured as number of employees in the relevant economic sector) at the previous

location or next location differ we take the minimum value. This is done to identify the

worst case that might be occurred.

Regarding other schedule-level attributes, the attributes consist of the number of

episodes of each particular activity and the total duration (for 10 activity categories,

ranging from work to touring). Assuming that the highest activity load is indicative for

available time, we take the maximum of these indicators. For example, if the number of

daily shopping episodes on the day of the male is 1 and that of the female is 3, then, the

number of daily shopping episodes used to describe the conditions for the choice is 3

for that particular household. A similar treatment also applies for the total duration of

each activity.

Household-level attributes of the joint activity model, obviously, are the same as that of

the independent activity model. For person-level attributes the information of male and

female are merged (age, driver license possession, work status) or become redundant

and undefined (gender). In case of joint activities, the definition of a space time prisms

takes into account the schedule settings of both persons involved. A location is within

reach only if it is in reach for both persons.

138

7.4 DATA

The data source used for empirical analysis of the new as well as the old model is

based on the 2004 Dutch National Travel Survey (MON data) which covers all of the

Netherlands. The survey is conducted on a regular basis to obtain travel and activity

information of residents in the Netherlands. The survey is a traditional trip-diary survey

and not an activity or time use survey but nevertheless includes relatively detailed data

on activities at trip destinations. The survey is conducted as a mail-out mail-back

survey. It is a household survey where diary data is collected of all household members

on a designated day. As a result, general information about household and individual

attributes such as, gender, age, vehicle ownership and driving license ownership, home

location, individual income, occupation, number of working hours per week, etc, were

collected. Diary days of individuals that do not include any trip are included as well.

Information for each trip includes start time, trip purpose, destination, activity type at

the destination, and transport mode. Situational variables and spatial geography are

also reported. All in all, this survey provides an exclusive data source to analyze

activity-travel behavior of Dutch residents.

The 2004 survey instrument was mailed to about 40,000 households and 29,221

households filled out a one-day travel/activity diary. Of these, 28,600 households fit

the criteria for being considered in ALBATROSS. The data was transformed to an

activity-diary data format for estimation purposes. The present study focus on the

location decision of household heads either in independent or joint activity. The

number of households performing an independent activity is 19,500, while there are

49,793 incidents of independent activities. A total of 5,017 two-head households

conduct a joint activity. The database contains 7,150 joint activity episodes. The data

were combined with extensive national datasets about the transport network and land-

use system. The study area is all of the Netherlands. Postcode areas are taken as the

unit of location (at a 4-digit level of which there are about 4000 in The Netherlands).

Land-use data include employment by economic sector for each postcode area. Data

about the transport system includes the road network for car and bike and zonal travel

times and travel costs for public transport modes. Finally, the database includes

opening hours of stores and parking facilities at locations.

7.5 OVERVIEW OF CONDITION AND ACTION VARIABLES

The definition of condition variables essentially depends on a particular decision under

consideration. The choice of condition variables is limited by the database that can be

used in the prediction phase. Furthermore, the choice is lead by theoretical

139

contemplations of which variables are prospectively relevant for a given decision. As

said, for this study, condition variables are related to household-level (H), person-level

(P), schedule-level (S) and activity-level (A). The choice of condition variables at

household-level and person-level are mostly used in either of the two sub-models.

At schedule level, condition variables illustrate the history of the decision process

where the model begins with an empty schedule, and then proceeds to the next

decisions that become available in the individual/household activity schedules. On the

level of the activity, the condition variables are related to specific information about

the activity concerned. In sum, condition variables are defined as input to the model

algorithm to find the rules that suit the output decision (action variables). The condition

variables that we used as input to the DT induction are portrayed in Table 7.1.

Table 7.1 shows the condition variables that we use for independent and joint activities.

The number of condition variables is nearly the same for both models, about 41

variables. However, since the activity is done together, gender is not relevant for joint

activities. Another variable that measures the maximum travel distance partner across

fixed activities in same time window is also not relevant. Hence these variables are left

out, remaining 39 variables for joint activities.

Urban density (of the residence location) is a household-level attribute that consists of

5 classes with ascending order from most densely to least densely areas. Household

composition is also a household-level attribute that consists of 5 categories: single-non-

worker, single-worker, double-single-worker, double-dual-workers, and double-non-

workers households. Nevertheless, in case of joint activities, only the three last

categories are relevant.

Age of the youngest child in a household is a household-level variable as well. It has 5

categories: no children, children less than 6 years old, children 6 – 11 years old, and

children 12 – 17 years old. Young people ≥ 18 years old are no longer considered as

children. Other household-level attributes are household income and car ownership.

Household income has 4 classes: low income (≤ € 16, 250), low-mid income (€ 16,251

– 23,750), mid-high income (€ 23,751 – 38,750) and high income (> € 38,750)

households. Car ownership accounts for the number of cars in the household and has 3

categories: no cars, 1 car, and 2 or more cars. Finally, day of the week consists of 7

days that begins from Monday to Sunday.

The following attributes are defined at the personal-level. Age of people has 5 classes:

younger than 35 years old, 35 – 54 years old, 55 – 64 years old, 65 – 74 years old and

75 years old or older.

140

TABLE 7.1 Condition Variables of Independent and Joint Activity

No Label Definition In Any Cases

1 Urb Urban size As is

2 Comp Household composition As is

3 Child Presence of young children As is

4 Day Day of the week As is

5 pAge Age As is

6 SEC Household income As is

7 Ncar Car ownership As is

8 Gend Gender Undefined – not used for joint act

9 Driver Driving license holder Is maximum (Joint is driver if at least

one person is driver)

10 Wstat Work status of Male Work status of male

11 PWstat Work status of Female Work status of female

12 Wodur Work duration Is maximum

13 Aty Activity type As is (except business and other)

14 Adur Activity duration As is (if different, take minimum)

15 Abt Activity start time As is (if different, take minimum)

16 VoisA Current activity = previous activity Is maximum priority (0=both home,

else minimum)

17 Voty Activity type of previous activity Is maximum priority (0=both home,

else minimum)

18 VoH Previous location is home Is product (1=both home, 0=otherwise)

19 Vosize Size of previous location Is minimum (worst case)

20 Vopark Parking price at previous location Is maximum (worst case)

21 Vohtt Travel time by car previous location - home Is maximum (worst case)

22 Vogord Order of municipality of previous location Is minimum (worst case)

23 NaisA Current activity = next activity Is maximum priority (0=both home,

else minimum)

24 Naty Activity type of next activity Is maximum priority (0= both home,

else minimum)

25 NaH Next location is home Is product (1=both home, 0=otherwise)

26 Nasize Size of next location Is minimum (worst case)

27 Napark Parking price at next location Is maximum (worst case)

28 Nahtt Travel time by car next location - home Is maximum (worst case)

29 Nagord Order of municipality of next location Is minimum (worst case)

30 Tavail Available time Is minimum (worst case)

31 LvoisLNa Previous location = next location Is product (1=both home, 0=otherwise)

32 Ptt Max travel distance partner across fixed activities in same

time window Undefined – not used for joint act

33 CarAv Car availability Is maximum (best case)

34 MxSizeD1 Max size of relevant sector in distance band 1 Arbitrary: take Male’s

35 MxSizeD2 Max size of relevant sector in distance band 2 Arbitrary: take Male’s 36 MxSizeD3 Max size of relevant sector in distance band 3 Arbitrary: take Male’s 37 MxSizeD4 Max size of relevant sector in distance band 4 Arbitrary: take Male’s 38 Dmin The nearest band where a location is available in prism Same for both

39 Dmax The farthest band where a location is available in prism Same for both

40 Smin The smallest location size category available in prism Same for both

41 Smax The largest location size category available in prism Same for both

141

Other personal-level attributes are driving license possession, and work status of male

and female. Driving license possession is a binary variable: yes or no. In case of joint

activities, the presence of at least one person with a driving license is indicative. The

work status attribute of the male-female heads indicates whether the person has no

work, part-time work, or full-time work. Those who work more than 32 hours per week

are considered full-time workers. In addition to these attributes, only one schedule-

level variable is identified, i.e. totals work duration of the male or female. In the

context of joint activity, the maximum work duration of either the male or female is

taken into consideration.

The subsequent attributes are defined at the activity-level (#13 - #41). There are 9 types

of activities being considered (except work activities) as possible independent activities.

These include business, bring/get, shop-1-store, shop-n-store, service, social, leisure,

touring, and other. As for joint activities, the model excludes business and other

mandatory activity meaning that 7 activities remain as possible activities for joint

participation. Activity duration and start time are continuous variables. However, they

are discretisized to four levels. Variables #16 - #22 cover information about the

previous activity setting, while variables #23 - #29 contain information about the

setting of the next activity. For instance, variable #16 states whether the current activity

is of the same type as the previous activity, and variable #23 reveals whether the

activity considered is of the same type as the next activity. In case of joint activities, if

the previous/next activity of the male/female is different (#16-17 and #23-24), the

order of activity priority is taken into consideration.

Variable #30 reveals the available time of doing non-work activities. If the available

times of the male and female differ, the minimum time available is taken into account

(for joint activities). Variable #31 contains information whether the previous location is

the same as the next location. In case of joint activities, if previous and next locations

of male and female are the same, it is noted as 1 and 0 otherwise. Variable #32 states

about the maximum travel distance of the partner across fixed activities in the same

time window, which is undefined in case of a joint activity (and not used). Variable

#33 is a binary variable indicating whether the car is available for the activity. In case

of a joint activity the maximum is taken. Variables #34-#37 reveal about the maximum

size of the relevant sector in distance bands 1-5, and this is the same for male and

female, since the space-time prisms are defined for the joint case. Variable #38 - #41

define availability characteristics of locations within the (joint) space-time prism.

Variable #38 indicates the nearest band where a location is available and variable #39

the farthest band. Similarly, regarding size, variables #40 and #41 indicate the smallest

and largest size across available locations in the prism. All in all, there are 41 condition

142

variables used for deriving the decision tree in case of independent activities, and 39

condition variables are used in case of joint activities (excluding gender and Ptt).

7.6 DECISION TREE INDUCTION AND IMPACT TABLE METHODS

ALBATROSS is a computational process model using a rule-based approach. It

consists of a large number of IF - THEN rules for each choice facet, indicating the

choices made by individuals depend on conditions in terms of socio-demographic

characteristics and other context variables. These rules are extracted from activity-

travel diary data using a tree induction method. Thus, the induction method identifies

the rules that describe which choices are made under which conditions. The basic

algorithm used in ALBATROSS is a CHAID-based tree induction method which

generates non-binary trees. The basic algorithm is appropriate for a categorical action

variable, as implied by the chi-square test statistic that is used as a split criterion. It

generates a decision tree by splitting the condition space on one condition variable at a

time into two or more subsets repeatedly, beginning with the entire data set. The split

that maximizes a significance value of the chi-square test across condition variables is

used for splitting if the split is significant. The process is repeated for each newly

created group until no more significant splits are found. In order to develop the

decision tree, 75% of the cases were used for training and the remaining cases were

used for validation. Readers interested in this topic are referred to Kass (1980).






of the tree induction. The method was developed for discrete decision trees, but can be

used with minor adjustment also for a continuous decision tree. The principle of the

proposed method is straightforward. Having derived a rule-based model from the

training data, the model is used to predict for each condition variable the impact on the

action variable. The model is applied to the training set as many times as there are

levels of the condition variable considered. In each run, each training case is assumed

to take on the level of the condition variable considered in that run. The mean and

standard deviation of predictions across training cases under that setting are recorded.

Repeating this process yields a table for each level of the condition variable a

distribution of the action variable defined by a mean and standard deviation. Further

details of this approach can be found in Arentze and Timmermans (2003).

143

7.7 DESCRIPTIVE ANALYSIS

Table 7.2 presents the percentages of performing independent activity by male/female

at the same location as the previous activity, the next activity, and at some other

location. Of 49,793 cases, the probability of performing activity at the same location as

the previous activity is 20,710 cases (41.6%) and at the same location as the next

activity is 19,292 cases (38.7%). Meanwhile the probability of conducted activity at

different location as the previous/next activity corresponding to a single stop trip from

a base location (given information available at the scheduling step), is 31,981 cases

(64.2%). In any location choice, females take higher percentage than that of the males.

Nevertheless, in particular in bring/get and daily shopping (shop-1) activities,

male/female is more likely to do multiple activities.

Table 7.3 presents the percentages of performing joint activity by male and female at

the same location as the previous activity, the next activity, and at some other location.

Of 7,150 cases, the probability of performing activity at the same location as the

previous activity is 2,058 cases (28.8%) and at the same location as the next activity is

1,746 cases (24.4%). Meanwhile the probability that the previous location is the same

as the next activity is 5,344 cases (74.7%). As can be seen, male-female tend to do

joint activities with any other joint activity (multiple activities) as well, in particular in

daily shopping (shop-1), social, leisure and touring. Touring takes the highest

percentage in case of the same as the previous location (26.2%). Daily shopping leads

in case of the same as the next location (25.8%). Additionally, social is dominantly

performed when the same as the previous and next activity location (31.7%).

TABLE 7.2 The Percentage of Performing Independent Activity at the Same

Location as Previous and/or Next Activity Current Activity Location =

Previous Activity Location

Current Activity Location

= Next Activity Location


≠ Previous/Next Location Activity

Male Female Male Female Male Female

Business 3.0 2.7 3.0 2.3 10.8 7.5

Bring/get 7.4 18.8 7.3 19.1 7.4 12.4

Shop-1 9.0 13.9 9.3 14.0 6.7 9.4

Shop-n 1.3 1.9 1.3 1.9 1.6 2.3

Service 1.7 2.1 1.8 2.2 1.7 2.2

Social 4.0 6.5 3.9 6.5 6.2 9.0

Leisure 4.7 5.9 4.6 5.5 5.7 7.3

Touring 5.6 7.3 5.8 7.7 1.8 2.4

Other 1.4 2.5 1.4 2.5 2.3 3.2

Total 38.3 61.7 38.4 61.6 44.3 55.7

# Cases 7,923 12,787 7,412 11,880 17,811 14,170

Total # cases 20,710 19,292 31,981

144

TABLE 7.3 The Percentage of Performing Joint Activity at the Same Location as

Previous and/or Next Activity

Activity Current Activity Location =

Previous Activity Location


= Next Activity Location

Current Activity Location ≠

Previous/Next Location

Bring/get 1.8 2.6 2.8

Shop-1 23.5 25.8 21.8

Shop-n 3.9 4.1 5.2

Service 3.0 2.3 5.1

Social 23.4 23.0 34.7

Leisure 18.1 17.4 19.8

Touring 26.2 24.9 10.7

# Cases 2,058 1,746 5,569

TABLE 7.4 The Percentage of Performing Independent and Joint Activity by

Available Distance and Location Size Band in Prisms

Independent Independent Distance (Travel

Time) Male Female

Joint Location

Size Male Female Joint

D1 1.1 0.9 16.9 S1 20.9 21.0 16.9

D2 2.0 1.7 23.7 S2 20.5 20.5 23.7

D3 4.4 4.0 24.2 S3 20.1 20.2 24.2

D4 12.6 11.6 21.1 S4 19.9 19.8 21.1

D5 79.8 81.7 14.1 S5 18.5 18.5 14.1

# Cases 9,075 12,884 22,905 33,728 48,627 22,905

Table 7.4 displays information about the percentages of performing independent and

joint activity by distance bands (the left side) and location size bands (the right side).

The probability of performing independent activity in distance 5 (D5) is the highest

among other distance bands, both for male (79.8%) and female (81.7%). Slightly

different from distance bands, location size shows quite similar results for each band,

in particular in independent activities. The probability of performing independent

activity where a location size is available in prisms, are 20% on average, in particular

from S1 – S3. Surprisingly, the probability of doing joint activity is the same, either

where the distance band is available in the prism or the location size band is available

in the prism.

145

7.8 RESULTS

For deriving the independent activities models, a total of 49,793 observations were

derived from the data set. About 75% of these cases were used for training and the

remaining cases were used for validation for each model. For deriving the joint activity

model, a total of 7,150 observations were available in the data set. About 75% of these

cases were used for training and the remaining cases were used for validation of the

model. Table 7.5 shows the decision tree results of the two models of each independent

and joint activity. In this section we will first discuss results for the independent

activities and next consider the joint-activities case.

In the context of independent activities models, the tree generated by CHAID consists

of 67 and 52 leaf nodes (decision rules) for the first and second model. In terms of

goodness-of-fit (hit ratio), model 1 that consists of 3 choice alternatives shows a better

performance than model 2. The hit ratio (based on a probabilistic assignment rule) of

the model compared to a null-model (a root-only decision tree) indicates a modest but

significant improvement: the hit-ratio of a null-model equals 0.484 and the hit ratio of

the tree after splitting equals 0.562. A Chi-square-based contingency coefficient of

0.489 proves that there is a moderately strong impact of the decision tree structure on

the action variable. The overall accuracy on the validation set is almost the same. It

only dropped slightly from 0.562 to 0.551 indicating no overfitting occurs.

TABLE 7.5 Results of Location Decision Tree Models

Independent Activity Joint Activity Indicator

Model 1 Model 2 Model 1 Model 2 N alts 3 25 3 25

N cases 13399 7350 4884 3713

N attr 41 41 39 39

N leafs 67 52 36 23

hit r(0) 0.484 0.041 0.620 0.045

hit r(t) 0.562 0.064 0.702 0.074

hit r(v) 0.551 0.052 0.695 0.066 2χ 4240.27 4237.07 1678.45 2638.80

C 0.490 0.604 0.506 0.645 Note: N alts : number of choice alternatives

N cases : number of observations in training data set

N attr : number of attributes

N leafs : number of leaf nodes 2χ : Chi-Square value

C : contingency coefficient

hit r(0) : expected ratio of correctly predicted cases (null model)

hit r(t) : expected ratio of correctly predicted cases (training set)

hit r(v) : expected ratio of correctly predicted cases (validation set)

146

Similar condition also applies to joint activities, where the accuracy on the validation

set is somewhat decreased from 0.702 (training set) to 0.695 (validation set). These fit

measures are even higher compared to the independent activities case indicating that

joint activity models perform substantially better than independent activities models.

Furthermore, the chi-square-based contingency coefficient of 0.506 (model 1) and

0.645 (model 2) also shows a significant and larger impact of the decision rules on the

action variables in the joint case.

Due to limited space and given the large number of decision rules, we cannot display

the entire set of results of the decision tree. Instead, in order to give a summary view of

the outcomes, we will discuss the results of the impact analysis in terms of Chi-square

measures.

7.8.1 Independent Activity

Table 7.6 displays the impact table for independent activity model for the first and

second model. For the first case, there are 3 choice alternatives (action variable),

whether the activity is conducted at the same location as the previous activity (in the

schedule), and the same location of the next activity or at some other location. In case

of a latter decision, a choice from the set of locations available in the prism is made. It

determines the location in terms of a combination of size class and distance class of the

postcode area which involves 25 choice alternatives.

As it appears in model 1 of independent activities, the condition when the previous

location is the same as next location (LvoisLNa) is by far the most important variable

for selecting relative location for conducting an independent activity, as indicated by

the highest Chi-square value. Activity type (Aty), activity duration (Adur), starting time

(Abt), and time availability (Tavail) also have a strong influence in selecting a relative

activity location. Several attributes related to the location, such as order of municipality

(Vogord), travel time from home to previous location (Vohtt), size of area (Vosize), and

previous activity (Voty) have an impact on selecting the activity location.

In addition, characteristics of the next location also have an impact, such as size of

next location (Nasize), travel time from next location to home (Nahtt), order of

municipality of next location (Nagord), and next location is home (NaH). Furthermore,

several personal and household attributes influence the choice of activity location, such

as, gender (Gend), urban size (Urb), Work status (Wstat and PWstat), household

composition (Comp), age (pAge), household income (SEC), day of the week (Day), and

car ownership (Ncar).

147

TABLE 7.6 Impact Table for Independent Activity

Model 1 Model 2

No Condition Variable Chi-Square No Condition Variable Chi-Square

1 LvoisLNa 5590,31 1 Vogord 3236,3

2 Aty 3311,57 2 LvoisLNa 2526,3

3 Adur 1958,47 3 Aty 2368,4

4 Vogord 1402,69 4 Comp 178,58

5 Vohtt 670,82 5 Nagord 172,6

6 Vosize 146,49 6 Urb 166,72

7 Gend 21,32 7 SEC 116,38

8 Urb 21,1 8 Vosize 51,18

9 Wodur 17,92 9 Adur 45,78

10 Voty 16,64 10 Vohtt 41,69

11 Nasize 14,76 11 pAge 26,51

12 Wstat 12,59 12 Day 13,37

13 Nahtt 9,52 13 Child 12,99

14 Nagord 6,96 14 Wstat 11,01

15 Comp 3,9 15 PWstat 10,77

16 Abt 3,74 16 Tavail 10,27

17 pAge 3,71 17 Gend 8,98

18 SEC 3,44 18 Vopark 7,49

19 Day 2,16 19 Napark 4,84

20 Ncar 1,75 20 Ncar 2,34

21 PWstat 1,71

22 NaH 1,54

23 Tavail 0,54

In terms of model 2, the number of influential variables appears slightly smaller than

that of model 1 (from 23 to 20 variables). Nevertheless, the significant variables are

almost the same, only the order of priority is somewhat different. In order to choose a

particular time-size band location, the condition when the previous location is the same

as next location (LvoisLNa) and the order municipality of previous location are the

most important variables. Other attributes of the previous activity location still play a

role, such as size of area (Vosize), travel time from home to previous location (Vohtt), and parking price (Vopark). Additionally, several characteristics pertaining to the next

location influencing the decision, such as order of municipality of the next location

(Nagord), and parking price (Napark). Other activity-level components such as activity

type (Aty) and activity duration (Adur) also play a role. The remainder of the influential

variables consists of person and household-level attributes.

148

7.8.2 Joint Activity

As shown in Table 7.5, the tree for joint activities consists of 36 and 23 leaf nodes

(decision rules) for the first and second model. Similar to the independent activity

model, the goodness-of-fit of model 1 is also better than that of model 2. The hit ratio

of model 1 compared to a null-model indicates a small but significant improvement:

the hit-ratio of a null-model equals 0.620 and the hit ratio of the tree after splitting

equals 0.702. The overall accuracy on the validation set is also almost the same, fall

somewhat from 0.702 to 0.695. A Chi-square-based contingency coefficient confirms

that there is a strong impact of the decision tree structure on the action variable, in

particular for model 2 (0.645). Although slightly lower than model 1, the result of

model 2 also shows improvement. The hit-ratio of a tree after splitting increases than

that of a null-model (0.074 and 0.045) and there is not much different on the validation

set accuracy (0.066). Overall, the performance of the joint activity model is better than

that of the model for independent activities.

Table 7.7 presents the impact table for the joint activity model for the first and second

model. As it appears, distance maximum in time band (Dmax) and travel time by car

from previous location to home (Vohtt) are by far the most important variables in both

models of joint activity. Other variables at the activity-level play an important role in

deciding on the choice of location for joint activities. Furthermore, other personal and

household-level attributes also influence the location choice decision, such as urban

size (Urb), household income (SEC), driver license holder (Driver), day of the week

(Day), work status of female (PWstat), and car ownership in a household (Ncar).

Similar to the model for independent activities, activity-level attributes also have a

major influence.

To sum, the number of variables influencing the activity location decision for joint

activities is slightly fewer than for independent model. The reason probably the sample

size of independent activities models is fairly larger than joint activities models.

Activity-level attributes have made significant contribution in both the independent and

joint activity models. Activity type (Atype) and activity duration (Adur) in particular,

are significant factors in both models. The tendency of choosing a particular activity

when the previous is the same as the next activity (LvoisLNa) is indicative of a multi-

purpose trip-chain. Moreover, the characteristics of the previous and the next location

also influence the location decision for independent activities. In contrast, distance

maximum in time band (Dmax), is the most influential attribute for doing joint

activities, suggesting that both adults heads prefer to travel jointly for a longer distance.

Furthermore, the travel time (by car) from previous location to home (Vohtt) also has

significant influence on joint activities decision.

149

TABLE 7.7 Impact Table for Joint Activity

Model 1 Model 2

No Condition Variable Chi-Square No Condition Variable Chi-Square

1 Dmax 2277.39 1 Dmax 9684.12

2 Vohtt 832.44 2 Vohtt 981.16

3 Aty 354.56 3 Nagord 557.47

4 Adur 157.44 4 Adur 217.77

5 NaH 103.17 5 Dmin 141.91

6 Vosize 88.6 6 NaH 129.32

7 Urb 25.26 7 LvoisLna 37.3

8 SEC 9.26 8 Urb 23.18

9 MxSizeD2 6.02 9 Tavail 22.25

10 Abt 4.81 10 Abt 13.9

11 Vogord 3.52 11 Driver 7.68

12 Day 3.44 12 PWstat 6.97

13 Nagord 1.58

14 VoisA 1.46

15 Ncar 1.03

16 Driver 0.6

Similar to independent activities models, activity duration (Adur) has major impacts in

both models of joint activities. This variable indicates that the decision on choosing a

location depends on activity duration either in independent or joint activity. Overall,

both models have a satisfactory performance. Nevertheless, the performance of the

model for joint activities is better than that of independent activities.

7.9 CONCLUSIONS

This study was intended to refine the ALBATROSS model. In the present paper, we

focus on activity location choice of male-female heads, in particular making a

distinction for those activities that are conducted independently and jointly. Each

model consists of 2 sub-models, where the first model relates to the decision whether

or not the activity is performed in the same location as the previous activity, whether

the activity is done at the same location as the next activity, or whether it is conducted

elsewhere. The second model relates to the last choice option in the first model that

comprises into 25 choice alternatives. It verifies the location in terms of a combination

of size - distance class of the postcode area. The size class depends on a particular

activity type and the size of available facilities at the activity location. Size is classified

150

into 5 categories based on employment in the relevant sector for the activity considered

and distance is classified in terms of a detour travel time (by car) also into 5 categories.

For those decisions, rule-based models using a CHAID-based algorithm were derived

from activity-trip diary data. The independent activities models included 67 and 52

condition-action rules for model 1 and model 2 respectively. Although slightly smaller

than the independent activities models, the joint activities models involve 36 and 23

condition-action rules. In both cases, the validity of the decision tree is reasonable in

the sense that the derived rules are readily interpretable and the overall goodness-of-fit

of the model on a validation set is acceptable as well. Furthermore, in both cases, a

substantial improvement in goodness-of-fit relative to a null model indicates that there

is a moderately strong association between condition variables defined at the household,

individual, activity and schedule level, on the one hand, and the location decisions, on

the other. Furthermore, the stability of performance on a validation set suggests that

derived rules are generalizable to hidden cases. These results suggest that the way of

structuring the household decisions as we proposed in this study has merits. Hence, by

refining the existing ALBATROSS in this way we expect that the accuracy and

sensitivity of predictions will be improved.

The tendency of conducting a particular activity at the same location as the previous

activity is higher for independent activities than joint activity. The same condition also

applies to activities that are conducted at the same location as the next activity. These

results imply that male and female are more likely to conduct multiple activities at one

particular location independently than jointly. These results do make sense, since the

activity-travel behavior of one person is different from the other person, even though

male-female couple living in the same household. Moreover, the passion of men is

fairly different from women.

Let’s give you one example. If a man and a woman travelling alone to the shopping

centre zone, they probably have different interests. The man could visit the electronic

shops, watch shops, clothing store (for men), and probably restaurant. From the woman

perspective, she could go to perfume store, jewelry store, beauty center, clothing store

(for women), and perhaps also restaurant. So, if they travel together, they would likely

have less similar purposes. In such cases, they may only go to the restaurant together,

while other activities will be performed alone. Underlying those descriptions, it is

reasonably to say why male or female prefer to conduct multiple activities when they

travel alone than travel together.

Activity-level attributes play a significant role in both the independent and joint

activities models. In terms of independent activities, in any location choice, females

151

take higher percentage than that of the males. Nevertheless, in particular in bring/get

and daily shopping (shop-1) activities, male/female is more likely to do multiple

activities. In terms of joint activities, both adult heads prefer to travel jointly for a

longer distance and that particular joint activity is performed before going home. Both

male and female tend to do joint activities with any other joint activity (multiple

activities), in particular in daily shopping (shop-1), social, leisure and touring.

In addition, activity type and activity duration are significant factors in independent

and joint activities models. The tendency of choosing a particular activity when the

previous is the same as the next activity is indicative of a multi-purpose trip-chain.

Furthermore, the characteristics of the previous and the next location also influence the

location decision for independent activities. In contrast, distance maximum in time

band, is the most influential attribute for doing joint activities, suggesting that both

adults heads prefer to travel jointly for a longer distance. Furthermore, the travel time

(by car) from previous location to home also has significant influence on joint activities

decision.

REFERENCES







Oriented Simulation System”. Transportation Research Part B, 38, 613-633.



Arentze, T.A. and Timmermans, H.J.P. (2007), “Robust Approach to Modeling Choice

of Locations in Daily Activity Sequences”. Paper presented in Transportation

Research Board Annual Meeting, Washington, D.C.

Chandraskharan, B. and Goulias, K.G. (1999), “Exploratory Longitudinal Analysis of

Solo and Joint Trip Making in the Puget Sound Transportation Panel”.

Transportation Research Record, 1676, 77–85.

Gliebe, F.J. and Koppelman, F.J. (2002), “A Model of Joint Activity Participation

between Household Members”. Transportation, 29, 49-72.

152

Golob, T., and McNally, M.G. (1997), “A Model of Activity Participation and Travel

Interactions between Household Heads”. Transportation Research B, 31(3), 177-194.


Categorical Data”. Applied Statistics, 29, 119-27.

153

Chapter 8

CAR ALLOCATION DECISIONS IN CAR-

DEFICIENT HOUSEHOLDS: THE CASE OF NON-

WORK TOURS

Anggraini, R., Arentze, T.A., and Timmermans, H.J.P., 2010. Paper is officially

accepted for publication in the Journal of Transportmetrica

ABSTRACT The activity-travel decisions of individuals in multi-person households are interrelated. This applies in particular to male-female household heads, as key decision makers in a household. As a result, any realistic model of travel behavior requires accommodating these interpersonal dependencies and household constraints. The present study examines such interactions in the context of the car allocation choice decision in car-deficient households as part of an activity-scheduling process, focusing on non-work tours. A CHAID-based algorithm is applied to derive a decision tree using a large activity diary data set recently collected in the Netherlands. The results show a satisfactory improvement in goodness-of-fit of the decision tree model compared to the null model. The gender seems still to play a role. A descriptive analysis indicates that men more often than women get the car for non-work tours for which a car allocation decision needs to be made. Tour-level attributes are shown to influence the household car allocation decision for non-work tours. The decision to allocate the car is considerably influenced by the longest distance (travel time) from home to a particular location in a tour of men and women. The probability that the men and women get the car increases with the increasing travel time monotonically. Socio-economic and situational factors have less influence to car allocation decision. Overall, men have more influence to the car allocation decision for non-work tour, as indicated by the number of influential variables that relates to the males in the impact table. The developed models will be incorporated in a refinement of the ALBATROSS model – an existing computational process model of activity-travel choice.

154

8.1 INTRODUCTION

Since many decades, mode choice models have received much interest in activity-

based transport demand modeling. Mode choice models are intended to identify which

transport mode people use to go to a particular destination. Virtually all mode choice

models assume that choosing a transport mode is the outcome of an individual decision

making process. However, it has been realized that many activity-travel decisions are

household decisions. Especially, in situations of constrained resources and

synchronization of household activities, models of individual decision making

imperfectly address the decision making between people belonging to the same

household. In car-deficient households, the number of available cars is a scarce

resource, and therefore the decision who will use the car in case of fully or partially

activity-travel episodes in time require a household decision. Although the importance

of household decision making has been realized in activity-based modeling from the

very beginning, models of household decision making are still relatively scarce in this

research community (see Timmermans, 2006 for an overview).

In this paper, car allocation decisions are considered as an element of a more

encompassing activity scheduling process. A large number of factors that potentially

influence car allocation decisions in car-deficient households are considered. These

factors relate to activity-schedule, space-time setting, and individual and household

characteristics. To get a better understanding how a decision is made in a particular

household, we focus on households consisting of two (male-female) heads household.

Both are drivers, and the household owns a single car.

On contributing to the still scarce literature on household activity-travel decisions, this

study will focus on car allocation decisions for non-work tours in car-deficient

households (e.g., more people with a driver license than cars). The paper will report the

conceptualization of the problem and present the empirical results of the car allocation

model for two household heads. In an earlier study (Anggraini, et al., 2008), we

examined car allocation decisions for work tours. In contrast, in this particular study,

we intend to identify how the car is allocated between household heads for non-work

tours. As opposed to work tours, travel for any activity episode or set of chained

activity episodes that does not include a work activity is considered a non-work tour.

The problem of modeling this allocation problem for non-work tours is more complex

than for work tours because the decision at this stage depends considerably on the

outcome of the previous stages in the scheduling process. Hence, the car can be

allocated to the male, female and none.

155

The paper is structured as follows: The subsequent section describes the data used for

extracting the decision trees. The following section explains the methodology: car

allocation model for non-work tours, decision tree induction method and impact table.

After this section, the results of some descriptive analyses will be discussed. This is

followed by a discussion of the empirical analysis of deriving a decision tree from the

MON data. The paper is wrapped up with drawing conclusions.

8.2 DATA DESCRIPTION

The data used originates from the Dutch National Travel Survey (MON = Mobiliteit Onderzoek Netherlands) collected in 2004 covering all of the Netherlands. The survey

is conducted on a regular basis to obtain travel and activity information of residents in

the Netherlands. It is a household survey where data is collected of all household

members for the diary day as well as general information about household and

individual attributes such as gender, age, vehicle ownership and driving license

ownership, home location, individual income, occupation, number of working hours

per week, etc. Respondents were also requested to give information about all trips

made on a designated day as well as on the activities conducted at trip destinations.

Information for each trip includes start time, trip purpose, destination, activity type at

the destination, and transport mode. Situational variables are reported as well. All in all,

this survey provides a comprehensive data source to analyze activity-travel behavior of

Dutch residents. In the data collection, 29,221 households filled out a one-day

travel/activity diary and 28,600 of these households fit the criteria for being considered

in ALBATROSS. The data were transformed to an activity-diary data format for the

current estimation purpose.

8.3 METHODOLOGY

8.3.1 Car Allocation Decisions

The car allocation decision model focuses on car-deficient households (i.e. the number

of drivers exceed the number of cars) and involves a joint decision between male-

female heads. As indicated, the total sample extracted from the MON data includes

28,600 households. Given the purpose of this study, only the following households and

days are relevant: (1) there are two heads in the household; (2) there is one car in the

household; (3) both heads are drivers and (4) both household heads have a non-work

activity on the day considered. As it appears, 3,190 households (and days) and 4,049

number of cases fit these criteria. The model includes the option that none of the

156

household heads uses the car, but some other means of transport instead. Hence, the

decision options are male, female, and none.

As indicated in the introduction, this study is part of a refinement of ALBATROSS

system, an existing operational activity-based model developed for the Dutch Ministry.

ALBATROSS considers several activity types in general: work, work-related such as

business and school, and non-work activities such as escorting, shopping (daily and

non-daily), service-related, social, leisure, and touring. Location decisions for those

activities are made at different moments in the sequential scheduling process. The car

allocation decision to non-work tours is made after the location choice decision. The

decision on car allocation can give significant information for transport mode choice in

the next stage of the process.

As a consequence of the sequential decision process, the available information for each

choice facet is limited. At the moment the car allocation decision for non-work tours is

made, the following information is available. First of all, the generation of a work

activity is executed in the ALBATROSS scheduling process. It includes information

about the number of work episodes, start time and duration, and the location and mode

choice of the work tour. Furthermore, it takes work-related activities (business and

other mandatory) into account. The successive process generates non-work activities,

such as escorting, shopping, service, social, leisure and touring. The first three

activities are considered to be a household task, while the last three activities are non-

household tasks. Household tasks are the type of activity that needs commitment from

household heads, such as delivering children to school. As a result, a task allocation

decision needs to be made.

Joint participation in activities also needs trade-offs between adult heads of household.

This can be applied to either household tasks or non-household tasks. Having defined

all activities in a priority order, trip-chaining is established. Additionally, the location

decision is simulated, and only after all these steps the car allocation decision for non-

work tours is established.

The process underlying the car allocation decision for non-work tours is schematically

shown in Figure 8.1. Primarily, we ensure that only households that are involved in

out-of-home non-work activities are being processed. Those households that do not

conduct any out-of-home activity or only out-of-home work activities will be

eliminated. In addition, only non-work activities performed by male and female are

taken into consideration. Further, only overlapping non-work activities of the male’s-

female’s that occurs in the same time slot is taken into account.

157

Overlapping tours are defined as a pair of tours conducted by respectively male and

female of which the start and/or end times of each tour (simulating use of a car for the

tour) defines a fully or partially overlapping episode. Instead of trip-based, tour-based

concept is assumed to define the car allocation decision between those overlapping

tours.

As we may know, a tour consists of a sequence of trips that starts and ends at a

particular location (i.e., home). We therefore have to determine the primary activity in

each tour. In order to identify the primary activity in a particular tour, we consider the

hierarchical order of activity priority. As mentioned, ALBATROSS considers 10

activity categories in priority order starting from work, business and other (mandatory)

activities. A group of non-work activities is considered, such as escorting, shopping

(daily and non-daily), service-related, social, leisure, and touring. Business and other

(mandatory) activity are also considered as non-work activities in this model.

START

Non-work

tours of both

heads is

overlapping

N

Car Allocation for

Non-Work Tour

Tou

k=1

Allocated to

Male, Female, or

None

k < K

N

STOP

Y

k = k + 1

Y

Non-work

activity is in

HH schedule

Y

N

STOP

STOP

k = index of # car allocation decisions K = # car allocation decisions

FIGURE 8.1 The Process of Car Allocation Decisions for

Non-Work Tour

158

Given the sequential process model underlying ALBATROSS, the car allocation

decision for non-work tours at this stage depends considerably on the outcomes of the

earlier stages in the process, such as work tour decision. A work tour decision is

deemed to have priority and the work tour decision is defined in a previous stage of the

process. Although work tour is an outcast in the present study, nevertheless, it exists as

condition variables, such as number of work activity and work duration (if any) of the

male and female.

To signify how to assign the case of car allocation decisions occurring in a household

on the day concerned, we take the following conditions into account. In principle, only

if the two heads of household having non-work activities it will be taken into account,

and the process is stopped otherwise. Further, only those non-work activity tours that

are performed at the same time period (have an overlap in time) are taken into

consideration. Note that, a single non-work tour of male or female is eliminated from

the process, because at this stage, all activities performed by male and female are

known in the schedule. In addition to that, if the non-work tour of one person is

overlapping with the work tour of the other person, the process is stopped as well,

since the work tour has been assigned earlier. Hence, only fully or partially overlapping

non-work activities are taken into account. In case the primary activity is a joint

activity, the car allocation decision is included also, to decide who the driver is.

In case one person has more than one activity in a tour, the underlying hierarchy of

activity priority is also taken into account to determine the primary activity of the tours.

The decision on who should use a car is executed in this stage. Hence, as mentioned,

the car can be allocated to male, female or none. If a household member is assigned the

car, it means this person is the driver. If the outcome of the model is none, it means that

either household member can still be a car passenger, with another driver than the two

household heads or that both can be non-car users.

Table 8.1 and Figure 8.2 show the case of one particular household that performs a

sequence of activities. As it appears the activity program of the male starts at 06:45 am

(Table 8.1). He further arriving back home at 15:10 pm. During the male’s work

activity, the female goes shopping. The time of conducting these two activities is

overlapping (case #1). However, we ignore such a case, due to the existence of a work

tour. We assume that a car allocation decision is already made for work tours during

this time slot. Further, the female leaves for the work place at 15:30 and is back home

at 18:00 by public transport, while during the same time period (16:30 – 17:30) the

male goes to a sport club. In this case (case #2), non-work tours of male and female are

performed at the same time period.

159

TABLE 8.1 Itinerary of Male-Female Heads in a Particular Household

Activity Time

Male Female

06:45 – 15:10 Work

11:00 – 14:00 Shopping

15:30 – 18:00 Work

16:30 – 17:30 Sport

19:00 – 21:00 Social

As this case is our concern in this study, the decision to allocate the car is necessary to

do. Hence the car allocation decision to non-work tour is executed in this case. After

arriving home, the male makes another tour to meet friend in a restaurant (case #3). In

this case, there are no overlapping episodes and hence, there is no decision for

allocating the car. In summary, among the three cases happened in this particular

household, only one case that is required to be executed, that is case #2.


CHAID-based tree induction method is used to identify the rules that describe which

choices (i.e., actions) are made under which conditions. CHAID (Chi-square Automatic Interaction Detector) generates non-binary trees, i.e., trees where more than

two branches can be attached to a single root or node, based on a relatively simple

800

1000

1200

1400

1600

1800

2000

2200

600

800

1000

1200

1400

1600

1800

2000

2200

600

Male

Female

Case #1 Case #2

Work Tour

Work Tour

Non-Work Tour

Non-Work Tour

Non-Work Tour

Case #3

FIGURE 8.2 An Example of Defining Car Allocation Decision Cases in

Household Schedules

160

algorithm that is particularly well suited for the analysis of larger datasets. CHAID

relies on the Chi-square test to determine the best next split at each step. CHAID

generates a decision tree by splitting subsets of the space into two or more nodes

repeatedly, beginning with the entire data set (Kass, 1980). The split that maximizes a

significance value of a Chi-square test - after adjustment for multiple testing

(Bonferroni adjustment) - across condition variables is used for splitting if the split is

significant. The process is repeated for each newly created group until no more

significant splits are found. In order to develop the decision tree, 75% of the cases are

used for training and the remaining cases were used for validation. The decision tree

can be interpreted as visualizing a series of decision heuristics that indicate which

decisions will be made (action) under particular situations.

8.3.3 Impact Tables









variables in rows and the frequency distribution across the levels of the target variable

(i.e., the action variable) in columns. The frequency table for a given condition variable

is generated by applying the model as many times as there are levels of the condition

variable. In each run, each training case is assumed to take on the level considered on

the condition variable. The frequency distribution across actions of the action variable

predicted under that setting is recorded. Repeating this process for each level of the

condition variable yields a frequency cross table of the condition variable against the

action variable. The impact of the condition variable is then measured as the Chi-

square for this frequency table. Further details of this approach can be found in Arentze

and Timmermans (2003).


Table 8.2 portrays the condition variables that were used as input to the tree-induction

algorithm. The condition variables concern household-level (including accessibilities),

person-level, tour-level variables, and schedule-level variables.

161

TABLE 8.2 Condition Variables for Car Allocation Model

No Level Acronym Variables Classifications

1 HH Urb Urban density 0:most densely , 4:least densely

2 HH Comp Household composition 2:double-1-worker, 3:double-2-worker,

4:double-non-worker

3 HH Child Age of the youngest children 0:no children, 1:<6, 2:6-11, 3:12-17 yrs

4 P Day Day of the week 0:Monday to 6:Sunday

5 P pAge Age of the oldest person in household

0:<35, 1:35-<55, 2:55-<65, 3: 65-<75, 4:75+

yr

6 HH SEC Socio-economic-class (SEC) 0:0-16,250 (low), 1:16,251-23,750 (low-mid),

2:23,751-38,750 (mid-high), 3:38,750+ (high)

7 P wstatm Working status – M 0:non-worker, 1:part-time, 2:full-time

8 P wstatf Working status – F 0:non-worker, 1:part-time, 2:full-time

9 T BTM Start time of doing non-work activity –

M (in hour)

0:≤1030, 1:1031-1333, 2:1334-1620.50,

3:>1620.50

10 T BTF Start time of doing non-work activity –

F (in hour) 0:≤1038, 1:1039-1343, 2:1344-1627, 3:>1627

11 S ntourM # tours in a day – M 1:1, 2:2, 3:3, 4:≥4

12 S ntourF # tours in a day – F 1:1, 2:2, 3:3, 4:≥4

13 S ntourHH # tours in a day – HH 2:2, 3:3, 4:4, 5:≥5

14 S nworkm # work activities in a day – M 0:0, 1:1, 2:≥2

15 S nworkf # work activities in a day – F 0:0, 1:1, 2:≥2

16 S nwohh # work activities in a day – HH 0:0, 1:1, 2:≥2

17 S nNWm # non-work activities in a day – M 1:1, 2:2, 3:3, 4:≥4

18 S nNWf # non-work activities in a day – F 1:1, 2:2, 3:3, 4:≥4

19 S nNWhh # non-work activities in a day – HH 2:2, 3:3, 4:4, 5:≥5

20 T nbusim # business episodes in tour – M 0:0, 1:1, 2:≥2

21 T nbusif # business episodes in tour – F 0:0, 1:1, 2:≥2

22 T nbrm # bring/get episodes in tour – M 0:0, 1:1, 2:≥2

23 T nbrf # bring/get episodes in tour – F 0:0, 1:1, 2:≥2

24 T nsh1m # shop-1-store episodes in tour – M 0:0, 1:1, 2:≥2

25 T nsh1f # shop-1-store episodes in tour – F 0:0, 1:1, 2:≥2

26 T nshnm # shop-n-store episodes in tour – M 0:0, 1:≥1

27 T nshnf # shop-n-store episodes in tour – F 0:0, 1:≥1

28 T nsocm # social episodes in tour – M 0:0, 1:1, 2:≥2

29 T nsocf # social episodes in tour – F 0:0, 1:1, 2:≥2

30 T nserm # service episodes in tour – M 0:0, 1:≥1

31 T nserf # service episodes in tour – F 0:0, 1:≥1

32 T nleim # leisure episodes in tour – M 0:0, 1:≥1

33 T nleif # leisure episodes in tour – F 0:0, 1:≥1

34 T ntoum # touring episodes in tour – M 0:0, 1:≥1

35 T ntouf # touring episodes in tour – F 0:0, 1:≥1

36 T nothm # other episodes in tour – M 0:0, 1:≥1

37 T nothf # other episodes in tour – F 0:0, 1:≥1

38 T distM The longest distance (travel time) in

each tour by car – M (in minute) 0:0, 1:1-6, 2:7-12, 3:13-20, 4:>20

39 T distF The longest distance (travel time) in

each tour by car – F (in minute) 0:0, 1:1-6, 2:7-11, 3:12-20, 4:>20

162

TABLE 2 (cont.)

No Level Acronym Variables Classifications

40 T TTptM Travel time ratio between PT & car – M (in

minute) 0:≤100, 1:101-244, 2:245-502, 3:>502

41 T TTptF Travel time ratio between PT & car – F (in

minute) 0:≤100, 1:101-236, 2:237-493, 3:>493

42 T TTcbM Travel time ratio between car & bike – M

(in minute) 0:≤25, 1:26-39, 2:40-100, 3:>100

43 T TTcbF Travel time ratio between car & bike – F (in

minute) 0:≤25, 1:26-40, 2:41-100, 3:>100

44 T NWdurM Duration of non-work tour – M ( in minute) 0:≤52, 1:53-120, 2:121-231, 3:>231

45 T NWdurF Duration of non-work tour – F ( in minute) 0:≤50, 1:51-113, 2:114-216, 3:>216

46 T trainM Train accessibility – M 0:no, 1:yes

47 T trainF Train accessibility – F 0:no, 1:yes

48 T busM Bus accessibility – M 0:no, 1:yes

49 T busF Bus accessibility – F 0:no, 1:yes

50 T rparkM Ratio # paid parking places to total #

parking places – M 0:0, 1:1-7, 2:8-15, 3:16-23, 4:>23

51 T rparkF Ratio # paid parking places to total #

parking places – F 0:0, 1:1-7, 2:8-15, 3:16-24, 4:>24

52 T pparkM Average price of parking – M 0:0, 1:1-7, 2:8-22, 3:23-44, 4:>44

53 T pparkF Average price of parking – F 0:0, 1:1-8, 2:9-22, 3:23-47, 4:>47

54 S mwork Male has work activity in a schedule 0:no, 1:yes

55 S fwork Female has work activity in a schedule 0:no, 1:yes

56 S wdurM Duration of work activity – M ( in minute) 0:0, 1:1-305, 2:306-495, 3:496-560,

4:>560

57 S wdurF Duration of work activity – F (in minute) 0:0, 1:1-245, 2:246-365, 3:366-515,

4:>515

58 S ncarAl # car allocation cases 1, 2, 3, 4, 5

59 T AtourM Primary activity in a tour – M 2:business to 10:other

60 T AtourF Primary activity in a tour – F 2:business to 10:other

61 T nAcToM # non-work episodes in a tour – M 1:1, 2:2, 3:≥3

62 T nAcToF # non-work episodes in a tour – F 1:1, 2:2, 3:≥3

Continuous condition variables, such as travel time, duration, and parking price, are

discretisized by using an equal-frequency interval method which divides a continuous

variable into n parts, in which each part contains approximately the same number of

cases. Household and individual attributes consist of the presence of young children in

a household, work status, socio-economic class (in Euro), household composition, age

of the person, urban density (number of home addresses per area unit in the zone where

the household lives classified on a 5-point scale) and the day of the week (no. 1-8 in

Table 2). Start time of a non-work activity is known at the moment the decision is

made (#9-10). The number of non-work tours performed by male and female in a day

is defined in #11-13. The number of work activity episodes (if any) that is performed

Note: M = Male, F = Female, HH = Household, PT = Public Transport

A = Activity, P = Person, S = Schedule

163

by male or female is also taken into account (#14-16). The number of non-work

episodes that is performed by male or female is taken into account (#17-19). The

number of each particular non-work activity in each tour is defined in #20-37.

Accessibility variables, such as travel time, train and bus connections, parking price

and free-paid parking place ratio were also used (#38-43 and #46-53). Note that, all of

those are indicated for the primary activity of the tour. They are calculated based on

national datasets of the transport system (car, bike/walk and public transport), parking

facilities and land-use system (employment data by sector and postcode area). They all

relate to the trip to the non-work location. If a non-work activity is conducted in the

same postcode area as where the person lives, then travel time is set to zero. Travel

time by car is included as a direct measure of accessibility, and refers to the longest

distance (travel time) in each tour if multiple activities are involved. Travel time ratios

between modes are used as indicators of relative accessibility by particular modes.

Ratios are used to allow the algorithm to identify impacts of modes more easily.

Work durations of each male and female (if any) are considered as schedule attributes

(#56-57). Meanwhile durations of non-work tours are defined as tour-level attributes in

#44-45. Variables #54-55 mention whether the male/female has a work activity on a

particular day. Further, variable #58 measures the number of car allocations for non-

work tours in the household. The type of primary activity of a tour is represented in

#59-60. Lastly, the number of non-work activities in each tour is represented in #61-62.

As a result, a total of 62 condition variables are defined. The action variable, the

outcome of the car allocation model, involves assigning the car to the male, female, or

none of the two household heads.

8.4 DESCRIPTIVE ANALYSIS

In this section we describe some descriptive analyses carried out to get a better

understanding of the characteristics of the sample after selecting car deficient

households. As discussed above, only a subset of households is relevant for the car

allocation model, because the problem concerns car allocation to non-work tours in car

deficient households. A total of 3,190 households were selected from the MON data,

yielding 4,049 relevant cases of car allocation decisions.

As mentioned above, with regard to allocating the car to non-work activities, instead of

allocating it for each overlapping activity episodes, we assume it for the overlapping

tours. In this study, tour consists of a group of activity episodes that normally start and

end at home at a particular time.

164

TABLE 8.3 Primary Activity of a Tour of Male – Female

No Primary Activity of the Tour Male (%) Female (%)

1 Business 2.69 0.89

2 Bring/get 3.31 3.83

3 Shop-1 24.57 28.06

4 Shop-n 5.63 6.59

5 Service 5.86 5.46

6 Social 23.61 24.50

7 Leisure 21.56 18.40

8 Touring 11.66 10.99

9 Other 1.11 1.28

Total 4,049

TABLE 8.4 Percentage of Getting a Car by Male/Female across Work Status

Male Female Work

Status Use other modes

(%)

Get the car (%) Use other modes

(%)

Get the car (%)

Non-worker 23.59 29.61 55.72 7.78

Part-time 2.79 3.33 15.68 2.52

Full-time 17.19 23.49 15.34 2.96

Total 43.57 56.43 86.74 13.26

1764 2285 3512 537 # cases

4,049 4,049

TABLE 8.5 Average Duration of Non-work Tour(s) across Work Status (in

minute)

Male Female

Work

Status

Average

duration of non-

work tours (min)

Standard

Deviation

(min)

Freq. Average duration of

non-work tours

(min)

Standard

Deviation

(min)

Freq.

Non-worker 160.47 168.79 2154 157.63 162.83 2571

Part-time 197.45 206.64 248 168.22 161.17 737

Full-time 184.76 180.71 1647 171.67 176.04 741

Total 172.62 176.67 4049 162.12 165.10 4049

165

In a single episode tour, the primary activity of the tour is that of the single activity

episode. For example, in a trip-chain of home-shopping-home, it is called a shopping

tour. Nevertheless, in a multi-episode tour, it seems quite complex to define which

activity should be called as a primary activity. Therefore, as said, we defined the

primary activity of the tour based on the hierarchy of activity priority as explained in

Section 8.3.1 and is shown in Table 8.3.

Table 8.3 displays the frequency distribution of non-work tours for which a car

allocation decision is to be made across the primary activity of a tour of the male and

female in the present study. For both male and female, shopping (one store) is the most

frequent activity on the tours that require a car allocation decisions (28.06% and

24.57%). It is followed by social purpose (24.50% for female and 23.61% for male)

and leisure purpose (21.56% and 18.40%). In overall, females have a lower percentage

of business and leisure tours for which a car allocation decision is needed compared to

males and higher percentages of tours for other purposes in the overlapping cases.

Table 8.4 shows the percentages of getting a car for non-work tours by male and

female across work status. The probability of the male getting the car is 56.43% and

female that gets a car is about 13.26% across 4,049 cases. In terms of work status, in

29.61% and 23.49% of the cases where a car allocation decision is involved the male is

a non-worker and full-time worker respectively, and gets the car. Meanwhile, in

55.72% of the cases where a car allocation decision is involved the female is non-

worker and uses another transport mode. The result indicates that male more often than

female uses a car in an overlapping non-work tour.

Table 8.5 shows the distribution of the average non-work tour(s) duration of male and

female by work status. Note that persons may conduct more than one non-work tour in

a day, and the figures presents the durations on a per-tour basis including the travel

time. As we can see, on average males and females have similar non-work tour

durations when a car allocation decision is involved. However, in each work status

group, the average duration of the males is higher than that of the females.

8.5 RESULTS

As indicated, for deriving the car allocation for non-work tours model, a total of 4,049

observations could be derived from the data set. About 75% of these cases (3,057)

were used for training and the remaining cases were used for validation. Given a

minimum group size of n=50 cases at parent nodes and a 5% alpha level, the tree

generated by CHAID consists of 28 leaf nodes (decision rules).

166

TABLE 8.6 Results of the Car Allocation Model to Non-Work Tours

Indicators Results

N alts 3

N cases 3,057

N attr 62

N leafs 28

hit r(0) 0.432

hit r(t) 0.544

hit r(v) 0.526

2χ 1,040.25

C 0.504

Note: N alts : number of choice alternatives

N cases : number of observations in training data set

N attr : number of attributes

N leafs : number of leaf nodes

hit r(0) : expected ratio of correctly predicted cases (null model)

hit r(t) : expected ratio of correctly predicted cases (training set)

hit r(v) : expected ratio of correctly predicted cases (validation set) 2χ : Chi-Square value;

C : contingency coefficient

TABLE 8.7 Impact Table of Car Allocation Decision to Non-Work Tour Model

No Variables IS ISmale ISfemale ISnone MSmale MSfemale MSnone

1 distM 2181,52 955.52 675.36 550.65 1.00 -0.99 -1.00

2 distF 493,19 4.39 271.27 217.53 -0.23 1.00 -1.00

3 AtourM 146,72 30.20 17.95 98.57 -0.87 -0.41 0.71

4 ntouf 37,19 9.53 2.48 25.19 -1.00 -1.00 1.00

5 TTcbF 30,22 1.38 13.20 15.65 -1.00 -1.00 1.00

6 ntoum 29,05 11.10 0.02 17.93 -1.00 1.00 1.00

7 AtourF 14,96 2.55 2.19 10.21 0.00 0.00 0.00

8 Urb 12,74 2.39 1.76 8.60 1.00 1.00 -1.00

9 RParkM 12,19 4.50 0.51 7.18 -1.00 -0.04 1.00

10 nsh1m 11,2 4.59 0.60 6.01 0.00 0.00 0.00

11 nNWm 8,44 0.39 3.54 4.50 0.00 0.00 0.00

12 nNWf 7,91 0.57 6.59 0.74 -0.87 0.37 0.02

13 nAcToF 4,06 0.13 1.91 2.01 0.00 0.00 0.00

14 TrAcM 3,56 1.48 0.31 1.77 1.00 -1.00 -1.00

15 pAge 3,49 0.62 2.87 0.00 -0.33 0.33 -0.33

16 NWdurM 2,34 0.85 1.21 0.28 1.00 -1.00 -1.00

17 BTM 1,95 0.79 0.08 1.07 -1.00 1.00 1.00

18 nleim 1,94 0.79 0.08 1.08 -1.00 1.00 1.00

167

The hit ratio (based on a probabilistic assignment rule and the training set) of the model,

compared to a null-model (a root-only decision tree) indicates a significant

improvement achieved by the tree: the hit-ratio of the null-model of 0.432 is

significantly increased to 0.544, as shown in Table 8.6. The overall accuracy on the

validation set is almost the same, only dropped slightly to 0.526 indicating that no

overfitting occurs. A Chi-square-based contingency coefficient of 0.504 confirms that

there is a moderately strong impact of the decision tree structure on the action variable.

To evaluate the quantitative impacts of each condition variable on the action variable,

Table 8.7 displays the impact table of the car allocation model for non-work tours. The

condition variables are listed in order of decreasing impact on the action variable

overall (the IS column). Note that ISmale, ISfemale, and ISnone show the size of the impact

for each action separately. The last three columns (the MS column) identify the

monotonicity measure of the condition variable across the action variable. The

condition variable that has a monotonically increasing impact on the frequency of a

particular action variable across the levels of the condition variable signify MS equals

1 and otherwise -1 if it has a monotonically decreasing impact. Any value in between

these extremes indicates that the impact is nonmonotonous in the direction indicated by

the sign across the range of the condition variable.

When we look at the differential impacts of types of condition variable, we see that the

tour-level variables have the strongest impact. The variable that gives by far the biggest

impact is the longest distance (travel time) from home to a particular destination in a

tour by male and female (distM and distF). The monotonicity measure (MSmale = 1.00

and MSfemale = 1.00) clarifies that with the increasing travel time, the probability that

the male and female get the car increases monotonically.

The monotonicity measure for the variable that gives the next biggest impact is the

primary activity of a tour performed by male (AtourM) in a day. It indicates that as the

activity code (from 2 to 10) goes up, the probability of none of the heads using the car

increases nearly monotonically (MSnone = 0.71). Note that an increase of activity code

means approximately a transition from mandatory to a discretionary nature of the

activity. The next variable that gives a significant impact is the number of touring

activities performed by the female in a tour (ntouf). It shows that an increasing number

of touring episodes decreases monotonically the probability that the male or female

uses the car (MSmale = -1 and MSfemale = -1). In addition to that, several other variables

at the tour-level that gives influence to the car allocation decision are the number of

activity episodes in each tour, such as touring, shop to multi store, and leisure episodes

performed by male (ntoum, nsh1m, and nleim). The primary activity of the tour

performed by female (AtourF), the number of activity episodes in a tour of female

168

(nAcToF) and start time of doing non-work activities by male (BTM) are also

influential tour-level variables. As such, the number of non-work episodes of male and

female in a day (nNWm and nNWf) and the duration of non-work tour of male

(NWdurM) also have an influence on the car allocation decision.

In terms of situational variables, we find that the most outstanding variable is

urbanization (Urb). The increasing level of urbanization variable, which indicates a

decreasing level of urban density, the tendency of male or female getting the car

increases monotonically (MSmale = 1 and MSfemale = 1). As for socio-demographic

variables, the only influential variable is the age of oldest head (male/female) in

houseshold. The probability of getting the car increases non-monotonically for the

female (MSfemale = 0.33) and decreases non-monotonically for the female (MSfemale =

0.33) with increasing age of the oldest head.

In relation to accessibility measures, there are some variables that have significant

impact to the car allocation decision. Those are travel time ratio between car and bike

of the female (TTcbF), ratio # paid parking places to total # parking places by male

(RParkM), and accessibility of train in the location of male (TrAcM).

8.6 CONCLUSIONS

This paper considered car allocation choice behavior in car-deficient households

explicitly in the context of an activity-scheduling process. Focusing on non-work

activities, a car allocation model based on rules derived from a large travel diary data

set using a CHAID-based induction algorithm was presented. The face-validity of the

decision tree model is good in the sense that the derived rules and impacts of condition

variables are readily interpretable. The overall goodness-of-fit of the model is

satisfactory. Although the performance on a validation set decreased slightly, the set of

decision rules seems stable across training and validation set to a satisfactory extent.

In 56.43% of the car-allocation decision cases men gets the car while in only 13.26%

of the cases women use the car. In the remaining cases (30.30%), none of the heads

uses the car. With regard to work status, in 53.10% of the cases where a car allocation

decision is involved, the male is a non-worker or full-time worker and uses the car.

Meanwhile, in 55.72% of the cases, the female is non-worker and uses another

transport mode. The result indicates that men more often than women in gets the car

when their non-work tours overlap. This result is almost similar to our previous study

on car allocation decisions for work tours (Anggraini, et al., 2008), where we found

169

that the probability of the men getting the car to work location is also higher than that

of the women.

In terms of the activity on the tour, the percentage of female and male doing shopping

(to one store) tour is the highest among other activities in car-allocation cases. It is

followed by social tour (24.50% for female and 23.61% for male) and leisure tour

(21.56% for male and 18.40% for female).

In terms of decision rules, as the impact table analysis showed, longest distance (travel

time) from home to a particular destination in a tour of male and female have an

important role in car-allocation decisions to non-work tours in two-driver, single-car

households. It indicates that the probability that the male and female get the car

increases with increasing travel time to a destination location. In addition to that, some

other variables that relates to activity-level have significant influence to the decision of

allocating the car. Those variables are the primary activity of a tour performed by male

and female in a day, and the number of episode of a particular activity such as touring,

shop to multi store, and leisure episodes performed by male and female in a tour. The

start time of doing non-work activity by male and its duration also have great influence

to car allocation decision. The probability of the male to get the car tends to decrease

monotonically when the start of doing non-work activity increases. This result is in

contrary to that of the females. Female’s probability to get a car increases when the

male starts doing non-work activity later. Overall, men have more influence on the car

allocation decision for non-work tours, as indicated by the number of influential

variables that relates to the male in the impact table.

Socio-economic variables have less influence on car allocation decisions, only age of

the person has an impact. The increase of the age of the oldest person in the household

will increase slightly the probability of women to get the car. Situational variables also

have less influence on the decision of allocating the car. The most influential variable

in this class is urban size, where the increasing level of urban size (from the most

densely to the least densely areas) will also significantly increase the probability of the

men and women to get the car. Slightly different from those two variable types,

accessibility measures variables have moderate influence to car allocation decision.

In general, the results indicate that gender still plays a role in car-deficient household’s

car allocation decisions. Men are still prevailing women in getting the car to non-work

tour. The decision from car allocation obviously is informative for transport mode

choice in the subsequent choice facet in the scheduling process. Hence, the car

allocation decision to non-work tour has proved that the model can be included as one

element in an activity scheduling process as what ALBATROSS done. ALBATROSS

170

proved to be a suitable framework for this. By understanding the car allocation

decision in car-deficient households, the mobility of the household members can also

be recognized well.

REFERENCES

Anggraini, R., Arentze, T.A., and Timmermans, H.J.P. (2008), “Car Allocation

between Household Heads in Car Deficient Households: A Decision Model”.

European Journal of Transportation and Infrastructure Research, 8(4), pp. 301-319.



Arentze, T.A. and Timmermans, H.J.P. (2003), “Measuring the Goodness-of-Fit of

Decision-Tree Models of Discrete and Continuous Activity-Travel Choice:

Methods and Empirical Illustrations”. Journal of Geographical Systems 5: 185-206.










Timmermans, H.J.P. (2006), Analyses and Models of Household Decision Making

Processes. In: Proceedings of the IATBR Conference, Kyoto (CD-ROM, 34 p.p).

171

Chapter 9

THE INTEGRATED MODEL

9.1 INTRODUCTION

ALBATROSS is a comprehensive activity-based model of transport demand, which

accounts for multiple facets of activity-travel behavior. As indicated, the aim of this

PhD project is to more systematically incorporate household decision making in the

model, replacing the earlier sequential approach. To that effect, based on a slightly

different scheduling process model, a new set of decision rules were derived from

empirical data. To judge the added value of this endeavor, ideally the heuristics

underlying the model should be compared. In that context, the model uses

individual/household decision trees that can be interpreted as decision rules to

predict/simulate a complete activity-travel pattern for each household head in a

synthetic population. Unfortunately, it is not possible to compare the model results at

the level of decision trees between the existing and new version because of the

involved change in the structure of the scheduling process. However, it is possible to

compare results based on the generated complete activity schedules; that is, on the

level of the integrated model.

In this chapter, we therefore evaluate the validity and sensitivity of the full, integrated

model in two respects. First, as a test of validity, we evaluate the extent to which the

172

model is able to reproduce observed frequency distributions and mobility indicators in

the MON dataset. We do not expect dramatic differences in that respect because both

versions derive the rules from the same data set. We expect that the new model is able

of reproducing the aggregated distributions as well as the existing model. Second, we

apply the model to a particular scenario of change in the Dutch population to evaluate

the sensitivity of the model. We apply both the existing and new version of the model

to a scenario using a synthetic population of the Netherlands. On this level, we do

expect a difference between the existing and new model. That is, we expect that the

new model will predict different effects due to the fact that it better takes into account

within-household interactions. The sensitivity analysis is also performed on the basis

of a set of relevant frequency and indicator variables of behavior.

The chapter is organized as follows. First, the results of the validity test using the

MON data will be reported. Predicted and observed aspects of activity-travel pattern

will be compared between the different versions of the model at the activity, tour and

schedule level. In addition, comparisons will be based on a set of performance

indicators that the model generates. Next, the sensitivity of the new version of the

model is compared to that of the old models, based on a scenario of increased

participation of women in the workforce. The chapter is completed with conclusions

and discussion.

9.2 TEST OF VALIDITY USING MON DATA

In order to test the performance of the comprehensive ALBATROSS model system,

we compare the goodness-of-fit of the predictions of the old and new version of the

integrated model on the Dutch National Travel dataset (MON). The old version uses

individual-based decisions and the new version uses household-based decisions where

appropriate. We expect that the result of the new version is not really different from

the old version in terms of frequency distributions and indicators. Both models should

be able to reproduce the aggregate distributions that are found in the MON data. In this

section, we compare the goodness-of-fit between the models.

9.2.1 Frequencies

To examine the validity of the model, the discrepancy between observed and predicted

data of the old version and new version of ALBATROSS system is analyzed. The Chi-

square ( 2

iχ ) measure can be used as a measure of difference between an observed and

predicted frequency distribution. It is defined as follows:

173

[9.1]

where i is an index of cell of the frequency table. Three Chi-square measures are used

to identify the differences between: (i) observed and predicted data of the old version

( 2

1χ ), (ii) predicted data of the old and new versions ( 2

2χ ), and (iii) observed and

predicted data of the new version ( 2

3χ ). As for (i) and (iii), the smaller the difference

the better the performance of the model is. Analysis (ii), on the other hand, indicates a

difference between the two models.

Table 9.1 displays the results of a frequency analysis of activity patterns generated by

ALBATROSS model for the MON sample for both the individual-based (old version)

and household-based (new version) decisions. The table illustrates the observed

frequencies in the MON data and predicted frequencies by the old version and new

version of ALBATROSS in terms of some variables. The variables shown here

represent the most relevant facets at the activity-level, tour-level, and schedule-level. Activity-level facets refer mainly to all main activity attributes. The frequency

distribution across activity types is fairly accurate. The discrepancy stays within a

range from 0 – 2 percent points. The only clear tendency in both models is that the

frequency of work activities is somewhat underpredicted. This is a known bias that is

due to the fact that ALBATROSS imposes the restriction of maximally two work

episodes per person. The Chi-square value that measures the discrepancy between

observed data and prediction of the new version ( 2

3χ = 145.32) proves a low

dissimilarity (given the large sample size we have), meaning that the new version

model predicts the activity type distribution accurately. Furthermore, the accuracy is

even somewhat better than the old version ( 2

3χ = 145.3 versus 2

1χ = 232.8). The time-

of-day distribution displays a relatively high dissimilarity which is caused by a shift

from day-time time slots to evening. Also, this bias is known and has been reported

before. The temporal constraints imposed on schedules cannot be fully accounted for in

the decision trees so that during scheduling a certain proportion of activities are shifted

from blocked to open time slots where open time slots are more likely to occur in the

evening. The old version and new version show similar predictions, as indicated by a

relatively low Chi-square value ( 2

2χ = 526.06). The bias is slightly stronger in the new

version as indicated by the higher value of the Chi-square. The explanation for this is

that joint activities have a higher probability to find an only feasible time slot in the

evening compared to independent activities, as they have to meet temporal constraints

of both persons at the same time, resulting in a somewhat larger shift.

∑−

=i

iii

ected

ectedobserved

exp

)exp( 2

2χ

174

TABLE 9.1 Some Relevant Variables at the Aggregate Level

Predicted Data (%) Activity Type

Observed

Data (%) Old Version New Version

Work 20.47 18.88 18.39

Business 5.80 6.43 5.79

Bring-Get 7.96 8.30 8.65

Shop-1 store 20.92 22.53 21.61

Shop-n store 4.07 4.46 3.97

Service 5.28 5.65 5.08

Social 13.17 11.67 13.86

Leisure 12.90 12.51 12.92

Tour 8.04 8.16 8.27

Other 1.39 1.41 1.47

Total 82,584 76,842 78,812 2

1χ = 232.83; 2

2χ = 250.29; 2

3χ = 145.32

Activity Time of Day

<=10 am 29.03 25.79 24.69

10-12 am 16.71 13.99 12.88

12-2 pm 15.23 12.93 11.96

2-4 pm 15.63 16.72 15.01

4-6 pm 9.84 11.84 12.29

> 6 pm 13.55 18.73 23.16

Total 82584 76842 78812 2

1χ =1334.11; 2

2χ = 526.06; 2

3χ = 3267.36

Trip-Chain Pattern

Single-stop 63.61 63.25 69.45

After-stop 13.44 15.54 13.06

Before-stop 13.44 15.54 13.06

In-between stop 9.52 5.67 4.43

Total 82584 76842 78812

2

1χ = 1015.18; 2

2χ = 678.59; 2

3χ = 1700.95

Activity Location

Home Zone 30.059 32.107 36.251

Home Municipality 29.475 25.797 25.043

Municipality order1 14.876 16.300 14.867





Total 82584 76842 78812

2

1χ = 1754.39; 2

2χ = 684.59; 2

3χ = 1430.83

175

TABLE 9.1 (cont.)

Predicted Data (%) First Tour Mode

Observed

Data (%) Old Version New Version

Car 43.31 47.31 44.55

Slow 42.86 38.09 40.23

Public 3.10 3.15 3.14

Car Passenger 10.73 10.66 11.31

Total 63627 60544 65027

2

1χ = 791.33; 2

2χ = 101.02; 2

3χ = 568.02

Number of Activity in a Tour

1 82.56 80.27 84.17

2 10.19 14.55 12.07

3 4.55 3.71 2.66

4 1.48 1.03 0.73

> 4 1.22 0.43 0.38

Total 63627 60544 65027

2

1χ = 831.31; 2

2χ = 348.20; 2

3χ = 887.54

Number of Tour in a Schedule

0 18.10 21.81 19.04

1 45.41 42.18 42.85

2 24.38 24.48 24.38

3 8.52 8.34 9.19

> 3 3.59 3.18 4.55

Total 46876 46593 46593

2

1χ = 229.46; 2

2χ = 221.47; 2

3χ = 109.83

Number of Non-Work Activity in a Schedule

0 31.58 34.94 31.90

1 30.69 28.60 32.17

2 19.24 17.99 17.75

3 9.91 9.88 9.43

4 4.84 5.16 4.75

> 4 3.74 3.42 3.99

Total 46876 46593 46593

2

1χ = 144.63; 2

2χ = 195.79; 2

3χ = 54.84

176

In terms of trip-chaining, both models predict the frequencies of the so-called After

stops and Before stops rather accurately but underpredict the Between-stops somewhat.

The underprediction is slightly bigger in case of the new model. Also, this can be

understood in terms of the increased difficulty of finding a feasible in-between time

slot due to additional constraints that joint activities bring along.

The last variable that is taken into account in this class is activity location. The activity

location that is the same as the home zone (Home Zone) is slightly overpredicted by

both models and a little more so by the new model. The frequency of other location

types (outside the home zone and within own municipality and outside the home

municipality in municipalities of different order) are predicted accurately. Again, this

slight difference between the old version and new version can be attributed to increased

constraints that joint activities must meet compared to independent activities. All in all,

the new version predicts location type frequencies slightly better than the old version

( 2

3χ = 1430.83 versus 2

1χ = 1754.39).

At the tour-level, a first variable considered is the transport mode of the first link of the

tour. Here, the old version and new version shows a similar prediction as indicated by a

low value of the Chi-square measure ( 2

2χ = 101.02). However, the prediction of the new

version seems somewhat better than the prediction of the old version as indicated by a

lower Chi-square value ( 2

3χ = 568.0 versus 2

2χ = 791.3). In terms of number of activities

in a tour, the two models perform approximately equally ( 2

3χ = 887.5 versus 2

2χ =

831.3). In both cases there is a slight underprediction of the multiple-activities tours

that might be related to the (imposed) underprediction of work activity episodes.

At the schedule-level, the prediction of the new model in terms of the number of tours

in a schedule shows a satisfying result ( 2

3χ = 109.83). It accurately predicts the

frequency distributions of schedules across numbers of tours on a day. Compared to

the old model the prediction is even more accurate. Finally, regarding the number of

activities in a schedule, the new model also shows an improvement in accuracy of the

prediction as indicated by the lower Chi-square value ( 2

3χ = 54.8 versus 2

2χ = 144.6). In

overall, the new model shows equal or better predictions in the frequency distributions

of the relevant variables, than the old model, except for time of day and trip-chaining.

9.2.2 Indicators

In addition to frequency distributions of relevant variables, we also calculated a set of

relevant indicators to examine the validity of the model. Again, we use the MON

177

sample and evaluate the dissimilarity between observed and predicted data of the old

version and new version of ALBATROSS. We consider the system total as well as the

mean across schedules, standard deviation, difference in means and t-value of

differences in means for each indicator. The significance of differences between means

is based on a two-sided independent samples t-value. The t-value is defined as follows:

[9.2]

where

Similar to what we did in case of frequency analysis, there are three aspects that we

want to compare. The three t-values are used to identify the differences between: (i)

observed and predicted data of the old version (t-value1), (ii) predicted data of the old

and new versions (t-value2), and (iii) observed and predicted data of the new version

(t-value3). As for (i) and (iii), smaller differences indicate better performance of the

model. Analysis (ii), on the other hand, indicates a difference between the two models.

Table 9.2 displays the observed values in the MON data and predictions of the old and

new version of ALBATROSS for the same sample in terms of a number of indicators

that are generally of interests. In terms of total travel time, the old and new model

show small differences in prediction, as signified by a low t-value (t-value2 = -2.1).

Although both models show an underprediction of average travel times (which is

known to the developers), the new model seems somewhat better than the old model

(t-value3 = 41.3 and t-value1 = 42.3). In terms of travel time for each mode, the new

version of ALBATROSS shows a better prediction than the old version for all

transport modes, except car driver. The prediction of number of tours and number of

trips by the new model are also more accurate than the old model. In terms of distance

traveled by each transport mode, the new model’s predictions are good for every

transport mode except slow modes (the latter is mainly due to some outliers of very

long slow mode travel times in the MON data). Only in terms of total distance, the

prediction of the old version is slightly better than the new version. Overall, for most

indicators, the new model performs slightly better than the old model.

2

2

2

1

2

1

21

n

S

n

S

XXt

+

−=

1X is the mean of sample 1 n1 = size of sample 1

2X is the mean of sample 2 n2 = size of sample 2

2

1S is the variance of sample 1

2

2S is the variance of sample 2

178

9.3 TEST OF SENSITIVITY

Another and perhaps more interesting test is whether the potentially improved

decisions mechanism at the household level make the model more sensitive to evaluate

policy scenarios that should be expected to influence household decision making. The

results of such a test are described below.

9.3.1 Synthetic Population

ALBATROSS has been developed to analyze the impacts of possible scenarios on

activity patterns and related travel demand. To that end, first a synthetic population

needs to be constructed for the whole Netherlands. The synthesis agent uses two sets

of data, namely national population statistics by zone (1308 zones) and a national

sample of households. The population statistics define the marginals and the sample

data the initial proportions of a multiway household attribute table that is generated

and fitted using an iterative proportional fitting method (IPF). Generated populations

by zone (1308 zones) are then allocated to the post code areas (3987 areas) within the

zone proportional to the known population sizes in postcode areas (Arentze and

Timmermans, 2005).

The results of the population synthesis procedure replace existing observed schedules.

The new set of observed schedules specifies for each case the day of the week, an

empty schedule for each person-day and household and person level data. It should be

noted that the synthesis module takes much computation time, because an IPF

procedure is run 1308 times, each for each zone, and involves fitting the data on both a

household and person level.

9.3.2 Scenario

In order to evaluate the sensitivity of the new ALBATROSS model, we develop a

scenario on the level of the synthetic population. The scenario considered here

involves an increase of women participation in the labor force of 65 % overall (labor

scenario) assuming the year 2000 as the base year. The scenario considered here

involves an increase of women participation in the labor force of 65 % overall (labor

scenario) assuming the year 2000 as the base year. The increase of 65% of labor

participation of women led to an increase of 41% of labor participation of women

household heads (apparently the increase is stronger for women that are not household

heads). So, this relates to women that are household heads and hence, occur in the

179

ALBATROSS population. This is a relatively strong increase, but it should be noted

that, in the scenario, the labor participation rate of women is still substantially less than

that of men. The ratio of part-time workers was not changed in the scenario meaning

that a much larger proportion of working women are part-time workers compared to

men. Due to correlations, the scenario population will also display differences in other

socio-demographic characteristics. Table 3 shows the differences between the baseline

and labor scenario for household composition, presence and age of children, car

possession, age of person, and work status of person. As side effects, there are shifts

towards higher income levels, no children in the household and an increase in car

possession. There are no noticeable differences in age distribution as age is a variable

that is constrained by population data in the synthesis. The differences between

populations will be discussed in more detail below.

To identify predicted effects, we compare the prediction under the scenario with the

prediction of the baseline for each of the two model versions. The results are displayed

in Table 9.3 and Table 9.4 for the old and new version respectively. Comparison with

the baseline reveals the effects of the scenario that each model predicts. In turn,

comparison of predicted effects between models reveals the extent to which the

models differ. An increase of sensitivity of the new model would emerge as a

difference in predicted effects. This means that a difference in predicted effects is

evidence for an improved sensitivity of the model (and a better prediction). We use the

same set of attributes and indicators as before for this analysis. Furthermore, we use a

standard functionality of ALBATROSS to reveal the variance of stochastic variation in

predictions. For each prediction, ALBATROSS calculates the mean and standard

deviation between subsets of the set of predicted schedules. The subsets are

determined based on a random partitioning of the set (in three subsets). Each table

shows information about the mean across subsets of the base-scenario (m0), difference

in means between base-scenario and labor-scenario as a percentage of m0 (m1-m0 %)

and the t-value of differences in means for each variable/indicator. The significance of

the differences between means is based on a two-sided, independent samples t-test.

Degrees of freedom are corrected for possible differences between the two

distributions. The significance of difference of means on a variable/indicator is

represented by a star symbol, where one star indicates significance at 5% alpha level

and two stars significance at the 2.5% level.

Table 9.3 illustrates the comparison between baseline and labor scenario in terms of

socio-demographic variables. As expected, with the increasing labor participation of

women, single-1-worker and double-2-worker households increase significantly with

14.32% and 32.67% respectively. With regard to household SEC (income), medium

and high income households increase with 3.15% and 7.19%, respectively.

180

TABLE 9.2 Observed and Predicted of the Old and New Versions

OBSERVED PREDICTED (OLD VERSIONS) PREDICTED (NEW VERSIONS) INDICATORS

Total N Mean Stdev Total N Mean Stdev Total N Mean Stdev t-value1

t-value2

t-value3

Total travel time 2997661 46876 63.9 71.0 2128941 46593 45.7 60.8 2167099 46593 46.5 57.4 42.3 -2.1 41.3

Travel time car

driver 1466223 46876 31.3 54.8 1085816 46593 23.3 40.7 1051020 46593 22.6 38.8 25.3 2.9 28.1

Travel time

public 249738 46876 5.3 30.8 210930 46593 4.5 28.1 215128 46593 4.6 27.4 4.2 -0.5 3.7

Travel time slow 873736 46876 18.6 39.5 601229 46593 12.9 38.6 633901 46593 13.6 35.1 22.4 -2.9 20.6

Travel time car

passenger 407964 46876 8.7 31.0 216631 46593 4.6 17.4 254214 46593 5.5 19.1 24.7 -6.7 19.3

Number of tours 63585 46876 1.4 1.0 60521 46593 1.3 1.0 65027 46593 1.4 1.1 8.3 -13.7 -5.7

Number of trips 146115 46876 3.1 2.4 137145 46593 2.9 2.4 143839 46593 3.1 2.5 11.1 -8.9 1.9

Ratio trips-tours 2.29795 2.26607 2.21199

Ratio single stop

tours - all tours 0.82559 0.80441 0.84167

Total travel

distance 1812815 46876 38.7 82.4 1632628 46593 35.0 64.6 1622956 46593 34.8 61.8 7.5 0.5 8.1

Distance car

driver 1115245 46876 23.8 62.7 1200032 46593 25.8 57.0 1138415 46593 24.4 53.6 -5.0 3.7 -1.7

Distance car

passenger 331312 46876 7.1 35.4 226063 46593 4.9 23.3 277702 46593 6.0 26.1 11.3 -6.9 5.5

Distance slow 223210 46876 4.8 34.3 79349 46593 1.7 6.9 80604 46593 1.7 6.5 18.9 -0.6 18.8

Distance public 143048 46876 3.1 24.2 127184 46593 2.7 24.1 126235 46593 2.7 22.5 2.0 0.1 2.2

181

TABLE 9.3 Comparison between Base-line and Scenario on Socio-Demographic

Characteristics

Household Composition m0 m1-m0 (%) sign t-value df

Single, 0-worker 52134 -11.81 ** -31.889 4

Single, 1-worker 42460 14.32 ** 88.292 3

Double, 1-worker 39852 -27.73 ** -104.422 3

Double, 2-worker 60545 32.67 ** 62.848 3

Double, 0-worker 32931 -26.2 ** -96.458 3

Total 227922 0.01 0.07 4

Household SEC

Minimum 61086 -7.45 ** -18.438 3

Low 55251 -2.7 ** -8.345 3

Medium 48576 3.15 ** 5.778 4

High 63009 7.19 ** 95.832 3

Total 227922 0.01 0.07 3

Presence and Age of Children

No child 164326 0.74 ** 6.743 4

< 6 yr 29190 -3.54 ** -5.845 4

6-<12 yr 17936 -1.58 ** -10.251 3

12-<17 yr 16470 0.71 1.013 2

Total 227922 0.01 0.07 3

Number of Cars in Household m0 m1-m0 (%) sign t-value df

No car 46327 -5.26 ** -9.764 4

One car 127312 -0.74 * -2.355 3

2 or more 54283 6.27 ** 20.098 3

Total 227922 0.01 0.07 3

Age of Person

< 35 yr 83562 -0.32 -0.697 2

35-<55 yr 156184 0.58 * 2.662 4

55-<65 yr 51397 1.56 ** 5.622 4

65-<75 yr 38943 -1.14 ** -12.128 4

75+ yr 31164 -2.8 ** -16.872 4

Total 361250 0.03 0.22 4

Person Work Status

Non-Worker 157847 -21.83 ** -196.857 3

Part Time Worker 53665 35.71 ** 89.985 2

Full Time Worker 149738 10.3 ** 28.384 3

Total 361250 0.03 0.22 4

182

TABLE 9.4 Predicted Scenario Effects on Some Variables/Indicators: Old Model

Version Activity Type m0 m1-m0 (%) sign t-value df

Work 114791 13.15 ** 30.473 2

Business 37685 9.89 ** 11.649 3

Bring-Get 50647 -4.98 ** -7.255 4

Shop-1-store 132147 -5.58 ** -12.04 2

Shop-n-store 26508 -6.54 ** -7.476 3

Service 30773 -1.33 -1.472 2

Social 72570 -0.66 -1.52 2

Leisure 76844 1.76 ** 3.99 4

Tour 47565 0.33 0.496 4

Other 9112 -6.84 ** -5.922 2

Total 598643 1.2 2.538 2

Activity Start Time

<= 10 am 154755 6.67 ** 12.981 2

10-12 am 81568 -4.39 ** -9.219 4

12-2 pm 78170 -3.08 ** -4.892 2

2-4 pm 97964 -2.69 ** -4.742 2

4-6 pm 72187 2.1 * 2.759 3

> 6 pm 113998 3.48 ** 7.455 3

Total 598643 1.2 2.538 2

Activity Trip Pattern

Single-stop 378913 1.16 * 2.996 2

After-stop 92716 0.9 1.371 3

Before-stop 92716 0.9 1.371 3

In-between stop 34298 3.28 ** 4.549 2

Total 598643 1.2 2.538 2

Activity Location

Home Zone 180365 -1.65 -2.86 2

Home Municipality 172704 1.7 ** 4.371 4

Municipality order1 92733 2.24 * 3.354 2

Municipality order2 55633 2.84 ** 11.294 2


Municipality order4 26996 2.16 1.861 3


Total 598643 1.2 2.538 2

183

TABLE 9.4 (cont.)

First Tour Mode m0 m1-m0 (%) sign t-value df

Car 219422 3.43 ** 6.726 2

Slow 184496 -1.34 ** -3.972 3

Public 17397 2.68 ** 5.748 4

CarPass 49528 -0.61 -1.05 4

Unknown 787 2.2 1.078 3

Total 471629 1.11 2.565 2


1 378913 1.16 * 2.996 2

2 68348 0.29 0.429 4

3 17259 1.91 1.845 2

4 4908 3.02 ** 8.55 2

> 4 2201 7.25 ** 6.347 2

Total 471629 1.11 2.565 2

Number of Tours in a Schedule

0 77177 -4.09 ** -12.867 2

1 153575 0.26 0.81 3

2 89193 3.41 ** 8.073 2

3 29699 0.97 1.699 4

> 3 11607 -3.82 ** -3.043 4

Total 361250 0.03 0.104 3

INDICATORS

Total travel time 17379946 -3.89 -0.625 2

Travel time car driver 7940591 4.56 ** 7.313 2

Travel time public 2867975 -36.74 -0.938 2

Travel time slow 4961321 0.15 0.352 3

Travel time car passenger 1588751 0.45 0.545 4

Number of tours 471629 1.11 2.565 2

Number of trips 1070272 1.16 2.55 2

Ratio trips-tours 2.269 0.05 1.864 3

Total travel distance 11586597 4.09 ** 7.014 2

Distance car driver 8409780 4.73 ** 6.799 2

Distance car passenger 1587591 1.01 1.046 3

Distance slow 672783 1.11 * 2.457 4

Distance public 916444 5.79 ** 7.887 3


184

TABLE 9.5 Predicted Scenario Effects on Some Variables/Indicators: New Model

Version

Activity Type m0 m1-m0 (%) sign t-value df

Work 115089 12.79 ** 32.421 3

Business 35932 10.78 ** 17.52 4

Bring-Get 53238 2.6 ** 6.752 4

Shop-1-store 133379 -3.15 ** -4.387 4

Shop-n-store 25328 -3.81 ** -6.136 3

Service 30111 2.2 ** 9.94 3

Social 89070 -1.21 ** -6.987 4

Leisure 81842 -0.69 -1.12 2

Tour 49668 -3.53 ** -6.355 3

Other 9506 -3.15 * -2.143 4

Total 623163 1.89 ** 7.448 2

Activity Time of Day

<= 10 am 152278 6.87 ** 42.423 2

10-12 am 78937 -3.38 ** -9.712 4

12-2 pm 74125 -1.4 ** -3.819 4

2-4 pm 94292 -1.83 * -3.021 3

4-6 pm 76311 3.14 ** 7.582 4

> 6 pm 147220 2.95 ** 10.241 3

Total 623163 1.89 ** 7.448 2

Activity Trip-Chain Pattern

Single-stop 428113 0.94 * 4.287 2

After-stop 83041 3.31 ** 8.936 4

Before-stop 83041 3.31 ** 8.936 4

In-between stop 28969 7.82 ** 8.038 4

Total 623163 1.89 ** 7.448 2

Activity Location

Home Zone 208622 -0.4 -0.912 4

Home Municipality 176818 2.31 ** 5.825 4






Total 623163 1.89 ** 7.448 2

185

TABLE 9.5 (cont.)

First Tour Mode m0 m1-m0 (%) sign t-value df

Car 225592 4.28 ** 13.694 3

Slow 207409 -1.31 ** -5.519 4

Public 18526 1.4 1.946 4

CarPass 57839 -0.78 * -2.483 4

Total 511154 1.32 ** 6.051 2


1 428113 0.94 * 4.287 2

2 62941 2.12 ** 10.25 4

3 14011 5.8 ** 5.137 4

4 4074 9.56 ** 8.777 3

> 4 2014 10.66 ** 4.806 2

Total 511154 1.32 ** 6.051 2

Number of Tours in a Schedule

0 64541 -6.46 ** -28.859 3

1 156377 1.24 ** 7.4 2

2 90293 2.12 ** 5.146 3

3 33563 1.53 ** 3.333 4

> 3 16475 -0.5 -0.832 4

Total 361250 0.03 0.22 4

INDICATORS

Total travel time 18044438 -3.38 -0.526 2

Travel time car driver 7820339 5.08 ** 21.663 4

Travel time public 2917420 -36.33 -0.944 2

Travel time slow 5320978 0.94 2.71 2

Travel time car passenger 1922107 0.2 0.387 3

Number of tours 511154 1.32 ** 8.975 2

Number of trips 1134317 1.63 ** 11.374 2

Ratio trips-tours 2.219 0.31 ** 7.856 3

Total travel distance 11872101 4 ** 12.347 4


Distance car passenger 2036742 0.36 0.536 3

Distance slow 708375 2.03 ** 4.718 3

Distance public 922991 5.2 ** 6.485 2

186

As for the presence and age of children, household composition changes only slightly.

It indicates that households with children aged under 12 years decreases with 1.58%

(6-11 years) and 3.54% (<6 years). As opposed to that, households with children over

12 years of age do not change significantly and no-child households show a small

increase of 0.74%. These results indicate that with the increasing labor participation of

women, the tendency of having children decreases, and hence, no-child households

increases.

In terms of car ownership, the prediction concludes that the possession of 2 or more

cars increase with 6.27%, whereas the number of households with no cars decreases

about 5.26%. In connection with person age, there are no noticeable changes as we

would expect since this variable is constrained by zonal population data in the

synthesis. In terms of work status of the person, as expected, the number of non-

workers among household heads decreases strongly (21.83%). On the other hand, the

number of part time workers increases significantly with 35.71%. Also the number of

full-time workers increases, all be it less substantial, with 10.3%. This reflects the fact

that in the baseline a relatively large proportion of women workers are part time

workers and this is maintained in the scenario.

Table 9.4 and Table 9.5 illustrate the comparison between the baseline and labor

scenario for the old version and new version respectively. Concerning the activity-level facets, both versions predict considerable shifts in frequency distributions as

consequences of the scenario. However, we are interested here in the differences in

prediction made by the old version (Table 9.4) and the new version (Table 9.5). In

predicting the number of work activities, both versions predict similar effects as we

would expect. However, in terms of household tasks activities (bring-get, service,

shop-1-store, and shop-n-store), the predictions of both versions are quite different.

The old version predicts a decrease of activities for all household task activities. The

new version predicts a slight increase in bring-get and service activities of 2.6% and

2.2% and smaller decreases of the other household activities compared to the old

version. The explanation might be that by making explicit allocation decisions

considering both schedules of the spouses, the new model might be better able to find a

time slot in either one of the two schedules for including a task activity. Since the old

version does not consider schedules of the spouses in combination it may fail to find a

time slot in the schedule of the person that is primary responsible for the task and omit

rather than re-allocate the activity. On the other hand, for non-task discretionary

activities, social, leisure and touring, the two model versions also predict rather

different effects. Note that non-task discretionary activities are relatively often

performed jointly in the baseline. When labor participation of women increases

according to the scenario, the models predict opposite effects. The old version predicts

187

no change or an increase depending on the specific type of discretionary activity. The

new version in contrast predicts a decrease at least for the social and touring activity

(the decrease of the leisure activity is not significant). Also this difference can probably

be attributed to a specific strength of the new model. With increasing work time of the

female, there will be fewer opportunities to find a time slot where the activities can be

conducted jointly. Given a preference to conduct them jointly, a decrease in

opportunities will lead to a decrease in these activities. This effect is predicted by the

new model. The old model treating activities independently does not impose the

requirement of finding a common time slot of (a subset of) the activities across the

schedules and, therefore, finds in more cases opportunities to schedule the activities.

In terms of time of day, there are no significant differences in predictions between the

old version and the new version. Both models predict an increase of activities with a

start time before 10 am of around 6-7%. This is an expected effect of an increase in

work activities, given that work activities tend to start at early time moments of a day.

In terms of trip-chaining, the new version predicts a stronger increase of activities on

an in-between stop (7.82% versus 3.48%). This result is consistent with the prediction

of the new model that more household-task activities are maintained in the scenario

and a tendency that these activities are combined with work activities. For example,

females tend to make multiple stops from home to work and stop by at school.

Regarding locations of activities, both versions again show similar results. The number

of activities conducted in the same postcode area where the person lives decreases as a

consequence of the scenario, whereas the choice of destinations outside the own

municipality slightly increases. Thus, the prediction points out that people tend to

travel longer from home when labor participation of women increases.

Regarding transport mode choice for tours, the two models predict more or less the

same effects. There are only slight differences which may not be significant. The new

version predicts a slightly stronger increase in car driver mode (4.28% versus 3.43%),

whereas the old version predicts a slightly stronger increase in public transport mode

than the new version do (2.68% versus 1.4%). These predictions are plausible, given

the increase in income, car possession, work activities and distance to destinations.

Furthermore, both models predict an increase of tours where multiple activities are

combined (more than 4). The new version predicts a slightly stronger increase of such

complex tours, which is consistent with the earlier finding that this model predicts a

stronger increase of activities conducted on in-between stops.

In terms of the prediction of indicator variables, the old and new versions give similar

results but at the same time display some notable differences. Total travel time

188

decreases about 3 - 4%. Percentage-wise, the models predict a small increase in travel

time for car driver (5.08% and 4.56%) and a strong decrease of travel time by public

transport (36 - 37%). The models predict different effects regarding the number of trips,

number of tours and ratio trips-tours. The old version predicts that there are no effects

on these variables. In contrast, the new version predicts a small but significant increase

in each of these variables. This difference reflects the differences that we saw in terms

of number of activities, trip-chaining and number of activities per tour. Hence, the

specific sensitivity of the new model is visible even at the level of aggregate mobility

indicators. As for distance traveled across all modes, the two models both predict an

increase of around 4 %. Based on the increase of work activities alone, one may have

expected a stronger increase in mobility. We should realize, however, that the increase

takes place primarily in the part time worker segment which is characterized by

relatively short home-work distances. Furthermore, it should be noted that non-work

activities decrease in this scenario. Distance traveled by public transport increases 5 -

6%, and travel distance by car driver increases 4 - 5%. According to those predictions,

there is a tendency of people traveling longer distances by car and public transport.

Finally, the model predicts different mobility effects for the weekend days and

weekdays (not shown) that can be interpreted too as a result of an increased sensitivity

of the new model.


This chapter considered the validity and sensitivity of the full ALBATROSS model by

comparing the performance of the old version and the new version. A validity test on

the basis of the MON data set established that the new version is able to predict

choice-facet frequency distributions and mobility indicators observed in the MON data

as accurately as the old version. The goodness-of-fit of the new version for most

choice facets appeared to be either equivalent or slightly better than the goodness-of-fit

of the old version. The only exceptions are time of day and trip-chaining. For these

facets the old model produced better results. The bias in time-of-day predictions in the

new version is probably due to the inclusion of joint activities. Joint activities might be

more feasible to do in the evening compared to independent activities, as a

consequence of coupling constraints, resulting in a fairly larger shift towards evening

hours. In relation to trip-chaining, the new model underpredicted the in-between stops.

This can be understood in terms of the increased difficulty of finding a feasible in-

between time slot due to additional constraints that joint activities bring along.

With regard to performance indicators, the new version of ALBATROSS shows better

predictions than the old version, for all transport modes, except car driver, in terms of

189

travel time for each mode. The prediction of the number of tours and number of trips by

the new model are also more accurate than the old model. In terms of distance traveled

by each transport mode, the new model’s predictions are better for every transport

mode except slow modes. Only in terms of total distance, the prediction of the old

version is slightly better than the new version.

In terms of a test of sensitivity, the new model proved to be more sensitive to the

impacts on situational and decision dimensions of activities, such as activity type, start

time, trip-chaining, location, etc. The scenario involved an increase in work activity

load in women’s schedules and the new model predicted somewhat different responses

that could be interpreted in terms of better representing opportunities and requirements

related to task allocation and joint activity participation. In sum, by considering

decisions of household heads on these dimensions in interaction, the system is able to

predict with increased sensitivity processes of activity re-scheduling in response to a

change. The results showed that this can lead to differences in prediction of activity

generation and travel choices that have an impact on aggregate mobility indicators (e.g.,

number of trips, shift in timing and transport mode) that are relevant for planning and

policy making. Other scenarios could be considered as well, but this case served our

purpose to evaluate the working of new model.

190

Chapter 10

CONCLUSIONS AND DISCUSSION

Household members traditionally share household resources and responsibilities. Since

demand for travel is derived from the necessity of individuals to perform out-of-home

activities, in particular in multi-person households, the activities of one person may

contribute to satisfying household needs and may therefore imply that other household

members do not need to make additional trips. The vast majority of studies and models

in activity-based analysis have been based on individual decision making. However, it

is more realistic to model activity-travel patterns from the perspective of a household to

capture the interrelationships and interdependencies of the activity-travel patterns of

household members. A focus on the household as the decision-making unit is relevant

for at least three decision problems: (i) the problem of allocating limited resources to a

household member; (ii) the problem of task allocation and (iii) the problem of joint

activity participation. A good example of the resource allocation problem is a

household with fewer cars than driver licenses. In that case, the household members

need to decide who will use the car to conduct a particular activity. It automatically

means that other household members cannot use the car at the same time. An example

of task allocation is a husband dropping off a child at school, while his spouse may be

responsible for grocery shopping. Both child care and grocery shopping satisfy

household needs, but only one household member needs to conduct these activities.

191

Joint activity participation requires synchronization of activity-travel patterns in time

and space. An example is a husband and wife intending to go for lunch together during

their work time who then need to decide on the timing, duration and location of the

activity that are suitable for both of them. Therefore, the activity-travel patterns of

persons in the same household are strongly interconnected.

This thesis argued that there is a need for incorporating household decision making in

urban travel demand modeling. The daily activity-travel patterns of individuals often

reflect interactions with other household members, in particular in the form of joint

activity participation, and task and resource allocation. Explicit representation of

household decision making in comprehensive activity-based models is a research

frontier. Although some models focusing on a limited number of aspects have been

suggested, most operational comprehensive models are based on individuals.

Considering that the household is the basic unit of decision making, modeling of

household behavior is important for researchers and policy makers alike. Logically,

representing intra-household interactions is motivated by the need to bring more

consistency in addressing the interdependencies that characterize activity-travel

behavior.

The aim of this thesis is to include aspects of household decision making in a more

rigorous manner in the ALBATROSS model. The original model already incorporated

aspects of household decision-making, albeit in an implicit way. The activity-travel

patterns of household adults were sequentially developed, incorporating the schedule of

the spouse. In this thesis, a more integral approach to household decision making was

systematically incorporated in the system. Using decision tree induction, as in the

original model, a series of decision tables were derived from the MON data. The model

considers all choice facets in generating the daily activity-travel patterns at the

individual level and household level. Facets that are analyzed at the individual level are

activity type selection, activity (tasks) allocation, start times, durations, trip-chaining,

locations, and transport modes. The household level considers joint activity selection,

car allocation to work tour, start times, durations, locations, and car allocation to non-

work tours. Note that all household level facets (except car allocation to work tour)

apply to non-work activities.

The household perspective requires some modification of the structure of the

scheduling process, underlying ALBATROSS. This new, household level version of

the model was compared to the old version by assessing the goodness-of-fit and the

sensitivity of the models. This comparison test was based on the complete generated

activity schedules. A full activity-travel pattern is generated on a continuous time scale

for each simulated individual, and hence, detailed information about trips, activities

192

and socio-demographic characteristics can be derived. To determine whether the new

model predicts choice-facet frequency distributions and mobility indicators as accurate

as the earlier version, a test of validity using the MON data set was conducted. The

results demonstrate that, overall, the predictions of the new version model are

equivalent or slightly better than the old model for most aspects. Only for some

attributes, the old model outperformed the new model.

The benefits of a comprehensiveness transport demand model, such as ALBATROSS,

are however not restricted to prediction. It also means an increase in range and detail in

information that can be generated and presented to the user for assessing the impact of

scenarios. Furthermore, the systematic and better handling of household decision

making implies that the model was expected to be more sensitive to scenarios that

likely impact the household decision making. To test this, a reasonable scenario on the

basis of the synthetic population was applied involving an increase of 41 % in labor

participation of women household heads (labor-scenario). The new model proved to be

more sensitive to the impacts on situational and decision dimensions of activities.

Specifically, the new model predicted different responses that could be interpreted in

terms of a better representation of opportunities and requirements related to task

allocation and joint activity participation. By considering decisions of household heads

on these dimensions, the system is able to predict with increased sensitivity

rescheduling processes of households in response to a change. In that sense, the

ALBATROSS system has become a more powerful operational system to predict and

assess urban travel demand due to an improved inclusion of household decision making

in the system.

The current version of ALBATROSS represents an operational model that can be

applied to asses the impacts of several policies. Furthermore, although ALBATROSS

is a comprehensive model and unique in its class, there are some limitations that have

not been implemented yet. Therefore, it is recommended that the following topics are

further developed and/or examined in future research:

1. adding the travel decision The decision to travel is closely related to the activity participation decision. The

decision of conducting activity independently or jointly was examined in this study.

In fact, the travel decision could be explicitly modeled as well. In doing so, a better

connection between activity and travel decisions may be established.

2. adding ride sharing decision The current version adds a new dimension in the system, i.e. car allocation decision.

We examined it in the context of work tours and non-work tours. It is also useful to

193

model the ride sharing decision between household heads that use their own car.

Although we model car driver and car passenger, this decision is not linked to

specific persons.

3. elaborate within-household interactions by adding the presence of other

household member (if any) Currently we are only concerned with the interaction between two household heads

in performing out-of-home activities. However, to some extent, the existence of

other household members, such as children, should also be taken into consideration.

Hence, joint participation in activity and travel may be more precise and more

sensitive by including other household members.

4. adding time allocation for spouses

Spouses could coordinate their work times, especially if they have small children.

For example, there may be evidence of turn-taking behavior or synchronization of

their work times. Such behavior has not yet been implemented in ALBATROSS.

5. incorporating in-home activities

The current Dutch National data (MON) includes information about out-of-home

activities only. In order extent within-household interactions, complementary

survey data on in-home activities allows an extension of the model. In the present

case, trip data was converted to activity data so that the in-home stay periods are

known as well. The only thing missing is the differentiation of the stay-at-home

duration by activity type (what type of activities is performed at home). Hence, the

in-home and out-of-home activities could be addressed explicitly.

Time and data limitations did not allow incorporating these aspects into the model.

However, those features could be implemented in future research to further enhance the

sensitivity of the model.

194

SUMMARY

Although the importance of households as a decision making unit has been recognized

in seminal work in activity-based analysis of transport demand, most comprehensive

models have relied on individual activity-travel patterns. The transformation of these

models to household level models and the explicit consideration of resource allocation,

task allocation and joint activity participation decisions is thus a challenge and research

frontier in this field of study. To contribute to this expanding field, the aim of this PhD

study is to develop such an activity-based model. More specifically, the slightly ad hoc

treatment of household decisions in the ALBATROSS model is replaced with a

systematic incorporation of household decisions. The new variant, based on the MON

2004 data, is compared in terms of goodness-of-fit and sensitivity with the previous

version of the model.

To this end, the thesis is organized as follows. Chapter 2 provides a review of past

research efforts concerning the determinants of household decision making. We discuss

how household decision making has been treated in comprehensive activity-based

models of transport demand. This line of research started with analytical studies on

household decision making taking into account car allocation and usage decisions.

Further literatures addressed task and time allocation decisions. They found that

household types, defined by the number of household heads and work status, strongly

influence activity time allocation and trip chaining. The presence of children in the

household has a positive effect on the duration of all out-of-home activities in

household trip chaining, except for the duration of out-of-home discretionary activities

of households having children under 5 years old. This suggests that the presence of

children induces more chaining of trips and more time allocated to these trip chains.

Households having more children of 16 years of age and over are more likely to spend

time in trip chaining for out-of-home subsistence activities. Finally, they found that

flexible work arrangements tend to be correlated with less trip chaining for the work

trip.

In addition to these studies, there is also a literature on joint activity participation.

Several studies have examined the effect of household attributes on joint activity-travel

behavior. They found that joint activities involving household heads are significantly

affected by the presence of children. Couples without children living at home are more

likely to pursue joint out-of-home non-work activities than couples with children. In

households with children, most joint activities between adults are at home. In addition,

the employment status of the household heads influences whether a joint activity

originates from home or from an out-of-home contact point. In additional to analytical

studies, existing comprehensive activity-based models are reviewed in this chapter in

195

terms of their inclusion and treatment of household decisions. Comprehensive in this

context means that the model allows predicting a combination of choice facets, at least

compatible with those underlying traditional four-step models: i.e. activity generation,

destination and transport mode choice. The discussion is restricted to fully operational

models. Over the years, many activity-based models have been suggested in the

literature, including constraints-based models, micro-simulation models, (nested logit)

utility-maximizing models, suites of advanced statistical models and rule-based models.

Most of these models either do not incorporate household decisions at all, or only in a

limited way.

Chapter 3 discusses the conceptual framework of this thesis for modeling household

activity-travel behavior. Because the thesis is an attempt of elaborating the

ALBATROSS model, we discuss this model in more detail, including its conceptual

framework. Further, we explain the entire process underlying the ALBATROSS

system and the inclusion of household decision making in the process, such as joint

participation, activity allocation, car allocation for non-work tours, and some other

choice facets. Household decision making is mostly applicable to non-work activities,

but the problem of car allocation is highly relevant for work tours in car-deficient

households. Further, we summarize the methodology that was used in this study:

decision tree induction using a CHAID-based induction algorithm being the core

method of ALBATROSS.

The remaining chapters then present the results of the various derived decision tress for

the sequential choice facets that together make up the ALBATROSS model. Chapter 4 describes the results for car allocation choice focusing on work activities. In this

analysis work-tours as opposed to work trips are considered. The car allocation model

focuses on car-deficient households (i.e., more drivers than cars present) and a joint

decision between the two heads (mostly, a female and male). We also assume that both

male-female are drivers and at least one of them has a work activity on the day

considered. Furthermore, the model includes the option that none of the household

heads uses the car, but some other means of transport. The results show that the

propensity of men driving a car to the work place is higher than that of women,

particularly, when women have no work activity or women’s work place is in the same

zone as the home location. This finding is consistent with the common notion that

women use a slow or public transport mode more often to travel to activity locations.

Women tend to use the car when men have no work activities or men work at home.

Chapter 5 reports the empirical derivation of a household decision model of activity

choice taking into account joint participation and task allocation between household

heads. These are considered household-level decisions given that they involve

196

commitments of multiple persons, in particular the two-head households. Of the 10

activity categories concerned, 7 activity categories (non-work activities) are used in

this study, i.e. bring/get, shopping to 1 store, shopping to multiple store, service-related,

social, leisure, and touring. The first four activities are deemed task allocation activities

and the rest are non-task activities (discretionary). Hence, two decision trees were

derived from diary data. The activity participation model, given the large number of

observations that could be derived from the data, included more than 300 condition-

action rules. The household task allocation model also involved an extensive set of

decision rules, involving more than 90 condition-action rules. In both cases, the

validity of the decision tree is satisfactory in the sense that the derived rules are readily

interpretable and the overall goodness-of-fit of the model on a validation set is

acceptable as well.

Chapter 6 focuses on the joint participation of male-female heads in non-work

activities and attempts to model the timing and duration decisions for these activities,

using decision tree induction. Decision tree results indicated that there were 17 and 31

condition-action rules derived for the duration model and start time model, respectively.

The improvement in S-value (a measure of prediction accuracy) relative to a null

model as well as an F-statistic indicates that there is a moderately strong association

between condition variables at household, individual, activity and schedule level, on

the one hand, and the decision, on the other. The S-value shows a more substantial

improvement in the start-time model compared to the duration model. The results show

that activity type has the most significant influence in both models. In addition, time

availability for non-work activities during morning off-peak periods has a strong

influence on start time decisions. The results also suggest that there is a substantial

influence of duration decisions on start time decisions. Joint participation of household

members in activities tends to lead to longer activity duration and earlier start times.

Overall, modeling timing and duration of joint activity participation decisions at the

household level proves to have some clear advantages.

Chapter 7 discusses the development of the household location choice model taking

into account the independent and joint activities, in particular non-work activities. In

ALBATROSS, location choice is modeled for independent and joint activity

participation of the household heads based on the concept of detour time. The detour

time of a candidate location for an activity is defined as the extra travel time required

to implement the activity in the context of the current activity schedule. There were

two decision tree models for both independent and joint activity categories. The first

model relates to the decision whether or not the activity is performed at the same

location as the previous activity, whether the activity is done at the same location as the

next activity, or whether it is conducted elsewhere. The second model relates to the last

197

choice option in the first model and comprises 25 choice alternatives. It verifies the

location in terms of a combination of size - distance classes. The size class depends on

a particular activity type and the size of available facilities at the activity location. Size

is classified into 5 categories based on employment in the relevant sector for the

activity considered and distance is classified in terms of a detour travel time (by car)

also into 5 categories. The tendency of conducting a particular activity at the same

location as the previous activity is higher for independent activities than for joint

activities. The same condition also applies to activities that are conducted at the same

location as the next activity. These results imply that males and females are more likely

to conduct multiple activities at one particular location independently than jointly.

These results do make sense, since the activity-travel behavior of one person is

different from the other person, even though male-female couples live in the same

household.

Chapter 8 is concerned with car allocation behavior for non-work activities. In this

study, the assumption is similar to the assumption in Chapter 4 where tours are taken

into account instead of trips. Travel for any activity episode or set of chained activity

episodes that does not include a work activity is considered a non-work tour. The

problem of modeling this allocation problem for non-work tours is more complex than

for work tours because the decision at this stage depends considerably on the outcome

of the previous stages in the scheduling process. Hence, the car can be allocated to

male, female or none. Further, only overlapping non-work activities of the male’s and

female that occur in the same time slot are taken into account. Overlapping tours are

defined as a pair of tours conducted by respectively male and female of which the start

and/or end times of each tour (simulating use of a car for the tour) defines a fully or

partially overlapping episode. As a tour consists of a sequence of trips that starts and

ends at a particular location (i.e., home), the primary activity in each tour needs to be

determined. In order to identify the primary activity in a particular tour, we consider a

hierarchical order of activity priority. In particular, 10 activity categories are

considered in order of priority starting from work, business and other (mandatory)

activities. A group of non-work activities is considered, such as escorting, shopping

(daily and non-daily), service-related, social, leisure, and touring. Since business and

other mandatory activities are not considered primary work activities, they are not dealt

with in the first stage of the scheduling process and, hence, they are also considered as

non-work activities in this model.

The results show a satisfactory improvement in goodness-of-fit of the decision tree

model compared to the null model. Gender seems to play an important role. A

descriptive analysis indicates that men more often than women get the car for non-

work tours for which a car allocation decision needs to be made. Tour-level attributes

198

are shown to influence the household car allocation decision for non-work tours. The

decision to allocate the car is considerably influenced by the longest distance (travel

time) from home to a particular location in a tour of men and women. The probability

that the men and women get the car monotonically increases with increasing travel

time. Socio-economic and situational factors have less influence on the car allocation

decision. Overall, men have more influence on the car allocation decision for non-work

tours, as indicated by the number of influential variables that relates to the males in the

impact table.

Chapter 9 discusses the results of the integrated model of ALBATROSS. In order to

test the performance of the ALBATROSS system based on all decision tables for the

assumed scheduling process, the validity and sensitivity of the integrated model were

evaluated and compared with the performance of the old model. First, the validity of

the model versions was compared by evaluating the extent to which the model is able

to reproduce observed frequency distributions and mobility indicators in the MON

dataset. In that sense, no major differences were expected. Instead, it was expected that

the new model is able to reproduce the aggregated distributions as well as the existing

model. A Second effort was to examine the sensitivity of the models by applying the

models to a particular scenario of change in the Dutch population. It was expected that

the new model was more sensitive to such scenarios. The scenario assumed an increase

of 41 % in labor participation of women household heads (labor scenario) assuming

the year 2000 as the base year. A fraction of 10% of the Dutch population in the year

2000 was generated using the synthesis module of ALBATROSS for the baseline and

the labor scenario. As expected, in the context of validity test, the new model showed

equal or slightly better goodness-of-fit for most choice facets, except for time of day

and trip-chaining. The new model proved to be more sensitive to facets such as activity

type, start time, trip-chaining, location, etc., in response to scenarios change. In

particular, the new model predicted somewhat different responses that could be

interpreted in terms of the better representation of opportunities and requirements

related to task allocation and joint activity participation. In sum, by considering

decisions of household heads in interaction, the system is able to predict with

increased sensitivity activity-travel rescheduling processes of households in response

to change.

199

APPENDIX:

DECISION TREES

TABLE AI-1.1 Car Allocation to Work Tour

TTbcF 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 2 2 2

TTcM 0 0 1 1 1 2 2 2 2 3-4 3-4 3-4 3-4 - - - - - - -

Day 0,3 1,4,2,5,6 - - - - - - - - - - - - - - - - - -

TTbcM - - 0-2 0-2 3-4 0-1 2-4 2-4 2-4 0-3 4 - - 0 0 1 2-4 - - -

DurM - - 0-2 3-4 - - - - - - - - - - - - - - - -

SEC - - - - - - 0-1 2-3 2-3 - - - - 0,1,2 3 - - - - -

Child - - - - - - 0 1,3,2 - - - - - - - - - - -

PParkM - - - - - - - - - 0-1 0-1 2-4 2-4 - - - - - - -

TTptM - - - - - - - - - - - 0-1 2-4 - - - - 0 0 1-4

TTcF - - - - - - - - - - - - - - - - - 0-1 2-4 -

TrAcF - - - - - - - - - - - - - - - - - - - 0

Male 0.075 0.267 0.539 0.351 0.315 0.735 0.614 0.547 0.32 0.728 0.536 0.291 0.535 0 0 0.613 0.262 0 0 0.442

Female 0 0 0 0 0 0 0 0 0 0 0 0 0 0.78 0.574 0.333 0.584 0.22 0.465 0.299

None 0.925 0.733 0.461 0.649 0.685 0.265 0.386 0.453 0.68 0.272 0.464 0.709 0.465 0.22 0.426 0.053 0.154 0.78 0.535 0.259

N 93 146 128 94 184 83 88 75 103 419 110 55 86 109 54 75 149 59 86 147

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20

200

TABLE AI-1.2 Car Allocation to Work Tour

TTbcF 2 3 3 3 3 3 3 4 4

TTbcM - 0 0 1 2-4 2-4 2-4 - -

TTptM 1-4 - - - - - - - -

TTcF - 0-1 2-4 - 0-1 0-1 2-4 - -

TrAcF 1 - - - - - - - -

TTcM - - - - 0-1 2-4 - - -

Mwork - - - - - - - 0 1

Male 0.296 0 0 0.712 0.272 0.547 0.378 0 0.474

Female 0.574 0.203 0.43 0.106 0.141 0.156 0.378 0.5 0.308

None 0.13 0.797 0.57 0.182 0.587 0.297 0.243 0.5 0.218

N 54 177 79 66 92 64 111 50 78

R21 R22 R23 R24 R25 R26 R27 R28 R29

201

TABLE AII-1.1 Household Activity Participation

HHact 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

nbr 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2

Child 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

ModeF 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 0,3,4 1 2 - -

ModeHH - - 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 0,1,3,4 2 - - - -

Dist1 - - 0-2 0-2 0-2 0-2 0-2 0-2 3 4-5 - - - - - - - - - -

SEC - - 0-2 0-2 0-2 0-2 0-2 0-2 0-2 0-2 3 3 3 3 - - - - - -

AgeF 0-2 3-4 - - 0-2 0-2 3-4 - - - - - - - - - - - - -

DrivF - - 0 1 1 1 1 - - - - - - - - - - - -

Ncar 0 0 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 - - - -

yOthM - - 0 0 0 0 0 0 0 0 0 0 0 0 1 - - - - -

Dist3 - - - - 0-1 2-5 - - - - - - - - - - - - - -

Time2F - - - - - - - - - - 0-1 0-1 2-4 2-4 - - - - - -

WstatF - - - - 0 0 0 1-2 - - - - - - - - - - - -

yWorkF - - - - - - - - - - - - 0 1 - - - - - -

Time5F - - 0-2 3-4 3-4 3-4 3-4 3-4 - - - - - - - - - - - -

Time1C - - - - - - - - - - 0 1-4 - - - - - - - -

Yes 0.022 0 0.092 0.019 0.108 0.042 0.02 0.017 0.071 0.035 0.063 0.011 0.06 0.102 0.118 0.153 0.101 0.268 0.436 0.259

No 0.978 1 0.908 0.981 0.892 0.958 0.98 0.983 0.929 0.965 0.937 0.989 0.94 0.898 0.882 0.847 0.899 0.732 0.564 0.741

N 274 402 98 483 194 284 342 354 393 2720 207 370 1979 147 119 85 159 138 427 205


202


HHact 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Day - 0,2,3,4 0,2,3,4 0,2,3,4 0,2,3,4 0,2,3,4 0,2,3,4 1 1 5,6 - - - - 0,1,3 2,5,6,4 0,5,6 1,4,2 1,4,2

nbr 3-5 0 0 0 0 0 0 0 0 0 1 1 1 1 1 2 2 2 2 2

Child 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

DurHH - - - - - - - - - - 0-3 4 - - - - - - - -

ModeF - - - - - - - - - - 0,4,1,2 0,4,1,2 3 - - - - - - -

NworkHH - - - - 0 1-4 - - - - - - - - - - - - -

AgeF - 0 0 0 0 1-4 1-4 - - - - - - - - - - - - -

AgeM - - - 0 1-4 - - 0 1-4 - - - - - - - - - - -

Urban - - - - - - - - - - 0-3 0-3 0-3 4 4 - - - - -

Time1F - 0-2 3-4 3-4 3-4 - - - - - - - - - - - - - - -

Comp - - 2 3,4 3,4 - - - - - - - - 2 3,4 2,4 2,4 3 3 3

Time3F - - - - - - - - - - - - - - - - - - 0-1 2-4

Yes 0.5 0.556 0.437 0.246 0.37 0.476 0.656 0.42 0.638 0.061 0.855 0.961 0.765 0.819 0.723 0.683 0.407 0.412 0.184 0.356

No 0.5 0.444 0.563 0.754 0.63 0.524 0.344 0.58 0.362 0.939 0.145 0.039 0.235 0.181 0.277 0.317 0.593 0.588 0.816 0.644

N 90 180 279 211 146 105 445 119 232 512 186 102 81 215 285 189 108 85 87 146


203


HHact 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Day 3 0,5,6,1,3 0,5,6,1,3 0,5,6,1,3 2,4 - - - 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 5 6 - - -

nbr 2 3 3 3 3 4 4 5 0 0 0 0 0 0 0 0 0 1 1 2

Child 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2

SEC - - - - - - - - - - - 0-2 3 - - - - - - -

DurF - - - - - - - - - - - - - 0-1 2-4 - - 0-1 2-4 -

NworkHH - 0-1 0-1 2-4 - - - - - - - - - - - - - - - -

AgeF - 0 1-4 - - - - - - - - - - - - - - - -

Dist2 - - - - - - - - 0-2 0-2 3-4 3-4 3-4 5 5

Comp 3 - - - - - - - - - 2,4 3 3 - - - - - - -

Time3C - - - - - 0-1 2-4 - - - - - - - - - - - - -

nEmp3 - - - - - - - - 0-3 4-5 - - - - - - - - - -

Yes 0.6 0.885 0.711 0.627 0.487 0.133 0.331 0.456 0.271 0.495 0.595 0.348 0.547 0.421 0.255 0.211 0.118 0.729 0.601 0.25

No 0.4 0.115 0.289 0.373 0.513 0.867 0.669 0.544 0.729 0.505 0.405 0.652 0.453 0.579 0.745 0.789 0.882 0.271 0.399 0.75

N 85 78 97 75 78 90 124 114 166 109 131 89 95 356 102 152 178 373 153 12.8


204


HHact 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2

Day - - 0,6,2,1,3,4 0,6,2,1,3,4 0,6,2,1,3,4 0,6,2,1,3,4 5 - - 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1

nbr 3 4-5 0 0 0 0 0 1 2-5 - - - - - - - - - - -

Child 2 2 3 3 3 3 3 3 3 - - - - - - - - - - 0

nsh1 - - - - - - - - - 0-1 0-1 0-1 0-1 0-1 0-1 2 3-5 0 0 1-5

DurHH - - - - - - - - - 0-1 0-1 0-1 0-1 0-1 0-1 0-1 0-1 2-3 2-3 2-3

Dist1 - - - 0-3 0-3 4-5 - - - 0 1-5 - - - - - - - - -

DurF - - - - - - - - - - - - - - - - - 0-2 3-4 -

DrivF - - 0 1 1 1 - - - - - - - - - - - - - -

Dist2 - - - 0-1 2-5 - - - - - - - - - - - - - - -

DrivM - - - - - - - - - - - - - 0 1 - - - - -

Dist3 - - - - - - - - - 0 0 1-4 1-4 5 5 - - - - -

WstatF - - - - - - - - - - - 0-2 1 - - - - - - -

Yes 0.601 0.25 0.011 0.108 0.035 0.136 0.218 0.481 0.327 0.283 0.535 0.542 0.471 0.351 0.471 0.308 0.451 0.485 0.265 0.255

No 0.399 0.75 0.989 0.892 0.965 0.864 0.782 0.519 0.673 0.717 0.465 0.458 0.579 0.649 0.529 0.692 0.549 0.515 0.744 0.745

N 153 128 95 213 200 610 170 135 98 226 127 1106 126 114 1173 409 237 709 82 141


205


HHact 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

Day 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 2,3 2,3 2,3 2,3 2,3

Child 1,3,2 - - - - - - - - - - - - - - - - - - -

nsh1 1-5 1-5 - - - - 0 0 0 0 0 0 1-5 1-5 1-5 0 0 0 0 0

DurHH 2-3 2-3 4 4 4 4 4 4 4 4 4 4 4 4 4 0-2 0-2 0-2 0-2 0-2

Dist1 - - - - - - - - - - - - - - - 0 0 1-4 5 5

Time4F - - 0 0 0 0 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 - - - - -

DrivM - - - - - - - - - - - - - - - - - - 0 1

Dist3 - - - - - - - - - - - - - - - 0 1-5 - - -

Urban - - - - - - 0-2 3-4 - - - - - - - - - - - -

Time1F - - - - - - - - 0 1-3 4 - - - - - - - - -

SizePop - - - 0-2 3-5 - - - 0-2 0-2 0-2 3-5 - - - - - - - -

Time4M - - - - - - - - - - - - 0-1 0-1 2-4 - - - - -

NworkM - - 0 1 1 2 - - - - - - - - - - - - - -

Comp 2 3,4 - - - - - - - - - - - - - - - - - -

Time4C - - - - - - - - - - - - - - - - - - - 0-1

Time5C - - - - - - 0-1 0-1 2-4 2-4 2-4 2-4 - - - - - - - -

yWorkF - - - - - - - - - - - - 0 1 - - - - - -

Yes 0.113 0.328 0.383 0.117 0.218 0.305 0.64 0.458 0.398 0.234 0.433 0.508 0.201 0.107 0.301 0.372 0.514 0.581 0.379 0.671

No 0.887 0.672 0.617 0.883 0.782 0.695 0.36 0.542 0.602 0.766 0.567 0.492 0.799 0.893 0.699 0.628 0.486 0.419 0.621 0.329

N 106 247 107 392 147 95 86 168 93 141 252 122 214 140 103 199 257 1032 95 85


206


HHact 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

Day 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3 2,3

nsh1 0 1 1 1 1 1 1 2 2 2 3-5 0 0 1-5 1-5 0 0 0 0 0

DurHH 0-2 0-2 0-2 0-2 0-2 0-2 0-2 0-2 0-2 0-2 0-2 3 3 3 3 4 4 4 4 4

Dist1 5 - - - - - - - - - - - - 0-3 4-5 - - - - -

SEC - - - - - - - - - - - - - - - 0-2 3 - - 0-2

Time3M - - - - - - - - - - - - - - - 0-2 0-2 3-4 - -

Time4F - - - - - - - - - - - - - - - 0 0 0 1-3 4

AgeF - - - - - - - - - - - 0 1-4 - - - - - - -

yBusiM - 0 0 1 - - - - - - - - - - - - - - - -

yWorkM - 0 0 0 1 - - - - - - - - - - - - - - -

AgeM - - - - - - - 0-3 0-3 4 - - - - - - - - - -

Ncar - - - - - 0-1 2 - - - - - - - - - - - - -

Dist2 - 0-3 4-5 - - - - 0-1 2-5 - - - - - - - - - -

DrivM 1 - - - - - - - - - - - - - - - - - - -

NworkM - - - - - - - - - - - - - - - - - - - 0-1

Comp - 2,3 2,3 2,3 2,3 4 4 - - - - - - - - - - - - -

Time4C 2-4 - - - - - - - - - - - - - - - - - - -

Yes 0.526 0.576 0.414 0.32 0.349 0.576 0.43 0.398 0.274 0.167 0.438 0.351 0.525 0.379 0.196 0.118 0.24 0.387 0.369 0.43

No 0.474 0.424 0.586 0.68 0.651 0.424 0.57 0.602 0.726 0.833 0.562 0.649 0.475 0.621 0.804 0.882 0.76 0.613 0.631 0.57

N 745 132 169 75 209 604 86 133 376 84 324 131 316 153 153 186 175 142 111 402


207


HHact 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

Day 2,3 2,3 2,3 2,3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4

nsh1 0 0 1-5 1-5 0 0 0 0 0 0 0 1 1 1 1 2 2 2 3-5 3-5

DurHH 4 4 4 4 - - - - - - - - - - - - - - - -

Dist1 - - - - - 0-1 2-5 2-5 2-5 - - - - - - - - - - -

SEC 3 - - - - - - - - - - - 0-2 0-2 3 0-1 2-3 2-3 - -

Time4F 4 4 - - - - - - - - - - - - - - - - - -

AgeF - - - - 0 1-2 1-2 1-2 1-2 3 4 - - - - - - - - -

DrivF - - - - - - - 0 1 - - - - - - - - - - -

yWorkM - - - - - - - - - - - 0 1 1 1 - - - 0 1

NworkF - - - - 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Time4M - - 0 1-4 - - - - - - - - 0 1-4 - - - - - -

NworkM 0-1 2 - - - - - - - - - - - - - - - - - -

WstatM - - - - - - - - - - - - - - - - 0 1,2 - -

Dist4 - - - - - - 0-2 3-5 3-5 - - - - - - - - - - -

Yes 0.528 0.648 0.191 0.307 0.509 0.542 0.786 0.747 0.624 0.559 0.411 0.565 0.245 0.443 0.467 0.299 0.557 0.381 0.52 0.263

No 0.472 0.352 0.809 0.693 0.491 0.458 0.214 0.253 0.376 0.441 0.589 0.435 0.755 0.557 0.533 0.701 0.443 0.619 0.48 0.737

N 288 88 456 257 220 236 98 87 545 245 90 536 106 97 152 197 97 160 252 76


208


HHact 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3

Day 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 6 6 6 6 0

nsh1 0 0 0 1-5 0-1 0-1 0-1 0-1 2 2 2 2 3 4-5 0 0 0 1-5 0

nshn - - - - - - - - - - - - - - - - - - - 0

SEC - - - - - - 0-1 0-1 2-3 - 0-2 0-2 3 - - - - - - -

Time3M - - - - - 0 1-4 1-4 1-4 - - - - - - - - - - -

DurF 0-2 0-2 3-4 - - - - - - - - - - - - - - - - -

DrivF - - - - - - - - - 0 1 1 1 - - - - - - -

NworkF 1 1 1 1 2 - - - - - - - - - - - - - - -

Ncar - - - - - - - - - - 0-1 2 - - - - - - - -

Dist3 - - - - - - - - - - - - - - - 0 1-5 - - -

Time4M 0-3 4 - - - - - - - - - - - - - - - - - -

WstatM - - - - - - 0 1,2 - - - - - - - - - - - -

Comp - - - - - - - - - - - - - - - 2,3 2,3 4 - -

Time3F - - - - - - - - - - - - - - - - - - - 0-2

Time5F - - - - - - - - - - - - - - - - - - - 0-3

Yes 0.429 0.596 0.311 0.302 0.491 0.364 0.503 0.604 0.64 0.232 0.306 0.446 0.495 0.556 0.461 0.132 0.057 0.026 0.385 0.013

No 0.571 0.404 0.689 0.698 0.509 0.636 0.497 0.396 0.36 0.768 0.694 0.554 0.505 0.444 0.539 0.868 0.943 0.974 0.615 0.987

N 105 146 193 288 112 99 469 369 1623 95 173 92 212 232 254 167 935 501 135 233


209


HHact 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3

Day 0 0 0 0 0 0 0 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3

nsh1 0 0 0 0 1-5 1-5 1-5 0 0 0 0 0 0 1 1 1 1 2-5 2-5 2-5

nshn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

DurHH - - - - - - - - - - - - - - - - - - - 0-3

ModeF - - - - - - - 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2

Dist1 - - - - - - - 0 1-3 4-5 - - - - - - - - -

SEC - - - - 0-2 3 - - - - - - - - - 0-1 2-3 - - -

yWorkM - - - - - - - - - - - - - - - - - 0 0 1

Ncar - - - - 0-1 0-1 2 - - - - - - - - - - - - -

Dist3 - - - - - - - - - - - - - - - - - 0 1-5 -

nEmp2 - - - - - - - 0 1-4 1-4 1-4 1-4 5 - - - - - - -

SizePop - 0 0 1-5 - - - - - - - - - 0 1-5 - - - - -

WstatM - - - - - - - - 0 0 0 1,2 - 0,1 0,1 2 2 - - -

Time3F 0-2 3-4 3-4 3-4 - - - - - - - - - - - - - - - -

Time1M - 0-2 3-4 - - - - - - - - - - - - - - - - -

Time5F 4 - - - - - - - - - - - - - - - - - - -

Yes 0.067 0.119 0.039 0.15 0.039 0.088 0.015 0.114 0.129 0.34 0.218 0.16 0.074 0.069 0.151 0.024 0.083 0.154 0.062 0.012

No 0.933 0.881 0.961 0.85 0.961 0.912 0.985 0.886 0.871 0.66 0.782 0.84 0.926 0.931 0.849 0.976 0.917 0.846 0.938 0.988

N 89 151 204 501 413 170 273 990 170 191 147 545 216 276 403 209 472 78 612 165


210


HHact 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3

Day 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 4 4 4 4 4 4 4 5 5 5 5 5

nsh1 2-5 - - - - - - - 0 0 0 1-5 1-5 1-5 1-5 0 0 0 1 1

nshn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

DurHH 4 - - - - - - - - - - - - - - - - - - -

ModeF 0,1,2 3,4 3,4 3,4 3,4 3,4 3,4 3,4 - - - - - - - - - - - -

Time4F - - - 0-1 0-1 2-4 2-4 2-4 - - - - - - - - - - - -

DurF - - - - - - - - 0-2 0-2 3-4 - - - - - - - - -

AgeF - - - - - - - - - - - - - - - - - - 0 1-4

yWorkM 1 0 0 1 1 1 1 1 - - - - - - - - - - - -

Dist2 - - - - - - - - - - - - - - - 0-1 2-4 5 - -

DrivM - - - - - - - - 0 1 - - - - - - - - - -

Dist3 - - - - - 0-2 3 4-5 - - - - - - - - - - - 0-2

Urban - - - - - - - - - - - 0-2 3-4 3-4 3-4 - - - - -

nEmp2 - - - - - - - - - - - 0-3 0-3 4-5 - - - - -

SizePop - 0-2 3-5 - - - - - - - - - - - - - - - - -

nEmp1 - - - - - - - - - - - - 0 1-5 - - - - - -

Time5F - - - 0-2 3-4 - - - - - - - - - - - - - -

Yes 0.071 0.109 0 0.012 0.048 0.021 0.138 0.066 0.106 0.211 0.082 0.133 0.094 0.04 0.146 0.235 0.407 0.172 0.321 0.261

No 0.929 0.891 1 0.988 0.952 0.979 0.862 0.934 0.894 0.789 0.917 0.867 0.906 0.96 0.854 0.765 0.593 0.828 0.679 0.739

N 84 331 102 427 208 189 87 319 94 693 144 362 351 299 103 183 204 233 84 111


211


HHact 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4

Day 5 5 5 6 0,3 1,4,2 5,6 - - - - - - 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3

Child - - - - - - - - - 0,3,2 1 - - - - - - - - -

nsh1 1 2-5 2-5 - 0 0 0 1-5 - - - - - 0 0 0 0 1-5 - -

nser - - - - - - - - - - - - - 0 0 0 0 0 0 0

nshn 0 0 0 0 1 1 1 1 1 1 1 2 3 0 0 0 0 0 0 0

DurHH - - - - 0 0 0 0 1 2-4 2-4 - - 0-1 0-1 0-1 0-1 0-1 0-1 0-1

Dist1 - - - - - - - - - - - - - 0 0 0 0 0 0 1-4

AgeF 1-4 - - - - - - - - - - - - - - - - - - -

yBusiM - - - - - - - - - - - - - - - - - - - 0

Dist2 - - - - - - - - - - - - - 0-2 0-2 0-2 0-2 0-2 3-5 -

Dist3 3-5 - - - - - - - - - - - - - 0 0 1-5 - - -

Urban - - - - - - - - - - - - - - 0 1-4 - - - -

nEmp2 - - - - - - - - - - - - - 0-3 4-5 4-5 4-5 - - -

Comp - 2 3,4 - - - - - - - - - - - - - - - - -

Yes 0.109 0.147 0.075 0.019 0.566 0.391 0.711 0.369 0.348 0.165 0.049 0.153 0.35 0.146 0.015 0.086 0.139 0.178 0.215 0.231

No 0.891 0.853 0.925 0.981 0.434 0.609 0.289 0.631 0.652 0.835 0.951 0.847 0.65 0.854 0.985 0.914 0.861 0.822 0.785 0.769

N 201 156 415 1594 106 215 159 317 92 334 123 476 117 158 132 81 79 304 144 1515


212


HHact 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4

Day 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3

nsh1 - 0 0 0 1-5 1-5 - - 0 1-5 1-5 1-5 - - - - - - - -

nser 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

nshn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

DurHH 0-1 0-1 0-1 0-1 0-1 0-1 2-3 2-3 2-3 2-3 2-3 2-3 4 4 4 4 4 4 4 4

Dist1 1-4 5 5 5 5 5 - - - - - - - - - - - - - -

SEC - - - - - - 0-1 0-1 2-3 2-3 2-3 2-3 - - - - - - - -

yBusiM 1 - 0 1 - - - - - - - - - - - - - - - 0

AgeM - - - - - - - - - 0 1-4 1-4 - - - - - - - -

Dist3 - - - - - - - - - - - - 0-4 5 - - - - - -

DurM - - - - - - - - - - - - - - - - - - - 0-1

NworkM - - - - - - - - - - 0-1 2 - - - - - - - -

Time3F - - - - - - 0-1 2-4 - - - - - - - - - - - -

nEmp1 - - - - - - - - - - - - 0-1 0-1 2 3-5 3-5 3-5 - -

Time3C - - - - - - - - - - - - 0-3 0-3 0-3 0-3 0-3 0-3 4 -

Time5C - 0-3 4 4 - - - - - - - - - - - - - - - -

WstatF - - - - 0-2 1 - - - - - - - - - - - - - -

yWorkF - - - - - - - - - - - - - - - 0 1 - - -

Time5M - - - - - - - - - - - - - - - 0-1 0-1 2-4 - -

Time1C - - - - - - - - - - - - 0-1 0-1 0-1 0-1 0-1 0-1 0-1 2-4

Yes 0.142 0.06 0.172 0.076 0.184 0.317 0.01 0.111 0.124 0.281 0.153 0.256 0.016 0.073 0.149 0.033 0 0.071 0.145 0.203

No 0.858 0.94 0.828 0.924 0.816 0.683 0.99 0.899 0.876 0.719 0.847 0.744 0.984 0.927 0.851 0.967 1 0.929 0.855 0.797

N 219 116 686 79 798 123 96 397 582 89 504 86 185 559 87 92 126 378 117 172


213


HHact 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4

Day 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 0,4,1,2,3 5 5 5 6 6 6 - - - - - - - - -

nbr - - - - - 0 0 1-5 - - - - - - - - - - - -

nsh1 - - - - - 0 1-5 - - - - - 0 0 0 1 2-5 - - -

nser 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1

nshn 0 0 0 0 0 0 0 0 0 0 0 1-3 - - - - - - - -

DurHH 4 4 4 4 4 - - - - - - - - - - - - - - -

AgeF - - - - - - - - - - - - 0-2 3-4 - - - - - -

yBusiM 1 - - - - - - - - - - - - - - - - - - -

yOthM - - - - - - - - - - - - - - - - - - - -

Dist2 - - - - - - - - 0 1 2-5 - - - - - - - - -

ModeM - - - - - - - - - - - - 0,1 0,1 0,1 0,1 0,1 2,4,3 2,4,3 2,4,3

Dist3 - - - - - - - - - - - - 0-4 0-4 5 - - - - -

Urban - - 0-3 0-3 4 - - - - - - - - - - - - - - -

DurM 0-1 2-4 2-4 2-4 2-4 - - - - - - - - - - - - 0-3 0-3 0-3

Time1F - 0-3 4 4 4 - - - - - - - - - - - - - - -

nEmp1 - - - - - - - - - - - - - - - - - 0 1-2 3-5

WstatF - - 0,1 2 - - - - - - - - - - - - - - - -

Time1C 2-4 2-4 2-4 2-4 2-4 - - - - - - - - - - - - - - -

Yes 0.105 0.068 0.037 0.179 0.151 0.048 0.092 0.174 0.017 0.099 0.047 0 0.425 0.644 0.366 0.29 0.373 0.283 0.135 0.3

No 0.895 0.932 0.963 0.821 0.849 0.952 0.908 0.856 0.983 0.901 0.953 1 0.575 0.356 0.634 0.71 0.627 0.717 0.865 0.7

N 114 614 161 78 423 421 753 109 347 141 1058 1158 167 37 186 293 292 145 126 150


214


HHact 4 4 4 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

Day - - - 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 3,4 3,4 3,4 3,4 3,4 3,4

nser 1 2 3 - - - - - - - - - - - - - - - - -

DurHH - - - - - - - - - - - - - - 0 0 0 0 0 0

ModeHH - - - 0 0 0 0 1,3 1,3 1,3 1,3 1,3 1,3 2,4 - - - - - -

Time3M - - - - - - - 0 0 0 1-3 4 4 - - - - - - -

yBusiM - - - - - 0 1 - - - - - - - - - 0 0 0 1

DrivF - - - 0 0 1 1 - - - - - - - - - - - - -

AgeM - - - - - - - - - - - - - - 0 1-4 1-4 1-4 1-4 1-4

Ncar - - - - - - - - - - - - - - - 0 1-2 1-2 1-2 1-2

ModeM 2,4,3 - - - - - - - - - - - - - - - - - - -

Dist3 - - - 0-2 3-5 - - - - - - - - - - - 0 1-5 1-5 -

nEmp2 - - - - - - - - - - - - - - - - - 0-2 3-5 -

DurM 4 - - - - - - - - - - - - - - - - - - -

Time1F - - - - - - - 0 1-4 1-4 - - - - - - - - - -

Time4M - - - - - - - - 0-1 2-4 - - - - - - - - - -

Time5M - - - - - - - - - - - 0-2 3-4 - - - - - - -

Yes 0.105 0.198 0.417 0.016 0.065 0.085 0.017 0.038 0.007 0.036 0.054 0.039 0.005 0.031 0.271 0.034 0.051 0.111 0.164 0.02

No 0.895 0.802 0.583 0.984 0.935 0.915 0.983 0.962 0.993 0.964 0.946 0.961 0.995 0.969 0.729 0.966 0.949 0.889 0.836 0.98

N 95 494 168 307 460 1736 173 312 1212 169 129 128 662 1081 96 179 178 1067 347 100


215


HHact 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

Day 3,4 3,4 3,4 3,4 3,4 3,4 3,4 3,4 3,4 3,4 5 5 5 5 5 5 6 6 6 6

nsh1 - - - - - - - - - - 0 0 0 1 1 2-5 - - - -

DurHH 1-2 1-2 1-2 1-2 3-4 3-4 3-4 3-4 3-4 - - - - - - - - - - -

Dist1 - - - - - - - - - - - - - - - - - - 0 1-5

SEC - 0-1 2-3 - - - - - - - - - - - - - - - - -

NworkHH - - - - - - - - - - - - - - - - - - 0 0

DrivF - - - - - - - - - - - 0 1 - - - - - - -

Dist3 - - - - - - - - - - - - - - - - 0 1-5 - -

Urban - - - - - - - - - - - - - - - - 0 0 1-4 1-4

Time1F 0-1 2-4 2-4 2-4 - - - - - - - - - - - - - - - -

Comp - - - - - - 2,4 2,4 3 2 3,4 3,4 - - - - - - -

Time3F - - - - - - 0-3 4 4 4 - - - - - - - - - -

nsoc - - - - - - - - - - - - - - - - 0 0 0 0

WstatF - - - - - - - - - - - - - 0,1 2 - - - - -

Time5F - - - - 0-2 3 4 4 4 4 - - - - - - - - - -

Time2C - 0-2 0-2 3-4 - - - - - - - - - - - - - - - -

Yes 0.122 0 0.088 0.011 0.016 0.075 0.055 0.002 0.033 0.029 0.117 0.154 0.255 0.126 0.039 0.167 0.061 0.176 0.209 0.308

No 0.878 1 0.912 0.989 0.984 0.925 0.945 0.998 0.967 0.971 0.883 0.846 0.745 0.874 0.961 0.833 0.939 0.824 0.791 0.692

N 115 86 193 265 428 107 271 435 91 381 205 117 423 333 103 635 114 131 254 987


216


HHact 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6

Day 6 6 6 6 6 6 0 0 0 0 0 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3

nbr - - - - - - - - - - - - - - - - - - - 0

nsh1 - - - - - - - - - - - - 0 1-5 1-5 - - - - -

Dist1 - - 0-3 4-5 - - - - - - - - - - - - - - - -

SEC - - - - - - - - - - - 0-1 2-3 2-3 2-3 - - - - -

NworkHH 1-4 - - - - - 0 0 0 1-4 1-4 0 0 0 0 1-4 1-4 1-4 1-4 1-4

yBusiM - - - - - - - - - 0 1 - - - - - - - - -

AgeM - 0 1 1 2 3-4 - - - - - - - - - - - - - -

Dist2 - - - - - - - - - - - - - - - - 0-2 0-2 3-5 3-5

Dist3 - - - - - - 0-2 3-5 3-5 - - - - 0 1-5 - - - - -

Urban 1-4 - - - - - - - - - - - - - - - - - - -

DurM - - - - - - - - - - - - - - - - - - 0-1 2-4

Comp - - - - - - - 2,4 3 - - - - - - 2,4 3 3 3 3

nsoc 0 1-2 1-2 1-2 1-2 1-2 - - - - - - - - - 0 0 0 0 0

Time1M - - - - - - - - - - - - - - - 0 0 0 0 0

yWorkF - - - - - - - - - - - - - - - - 0 1 - -

Yes 0.147 0.267 0.211 0.08 0.233 0.037 0.029 0.063 0.148 0.005 0.027 0.042 0.123 0.12 0.045 0.008 0.083 0.031 0.062 0.004

No 0.853 0.733 0.789 0.92 0.767 0.963 0.971 0.937 0.852 0.995 0.973 0.958 0.877 0.88 0.955 0.992 0.917 0.969 0.938 0.996

N 95 75 90 113 103 82 308 492 81 1109 149 1365 600 83 619 760 120 194 81 525


217


HHact 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6

Day 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 4 4 4 4 5 5 5 5 5 5 5 5 5 6 6

nbr 1-5 - - - - - - - - - - - - - - - - - - -

Child - - - - - - - - - - - - - - 0 1,3,2 - - - -

nsh1 - - - - - - - - - - 0 0 1 1 2-5 2-5 2-5 2-5 - -

nshn - - - - - - - - - - 0 1-3 - - - - - - - -

DurHH - - - - - - - - - - - - - - - - - - 0 0

Dist1 - - - - - - - - - - - - - - - - - - 0 0

SEC - - - - - - - - - - - - 0-2 3 - - - - - -

NworkHH 1-4 1-4 1-4 1-4 1-4 - - - - - - - - - - - - - - -

DrivF - - - - - 0 1 0 1 1 1 1 1 1 1 1 - -

yWorkM - - - - - 0 0 1 - - - - - - - - - - - -

AgeM - - - - - - - - - - - - - - 0-1 0-1 2-4 2-4 - -

Dist2 3-5 - - - - - - - - - - - - - - - - - - -

nlei - - - - - 0 0 0 1-2 - - - - - - - - - - -

DurM 2-4 - - - - - - - - - - - - - - - - - - -

Comp 3 3 - - - - - - - - - - - - - - - - - -

nEmp1 - - - - - - - - - - - - - - - - - - 0-1 2-3

Dist4 - - - - - - - - - - - - - - - - 0-3 4-5 - -

nsoc 0 0 0 0 1-2 - - - - - - - - - - - - - - -

Time1M 0 1 2 3-4 - - - - - - - - - - - - - - -

Yes 0.035 0.039 0 0.026 0.115 0.035 0.086 0.032 0.21 0.044 0.194 0.074 0.033 0.096 0.215 0.107 0.104 0 0.089 0.2

No 0.965 0.961 1 0.974 0.885 0.965 0.914 0.968 0.79 0.956 0.806 0.926 0.967 0.904 0.785 0.893 0.896 1 0.911 0.8

N 141 488 287 1056 87 260 853 920 124 273 439 135 209 166 163 196 77 79 79 130


218


HHact 6 6 6 6 6 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

Day 6 6 6 6 6 0,4 0,4 0,4 0,4 0,4 0,4 0,4 0,4 0,4 0,4 0,4 0,4 0,4 0,4 1,2,3 1,2,3 1,2,3 1,2,3

nbr - - - - - - - - - - - - - - - - - - - 0 0 1-5 -

DurHH 0 0 0 0 1-4 - - - - - - - - - - - - - - - - - -

ModeHH - - - - - 0 0 0 0 0 1,3 1,3 1,3 1,3 1,3 2,4 2,4 2,4 2,4 0 0 0 0

Dist1 0 1-5 1-5 1-5 - - - - - - - - - - - - - - - 0-3 4-5 - -

ntou - - - - - 0 0 0 0 1-2 - - - - - - - - - 0 0 0 1-2

yWorkM - - - - - - - - - - - - - 0 1 - - - - - - - -

Ncar - - - - - - - - - - - - - - - 0-1 0-1 0-1 2 - - - -

Dist2 - - - - - - 0-1 2-5 - - 0-3 0-3 4 5 5 - - - - - - - -

Urban - 0-1 2-4 - - 0 1-2 1-2 3-4 - - - - - - - - - - - - - -

nEmp2 - - - - - - - - - - 0 1-5 - - - - - - - - - - -

SizePop - - - - - - - - - - - - - - - 0 1 2-5 - - - - -

Comp - 2,3 2,3 4 - - - - - - - - - - - - - - - - - - -

nEmp1 4-5 - - - - - - - - - - - - - - - - - - - - - -

Yes 0.074 0.24 0.143 0.252 0.028 0.031 0.039 0.127 0.047 0.164 0.012 0 0.027 0.03 0.005 0.005 0.005 0.003 0.051 0.037 0.068 0 0.162

No 0.926 0.76 0.857 0.748 0.972 0.969 0.961 0.873 0.953 0.836 0.988 1 0.973 0.97 0.995 0.995 0.954 0.997 0.949 0.963 0.932 1 0.838

N 231 204 722 445 106 223 129 221 1169 128 83 603 185 100 603 207 108 304 118 1013 1259 230 142

R341 R342 R343 R344 R345 R346 R347 R348 R349 R350 R351 R352 R353 R354 R355 R356 R357 R358 R359 R360 R361 R362 R363

219


HHact 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

Day 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6

Child - - - - - - - - - - - - - - 0,1 0,1 0,1 0,1 0,1 0,1 2,3 2,3 2,3

nsh1 0-1 2-5

- - - - 0 0 1-5 1-5 1-5

- - - - - - - - - - - -

ModeHH 1,3,2 1,3,2 1,3,2 1,3,2 1,3,2 4

- - - - - - - - - - - - - - - - -

Dist1 - - - - - - - - - - - - - - - - - - - - 0-1 2-4 5

SEC - - - 0-1 2-3

- - - - - - - - - 0,1 0,1 2-3 2-3 2-3 2-3

- - -

ntou - - - - - - 0 0 0 0 0 0 0 1-2

- - - - - - - - -

DrivF - - - - - - 0 1

- - - - - - - - - - - - - - -

Ncar - - - - - - - - 0-1 2

- - - - - - - - - - - - -

WstatM 0,2 0,2 0,2 1 1

- - - - - - - - - - - 0 0 1,2 1,2

- - -

nsoc - - - - - - 0 0 0 0 0 1-2

- - - - - - - - - - -

nEmp3 - - - - - - - - - - - - - 0-2 3-5

- - - - - - -

WstatF - - - - - - - - 0,1 0,1 2

- - - - - - - 0,1 2

- - -

yWorkF - - - - - - 0 0 0 0 0 0 1

- - - - - - - - - -

yOthF 0 0 1

- - - - - - - - - - - - - - - - - - - -

Yes 0 0.008 0.011 0.037 0 0.008 0.037 0.101 0.048 0.006 0.083 0.02 0 0.182 0.145 0.053 0.132 0.263 0.174 0.099 0.03 0.148 0.064

No 1 0.992 0.989 0.963 1 0.992 0.963 0.899 0.952 0.994 0.917 0.98 1 0.818 0.855 0.947 0.868 0.737 0.826 0.901 0.97 0.852 0.936

N 2105 258 94 82 209 898 107 337 374 175 180 250 119 88 380 187 76 259 391 222 99 115 140

R364 R365 R366 R367 R368 R369 R370 R371 R372 R373 R374 R375 R376 R377 R378 R379 R380 R381 R382 R383 R384 R385 R386

220

TABLE AII-2.1 Household Activity Allocation

durM 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

nsh1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

nbr 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

durF 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

yBusiM 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

nshn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1-3

wstatF 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 2 2 2 2 2 -

Acty 1 1 1 2,4 2,4 2,4 2,4 2,4 2,4 2,4 2,4 2,4 3 3 - - - - - -

Child 0,3 1,2 1,2 - - - - - - - - - - - - - 0 - 1,3,2 -

Comp - 2 3,4 - - - - - - - - - - - - - - - - -

nser - - - 0 0 0 0 0 0 0 0 1-3 - - - - - - - -

SEC - - - 0 0 1 1 1 1 2-3 2-3 - - - - - - - - -

SizePop - - - 0-3 4-5 - - - - - - - - - - - - - - -

AgeM - - - - - 0-2 0-2 3-4 3-4 - - - - - - - - - - -

drivF - - - - - 0 1 - - - - - - - - - - - - -

ncar - - - - - - - 0 1 - - - - - - - - -

Urb - - - - - - - - - 0-1 2-4 - - - - - - - - -

nEmp2 - - - - - - - - - - 0-1 2-5 - - 0-1 2-5 - -

wstatM - - - - - - - - - - - - - - 0-1 0-1 2 2 2 -

Day - - - - - - - - - - - - - - 0,1,2,3,6 4,5 - - - -

Male 0.514 0.304 0.512 0.487 0.333 0.398 0.336 0.388 0.39 0.404 0.468 0.387 0.229 0.38 0.311 0.116 0.247 0.341 0.4 0.326

Female 0.204 0.565 0.317 0.258 0.179 0.216 0.391 0.359 0.226 0.25 0.287 0.493 0.307 0.231 0.485 0.524 0.376 0.179 0.425 0.568

Both 0.282 0.13 0.171 0.255 0.487 0.386 0.273 0.252 0.384 0.346 0.244 0.12 0.464 0.389 0.205 0.361 0.376 0.48 0.175 0.105

N 245 92 82 353 78 88 271 103 461 280 1023 75 179 208 132 147 93 123 80 190


221


durM 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

nsh1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1

nbr 0 0 0 0 0 0 0 1 1 1 1 2 2 2 3-5 3-5 - 0 0 0

durF 0 0 1 2 3 3 4 - - - - 0-1 0-1 2-4 - - - - - -

yBusiM 1 1 - - - - - - - - - - - - - - - 0 0 0

wstatF - - - - - - - - - - - - - - - - - 0,1 2 2

Acty - - - - - - - - - - - - - - - - - 1,2 1,2 1,2

Child - - - - - - - 0 1,2,3 1,2,3 - 0-3 1-2 - - - - - - -

SizePop - - - - - - - - - - - - - - - - - - 0-1 2-5

ncar 0-1 2 - - - - - - - - - - - - - - - - - -

nEmp2 - - - - - - - - 0-2 3-5 - - - - - - - - - -

time1C - - - - 0-2 3-4 - - - - - - - - - - - - - -

yWorkF - - - - - - - - - - - - - - 0 1 - - - -

time1F - - - - - - - - - - - - - - - - 0-3 4 4 4

durHH - - - - - - - 0-1 0-1 0-1 2-4 - - - - - - - - -

Male 0.407 0.196 0.444 0.51 0.702 0.524 0.788 0.511 0.205 0.39 0.764 0.415 0.222 0.696 0.156 0.633 0.661 0.465 0.462 0.229

Female 0.458 0.768 0.477 0.351 0.228 0.435 0.185 0.38 0.712 0.568 0.236 0.5 0.763 0.293 0.834 0.367 0.333 0.412 0.385 0.542

Both 0.136 0.036 0.079 0.139 0.07 0.035 0.026 0.109 0.082 0.042 0 0.085 0.015 0.011 0.01 0 0.006 0.123 0.154 0.229

N 118 138 151 151 114 85 189 229 219 118 110 106 194 92 199 90 177 876 78 83


222


durM 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1

nsh1 1 1 1 1 2 2 3-5 3-5 3-5 0 0 1-5 - - - - - - -

nbr 0 0 1 2-5 - - - - - 0 0 0 1 2-5 - - - 0 1-5

durF - - - - - - - - - 0-1 0-1 0-1 0-1 0-1 2 3-4 - - -

yBusiM 0 1 - - - - - - - 0 0 0 0 0 0 0 1 1 1

wstatF - - - - 0,1 2 0,1 0,2 2 0,1 2 - - - - - - 0,1 0,1

Acty 3,4 - - - - - - - - - - - - - - - - - -

Dist1 - - - - - - 0-3 4-5 - - - - - - - - - - -

time1F 4 - - - - - - - - - - - - - - - 0-1 2-4 2-4

Male 0.394 0.268 0.405 0.176 0.311 0.416 0.284 0.19 0.522 0.445 0.198 0.249 0.271 0.147 0.49 0.691 0.314 0.115 0.037

Female 0.525 0.695 0.583 0.775 0.627 0.468 0.679 0.81 0.435 0.411 0.581 0.72 0.712 0.853 0.402 0.272 0.676 0.858 0.963

Both 0.081 0.037 0.012 0.049 0.062 0.117 0.037 0 0.043 0.144 0.221 0.032 0.017 0 0.108 0.037 0.01 0.027 0

N 617 82 84 102 598 154 190 211 92 299 86 189 118 136 102 136 105 261 161

R41 R42 R43 R44 R45 R46 R47 R48 R49 R50 R51 R52 R53 R54 R55 R56 R57 R58 R59

223


durM 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3

nbr - 0 0 0 0 1-5 1-5 1-5 1-5 0 0 0

durF - 0 0 0 0 0 0 0 0 1-2 1-2 1-2 3 3 4 0 0 0

yBusiM 1 - - - - - - - - - - - - - - - - -

wstatF 2 - - - - 0,1 0,1 0,1 2 - - - - - - - - -

Urb - - - - - - - - - - - - - - - 0-3 4 -

time1F 2-4 - - - - - - - - - - - - - - - - -

time4M - 0 0 1-3 4 0-1 0-1 2-4 - - - - - - - 0-1 0-1 2-4

time5M - 0-2 3-4 - - - - - - - - - - - - - - -

modeF - - - - - - - - - 0,1,4 2,3 2,3 - - - - - -

time4F - - - - - - - - - - - - 0-1 2-4 - - - -

Dist3 - - - - - 0-2 3-5 - - - 0-4 5 - - - - - -

Male 0.199 0.123 0.226 0.286 0.178 0.052 0.005 0.099 0.212 0.333 0.095 0.25 0.433 0.227 0.594 0.106 0.154 0.301

Female 0.789 0.845 0.77 0.639 0.779 0.948 0.995 0.901 0.788 0.667 0.853 0.74 0.433 0.667 0.338 0.839 0.831 0.667

Both 0.012 0.032 0.004 0.075 0.043 0 0 0 0 0 0.053 0.01 0.134 0.107 0.068 0.055 0.014 0.032

N 166 252 226 147 208 116 185 121 132 132 95 96 97 75 133 199 356 93

R60 R61 R62 R63 R64 R65 R66 R67 R68 R69 R70 R71 R72 R73 R74 R75 R76 R77

224


durM 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4

nsh1 - - - - - - - - - - - - - - - - -

nbr 1-5 1-5 1-5 - - - - 0 0 0 1-5 1-5 1-5 - - - -

durF 0 0 0 1-2 1-2 3 4 0 0 0 0 0 0 1-3 1-3 1-3 4

wstatF 0 1,2 1,2 - - - - - - - - - - - - - -

SizePop - - - - - - - - - - 0 1-5 1-5 - - - -

ncar - 0-1 2 - - - - - - - - - - 0-1 2 -

Urb - - - - - - - - - - - - - 0-1 2-4 2-4 -

time4M - - - - - - - 0 0 1-4 - - - - - - -

time5M - - - - - - - 0-2 3-4 - - - - - - - -

Dist3 - - - - - - - - - - - 0-3 4-5 - - - -

AgeF - - - 0 1-4 - - - - - - - - - - - -

Male 0.023 0.057 0.177 0.307 0.139 0.293 0.452 0.083 0.12 0.206 0 0.023 0.098 0.0284 0.175 0.062 0.327

Female 0.977 0.943 0.81 0.667 0.855 0.672 0.423 0.912 0.831 0.724 1 0.977 0.092 0.695 0.817 0.932 0.591

Both 0 0 0.013 0.027 0.006 0.034 0.125 0.004 0.049 0.065 0 0 0 0.021 0.008 0.006 0.082

N 171 106 79 75 173 116 104 480 142 107 198 171 92 95 126 176 110

R78 R79 R80 R81 R82 R83 R84 R85 R86 R87 R88 R89 R90 R91 R92 R93 R94

225

TABLE AIII-1 Duration

acty 1 2,4,7 2,4,7 2,4,7 2,4,7 2,4,7 2,4,7 3 5 5 5 5 6 6 6 6

durtot - 0 0 0 0 0 1-4 - 0-1 0-1 2-4 2-4 0-1 0-1 0-1 2-4

availT3 - 0-1 2-4 2-4 2-4 2-4 - - 0-3 4 - - - - - -

AgeM - - 0-1 2 2 3-4 - - - - - - - - - -

AgeF - - - 0-1 2-4 - - - - - - - - - -

ncar - - - - - - - - - - 0-1 2 0-1 0-1 2 -

Urban - - - - - - - - - - 0-2 3-4 - -

m 35.25 47.264 82.948 144.45 85.118 74.524 88.126 56.746 105.23 157.083 144.151 174.736 171.6 134 179.5 120.4

S 50.632 43.87 89.336 216.77 101.59 92.196 107.09 81.441 75.278 80.994 109.767 114.721 156 113 158.9 86.86

N 108 144 495 75 245 584 95 497 121 952 205 121 168 272 197 337

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16

226

TABLE AIII-2.1 Start-Time

availT2 0 0 0 0 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4

acty 1,4,2,3 1,4,2,3 5,6,7 5,6,7 1,6 1,6 1,6 1,6 1,6 1,6 2,3,4 2,3,4 2,3,4 2,3,4 2,3,4 2,3,4 2,3,4

child 0 1,2,3 - - - - - - - - - - - - - - -

availT4 - - 0-2 3-4 - - - - - - - - - - - - -

nlei - - - - 0 0 0 0 0 1-2 - - - - - - -

Day - - - - 0,2,6 0,2,6 0,2,6 0,2,6 1,3,4,5 - - - - - - - -

durasi - - - - 0-2 0-2 3 4 - - 0 0 1-4 1-4 1-4 1-4

Comp - - - - 2-3 4 - - - - - - - - - - 2,4

durtot - - - - - - - - - - 0 0 0 0 0 0 1-4

ageM - - - - - - - - - - 0-2 3-4 - - - - -

SEC - - - - - - - - - - - - 0-1 2-3 2-3 2-3 -

Dist3 - - - - - - - - - - - - - 0-2 3-5 3-5 -

ncar - - - - - - - - - - - - - - 0-1 2 -

m 1005.3 890.9 1172.3 1066.82 824.88 737.25 708.9 782.7 854.06 989.69 797.87 733.39 713.71 757.13 710.38 746.14 783.54

S 164.27 247.09 96.88 152.09 229.28 204.25 190.75 223.36 219.77 176.59 169.68 161.7 142.21 142.76 133.48 135.52 137.51

N 130 80 216 96 112 94 115 87 399 140 113 84 529 209 206 131 237


227

TABLE AIII-2.2 Start-Time

availT2 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4 1-4

acty 2,3,4 5 5 5 5 5 5 5 5 7 7 7 7 7

Day - 0-6 0-6 0-6 1,3,2 4,5 4,5 - - - - - - -

durasi - - - - - 0-3 4 - - 0-1 2-4 0-1 0-1 2-4

Comp 3 - - - - - - - - - - - - -

durtot 1-4 - - - - - - - - 0 0 1-4 1-4 1-4

ageM - - - - - - 0-1 2-4 - - - - - -

SEC - 0-1 2-3 2-3 - - - - - - - - - -

dursoc - 0-1 0-1 0-1 0-1 0-1 0-1 0-1 2-4 - - - - -

nEmp1 - - 0-1 2-5 - - - - - - - - - -

m 841.17 813.79 884.32 827.05 875.26 892.54 1035.2 949.53 1016.8 856.75 750.47 900.86 964.04 840.67

S 126.03 164.09 157.52 171.48 211.28 195.8 186.1 227.6 155.42 151.69 125.54 176.61 166.45 131.46

N 101 159 163 144 247 188 85 78 122 183 116 131 75 78

R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31

228

TABLE AIV-1.1 Location Choice Model – Independent Activity – 3 alternatives

LvoisLna 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1

Vohtt 0 0 1 1 2-3 2-3 2-3 2-3 2-3 4 4 4 4 4 - - - - - -

Nahtt 0-2 3-4 - - - - - - - - - - - - - - - - - -

Nagord - - 0 1-4 - - - - - - - - - - - - - - - -

Vosize - - - - 0-2 0-2 0-2 3-4 3-4 - - - - - - 0-2 3-4 - - -

NaH - - - - 0 1 1 - - - - - - - - - - - - 0

NaSize - - - - - 0-1 2-4 - - - - - - - - - - - - -

VoisA - - - - - - - 0 1 - - 0 1 - - - - - - -

Voty - - - - - - - - - 0,1 2 3 3 3 - - - - - -

MxSizeD4 - - - - - - - - - - - 0-2 0-2 3-5 - - - - - -

Adur - - - - - - - - - - - - - - 0 0 0 0 0 0

Aty - - - - - - - - - - - - - - 0,1,10,3 0,1,10,3 0,1,10,3 2,5 2,5 4,6,7,9

Vogord - - - - - - - - - - - - - - 0 1-4 1-4 - - 0

Comp - - - - - - - - - - - - - - - - - 0,4,1 2,3 -

same as

previous 0.237 0.179 0.421 0.4 0.275 0.257 0.158 0.346 0.41 0.099 0.092 0.219 0.328 0.453 0.792 0.578 0.642 0.497 0.678 0.462

same as next 0.289 0.128 0.368 0.183 0.087 0.151 0.27 0.137 0.049 0.113 0.021 0.08 0.017 0.047 0.002 0.6 0 0 0 0

other 0.474 0.694 0.24 0.417 0.638 0.592 0.566 0.517 0.542 0.789 0.887 0.701 0.655 0.5 0.205 0.427 0.308 0.503 0.322 0.538

N 97 196 95 240 207 331 152 315 144 142 142 187 116 128 307 216 169 169 326 169


229


LvoisLna 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Vosize - - - - 0-3 4 - - - - - - - - - - 0-1

NaH 1 1 1 1 - - - - - - - - - - - - -

Adur 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1

Aty 4,6,7,9 4,6,7,9 4,6,7,9 4,6,7,9 4,6,7,9 4,6,7,9 4,6,7,9 4,6,7,9 4,6,7,9 4,6,7,9 8 8 8 0,1,10,3,5,8 0,1,10,3,5,8 0,1,10,3,5,8 2,4,6,7,9

Vogord 0 0 0 0 1 1 2-4 2-4 2-4 2-4 - - - 0 0 0 0

Comp - - - - - - - 0,1 2,4,3 - - - - - - - -

pAge 0,1 0,1 2,4,3 2,4,3 - - - - - - - - - - 0,3,4 1,2 -

Tavail 0,1 2-4 0-2 3-4 - - - - - - - - - - - - -

Naty - - - - - - 0 0 0 1,2,3 - - - - - - -

Driver - - - - - - 0 1 1 - - - - - - - -

SEC - - - - - - - - - - 0 1,2,3 1,2,3 - - - -

Wstat - - - - - - - - - - - 0 1,2 - - - -

Gend - - - - - - - - - - - - - 0 1 1 0

same as

previous 0.675 0.527 0.789 0.654 0.369 0.598 0.49 0.241 0.413 0.14 0.657 0.852 0.759 0.723 0.722 0.529 0.454

same as next 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

other 0.325 0.473 0.211 0.346 0.631 0.402 0.51 0.759 0.587 0.86 0.343 0.148 0.241 0.277 0.278 0.471 0.546

N 154 226 180 130 401 107 149 145 286 164 181 310 224 329 90 155 205


230


LvoisLna 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Vosize 2-4 - - 0-1 2-4 - - 0-1 2-4 - - - - - - - -

Adur 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2

Aty 2,4,6,7,9 2,4,6,7,9 0,1,10,2 3,4,8 3,4,8 5,7,9,6 5,7,9,6 5,7,9,6 5,7,9,6 0,1,10,8,2,3 4,6,9 5,7 - - - - -

Vogord 0 0 1-2 1-2 1-2 1-2 1-2 - - 3-4 3-4 3-4 0 0 0 0 0

Comp - - - - - - - - - - - - - - - 0,1,3 2,4

pAge - - - - - 0,3,2 1,4 - - - - - - - - - -

Tavail - - - - - - - - - - - - 0-2 3-4 3-4

Wstat - - - - - - - - - - - - 0 1,2 - - -

Gend 0 1 - - - - - - - - - - - - - - -

Urb - - - - - 0,1,2 0,1,2 3,4 3,4 - - - - - - - -

Child - - - - - - - - - - - - 0 0 0 0 0

Day - - - - - - - - - - - - 0,4 0,4 1,3,2,5,6 1,3,2,5,6 1,3,2,5,6

same as

previous 0.599 0.362 0.642 0.353 0.57 0.152 0.308 0.25 0.447 0.405 0.118 0.216 0.597 0.427 0.417 0.224 0.378

same as next 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

other 0.401 0.638 0.567 0.647 0.43 0.848 0.692 0.75 0.553 0.595 0.882 0.784 0.403 0.573 0.583 0.776 0.622

N 202 199 485 306 179 132 120 128 132 252 119 125 139 103 312 152 156


231


LvoisLna 1 1 1 1 1 1 1 1 1 1 1 1 1

Nagord - - - - 0 1-4 - - - - - - -

Vosize - - - - - - 0 1-4 - - - - -

Adur 2 2 2 2 2 2 2 2 2 2 2 2 2

Aty - - 0,1,10,8,2,5 0,1,10,8,2,5 0,1,10,8,2,5 3,6,9,7,4 - - - - - - -

Vogord 0 0 1 1 1 1 1 1 1 2-4 2-4 2-4 2-4

Comp - - 0,2,3 0,2,3 0,2,3 0,2,3 1,4 1,4 - - - - -

Tavail - - 0-3 0-3 0-3 0-3 0-3 0-3 4 - - 0 1-4

SEC - - - 0,1,2 3 - - - - - - - -

Wstat - - - - - - - - - 0 1,2 - -

Urb - - 0,2,3 1,4 1,4 - - - - - - - -

Child 1,2 3 - - - - - - - 0 0 1,2,3 1,2,3

same as

previous 0.517 0.314 0.347 0.623 0.394 0.24 0.153 0.32 0.213 0.226 0.14 0.44 0.289

same as next 0 0 0 0 0 0 0 0 0 0 0 0 0

other 0.483 0.686 0.653 0.377 0.606 0.76 0.847 0.68 0.787 0.774 0.86 0.56 0.711

N 575 159 170 154 109 167 111 122 197 561 350 159 402

R55 R56 R57 R58 R59 R60 R61 R62 R63 R64 R65 R66 R67

232


LvoisLna 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1

Nagord 0 0 1-3 1-3 1-3 1-3 1-3 1-3 4 - - - - - - - - - - -

Naty 0,2 1,3 - - - - - - - - - - - - - - - - - -

NaSize - - 0-1 0-1 0-1 2 3 4 - - - - - - - - - - - -

pAge - - 0,4,1 0,4,1 2,3 - - - - - - - - - - - - - - -

VoSize - - 0-1 2-4 - - - - - - - 0 1-4 1-4 1-4 1-4 0-1 0-1 0-1 0-1

Vogord - - - - - - - - - 0 0 0 0 0 0 0 0 0 0 0

Urb - - - - - - - - - 0,1,2 3 4 4 4 4 4 4 4 4 4

Adur - - - - - - - - - - - 0-1 0-1 0-1 0-1 0-1 2 2 2 2

Comp - - - - - - - - - - - - 0,1,3 0,1,3 2 4 - - - -

Aty - - - - - - - - - - - - - - - - 0,1,10,4,5,9 2,6,8,3,7 2,6,8,3,7 2,6,8,3,7

Wstat - - - - - - - - - - - - - - - - - 0,2 0,2 1

SEC - - - - - - - - - - - - - - - - - 0,1 2,3

Day - - - - - - - - - - - - - - - - - - - -

Ncar - - - - - - - - - - - - - - - - - - - -

Pwstat - - - - - - - - - - - - 0,1 2 - - - - - -

Z1 0.069 0.078 0.093 0.093 0.113 0.074 0.072 0.03 0.016 0.027 0.006 0.025 0 0.014 0 0.011 0.052 0 0 0

Z2 0.056 0.055 0.056 0.023 0.163 0.107 0.048 0.107 0.033 0.047 0.038 0.025 0.034 0.027 0.026 0 0.006 0.008 0 0

Z3 0.089 0.133 0.131 0.085 0.087 0.149 0.048 0.148 0.085 0.081 0.026 0.014 0 0.014 0.043 0.011 0.039 0 0.014 0.01

Z4 0.046 0.055 0.098 0.093 0.05 0.041 0.084 0.059 0.077 0.081 0.045 0.025 0.017 0.027 0.026 0.078 0.032 0.008 0.007 0.01

Z5 0.056 0.039 0.145 0.116 0.062 0.05 0.133 0.118 0.203 0.027 0.032 0.022 0 0.014 0 0.022 0 0 0 0.019

Z6 0.056 0.039 0.061 0.023 0.013 0.041 0.012 0.036 0.02 0.014 0.064 0.084 0.026 0.007 0.052 0.056 0.039 0 0.014 0.01

Z7 0.043 0.023 0.019 0 0.075 0.041 0.012 0.018 0.049 0.034 0.032 0.025 0.026 0.02 0.069 0.033 0.045 0.016 0 0.058

Z8 0.056 0.016 0.042 0.039 0.025 0.058 0.048 0.03 0.053 0.135 0 0.025 0.017 0.007 0.017 0.033 0.052 0.008 0.007 0.029

Z9 0.026 0.086 0.023 0.054 0 0.041 0.048 0.012 0.045 0.027 0.058 0.039 0.051 0.027 0 0.089 0.039 0.025 0.022 0.049

Z10 0.026 0.008 0.042 0.116 0.087 0.041 0.096 0.077 0.085 0.014 0.038 0.017 0 0 0.026 0.011 0 0.016 0.022 0.019

Z11 0.053 0.023 0.033 0.023 0.013 0.025 0.036 0.018 0.02 0.047 0.038 0.081 0.026 0.02 0.034 0.033 0.052 0.008 0.022 0.058

Z12 0.049 0.016 0.033 0 0.062 0.025 0.048 0.03 0.012 0.027 0.032 0.034 0.043 0.054 0.017 0.022 0.052 0.098 0.043 0.029

Z13 0.039 0.008 0.005 0.039 0.013 0.05 0 0.018 0.045 0.054 0.051 0.031 0.085 0.047 0.078 0.122 0.026 0.025 0.007 0

Z14 0.03 0.031 0.019 0.008 0.025 0.008 0 0.018 0.016 0.027 0.026 0.031 0.068 0.047 0.112 0.033 0.019 0.016 0.007 0.039

Z15 0.02 0.078 0.033 0.047 0.037 0.017 0.036 0.041 0.053 0.047 0.038 0.014 0.017 0.041 0.017 0.011 0.045 0.049 0.029 0.039


233

TABLE AIV-2.1 (cont.)

LvoisLna 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1

Nagord 0 0 1-3 1-3 1-3 1-3 1-3 1-3 4 - - - - - - - - - - -

Naty 0,2 1,3 - - - - - - - - - - - - - - - - - -

NaSize - - 0-1 0-1 0-1 2 3 4 - - - - - - - - - - - -

pAge - - 0,4,1 0,4,1 2,3 - - - - - - - - - - - - - - -

VoSize - - 0-1 2-4 - - - - - - - 0 1-4 1-4 1-4 1-4 0-1 0-1 0-1 0-1

Vogord - - - - - - - - - 0 0 0 0 0 0 0 0 0 0 0

Urb - - - - - - - - - 0,1,2 3 4 4 4 4 4 4 4 4 4

Adur - - - - - - - - - - - 0-1 0-1 0-1 0-1 0-1 2 2 2 2

Comp - - - - - - - - - - - - 0,1,3 0,1,3 2 4 - - - -

Aty - - - - - - - - - - - - - - - - 0,1,10,4,5,9 2,6,8,3,7 2,6,8,3,7 2,6,8,3,7

Wstat - - - - - - - - - - - - - - - - - 0,2 0,2 1

SEC - - - - - - - - - - - - - - - - - 0,1 2,3 -

Pwstat - - - - - - - - - - - - 0,1 2 - - - - - -

Z16 0.056 0.008 0.009 0.008 0.05 0 0.024 0.006 0.004 0.014 0.032 0.073 0.085 0.027 0.069 0.011 0.09 0 0.145 0.039

Z17 0.039 0.031 0.009 0.008 0.025 0.05 0.012 0.012 0.028 0.007 0.006 0.073 0.103 0.027 0.078 0.056 0.071 0.098 0.058 0.039

Z18 0.043 0.016 0.019 0.016 0.013 0.041 0 0.036 0.024 0.014 0.045 0.053 0.077 0.054 0.034 0.033 0.052 0.057 0.036 0.01

Z19 0.02 0.062 0.014 0.008 0 0.017 0.048 0.024 0.012 0.02 0.038 0.053 0.06 0.095 0.06 0.067 0.039 0.107 0.036 0.049

Z20 0.016 0.008 0.014 0.008 0.025 0 0.06 0.006 0.037 0.068 0.058 0.053 0.034 0.047 0.026 0.022 0.032 0.025 0.058 0.117

Z21 0.023 0.047 0.037 0.008 0 0.025 0.012 0.041 0.012 0.02 0.051 0.079 0.103 0.054 0.06 0.089 0.019 0.115 0.065 0.087

Z22 0.03 0.031 0.005 0.008 0 0.025 0.012 0.047 0.012 0.02 0.064 0.031 0.026 0.074 0.095 0.022 0.045 0.107 0.094 0.078

Z23 0.023 0.055 0.019 0.07 0.025 0.033 0.048 0.03 0.024 0.047 0.064 0.045 0.009 0.108 0.017 0.044 0.032 0.033 0.138 0.107

Z24 0.013 0.039 0.014 0.031 0.025 0.008 0.024 0.006 0.016 0.034 0.071 0.022 0.043 0.108 0.026 0.033 0.065 0.033 0.101 0.058

Z25 0.023 0.016 0.028 0.078 0.013 0.033 0.0336 0.036 0.016 0.068 0.045 0.022 0.051 0.041 0.017 0.056 0.058 0.131 0.072 0.049

N 304 128 214 129 80 121 83 169 246 148 156 356 117 148 116 90 155 122 138 103


234


LvoisLna 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

pAge - - - - - - - - - - - 0,1 2,4,3 - - - - - - -

VoSize 2 3-4 - - - - - - - - - - - - - - 0 1-4 - -

Vogord 0 0 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2

Urb 4 4 - - - 0,3,4 1,2 0,2,1 3,4 3,4 - - - - - - - - - -

Adur 2 2 0 1 2 4 - - - - - - - - - - - - 0 1

Comp - - - - - - - - - - - - - - 0,4 0,4 0,4 0,4 1,3 1,3

Aty - - 0,1,10,6 0,1,10,6 0,1,10,6 2 2 3 3 3 4,5 7 7 8,9 - - - - - -

SEC - - - - - - - - - - - - - - 0 1,2,3 - - - -

Day - - - - - - - - 0,2,5 1,6,3,4 - - - - 0,4,5,6 0,4,5,6 1,3,2 1,3,2 - -

Z1 0 0.006 0.03 0 0.021 0.006 0.036 0.024 0.022 0.065 0.056 0.055 0.043 0.044 0.024 0.033 0.038 0.007 0.015 0.019

Z2 0 0.006 0.023 0.087 0.026 0 0 0.087 0.011 0.013 0.014 0.039 0.085 0.082 0.012 0.057 0.025 0.035 0.015 0.028

Z3 0 0.006 0.075 0.013 0.041 0.006 0.048 0.016 0.011 0 0.021 0.07 0.017 0.088 0.012 0 0.038 0.035 0 0.028

Z4 0.022 0.025 0.068 0.013 0.015 0.035 0.048 0.071 0 0.091 0.049 0.109 0.077 0.088 0.048 0.033 0.013 0.014 0.007 0.019

Z5 0 0 0.008 0.013 0.031 0.006 0 0.031 0.055 0.026 0.007 0.039 0.06 0.038 0.06 0.016 0.038 0.035 0.052 0.009

Z6 0.011 0 0.023 0.1 0.052 0.029 0.012 0.031 0.077 0 0.028 0.062 0.009 0.033 0 0.033 0.038 0 0.037 0.028

Z7 0.045 0 0.113 0.1 0.046 0.012 0.012 0.134 0 0.026 0.035 0.016 0.103 0.06 0.024 0.09 0.038 0.056 0.067 0.046

Z8 0.067 0 0.105 0.087 0.057 0.035 0.12 0.031 0 0.078 0.028 0.047 0.017 0.121 0.083 0.041 0.127 0.084 0.007 0.056

Z9 0.011 0.051 0.015 0.013 0.036 0.012 0.12 0.055 0.022 0.013 0.104 0.008 0.068 0.049 0 0.057 0.013 0.014 0.075 0.019

Z10 0 0 0 0.05 0.046 0.029 0.096 0.055 0.066 0.091 0.062 0.039 0.068 0.06 0.202 0.025 0.013 0.049 0.075 0.056

Z11 0.034 0.013 0.053 0.025 0.021 0.069 0.024 0.031 0.055 0.078 0.097 0.008 0.026 0.044 0.012 0.049 0.063 0.049 0.082 0.046

Z12 0.011 0.013 0.068 0.05 0.046 0.064 0.024 0.102 0.088 0.039 0.014 0.016 0.009 0.022 0.012 0.016 0.025 0.063 0.03 0.056

Z13 0.101 0.032 0.09 0.037 0.031 0.104 0.072 0.016 0.033 0.091 0.056 0.039 0.009 0.016 0.06 0.025 0.063 0.056 0.037 0.046

Z14 0.056 0.045 0.023 0.025 0.041 0.023 0.024 0.024 0.022 0.052 0.042 0.039 0.017 0.022 0.024 0.057 0.025 0.098 0.015 0.056

Z15 0.022 0.019 0.045 0.037 0.036 0.052 0.06 0.031 0.121 0.013 0.062 0.023 0.043 0.016 0.012 0.016 0.063 0.182 0.127 0.037


235


LvoisLna 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

pAge - - - - - - - - - - - 0,1 2,4,3 - - - - - - -

VoSize 2 3-4 - - - - - - - - - - - - - - 0 1-4 - -

Vogord 0 0 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2

Urb 4 4 0,3,4 1,2 0,2,1 3,4 3,4 - - - - - - - - - -

Adur 2 2 0 1 2 4 - - - - - - - - - - - - 0 1

Comp - - - - - - - - - - - - - - 0,4 0,4 0,4 0,4 1,3 1,3

Aty - - 0,1,10,6 0,1,10,6 0,1,10,6 2 2 3 3 3 4,5 7 7 8,9 - - - - - -

SEC - - - - - - - - - - - - - - 0 1,2,3 - - - -

Day - - - - - - - - 0,2,5 1,6,3,4 - - - - 0,4,5,6 0,4,5,6 1,3,2 1,3,2 - -

Z16 0.101 0.045 0.015 0.037 0.057 0.081 0.012 0.016 0.055 0.039 0.035 0.039 0.026 0.016 0.036 0.025 0.025 0.014 0 0.019

Z17 0.034 0.102 0 0.025 0.026 0.017 0.012 0.016 0.033 0.052 0.007 0.008 0.034 0.005 0.083 0.049 0.038 0 0.067 0.074

Z18 0.09 0.038 0.03 0.087 0.031 0.046 0.024 0.016 0.033 0.013 0.056 0.031 0.06 0.038 0.071 0.049 0.013 0.042 0.06 0.046

Z19 0.056 0.057 0.015 0.037 0.036 0.029 0.024 0.016 0 0.013 0.021 0.047 0.026 0.038 0.024 0.049 0.038 0.014 0.052 0.074

Z20 0.056 0.045 0.03 0 0.077 0.017 0.036 0.031 0.022 0.013 0.028 0.07 0.051 0.005 0.06 0.049 0.076 0.049 0 0.056

Z21 0.045 0.121 0.038 0.087 0.021 0.069 0.084 0.071 0.099 0.026 0.056 0.023 0.026 0.011 0.012 0.041 0.038 0.021 0.09 0.019

Z22 0.045 0.096 0.038 0.037 0.036 0.11 0 0.047 0.033 0.026 0.014 0.031 0.068 0 0.06 0.049 0.038 0.021 0.022 0.037

Z23 0.079 0.115 0.038 0.025 0.077 0.081 0.048 0 0.077 0.026 0.014 0.016 0.051 0.011 0.012 0.057 0 0.014 0.007 0.028

Z24 0.056 0.083 0.023 0.013 0.041 0.035 0.012 0.39 0.011 0.078 0.035 0.07 0.009 0.049 0.036 0.025 0.063 0 0.037 0.037

Z25 0.056 0.083 0.038 0 0.052 0.035 0.048 0.008 0.055 0.039 0.062 0.055 0 0.038 0.024 0.057 0.051 0.049 0.022 0.065

N 89 157 133 80 194 173 83 127 91 77 144 128 117 182 84 122 79 143 134 108


236


LvoisLna 1 1 1 1 1 1 1 1 1 1 1 1

VoSize - - - 0 1-4 - - 0 0 1-2 3-4 3-4

Vogord 2 2 2 3 3 3 3 4 4 4 4 4

Urb - 0,2 1,3,4 - - - - - - - - -

Comp 1,3 2 2 - - - - - - - - -

Wstat - - - 0,1 0,1 2 - - - - - -

SEC - - - 0,3,1 0,3,1 0,3,1 2 0,1 2,3 - - -

Ncar - - - - - - - - - - 0,2 1

Z1 0 0 0.042 0 0.009 0 0.013 0.011 0.025 0.02 0.014 0

Z2 0.005 0.022 0 0.026 0.022 0 0.007 0.011 0.017 0.045 0.007 0.024

Z3 0 0.022 0 0.009 0.013 0.028 0.007 0.032 0.059 0.03 0.021 0.006

Z4 0.016 0.055 0 0.026 0.058 0.014 0.034 0.147 0.017 0.025 0.048 0.018

Z5 0.016 0.022 0.014 0.017 0 0.007 0.007 0.063 0.034 0.02 0.048 0.024

Z6 0.01 0.022 0.021 0.026 0.004 0 0.02 0 0.008 0.03 0 0

Z7 0.021 0.022 0.099 0.078 0.04 0.028 0.027 0.032 0 0.03 0.007 0.012

Z8 0.026 0.11 0.014 0.052 0.018 0.014 0.04 0.021 0.068 0.045 0.007 0.024

Z9 0.031 0.033 0 0.034 0.04 0.014 0.034 0.011 0.034 0.061 0.028 0.029

Z10 0.031 0.055 0.042 0.026 0.085 0.07 0.114 0.105 0.042 0.035 0.117 0.071

Z11 0.057 0.11 0.063 0.052 0.036 0.021 0.034 0 0.034 0 0 0.041

Z12 0.031 0 0.035 0.069 0.054 0.042 0.04 0 0.025 0.035 0.034 0.029

Z13 0.047 0.011 0.063 0.026 0.071 0.049 0.02 0.042 0.102 0.066 0.014 0.018

Z14 0.031 0.099 0.113 0.034 0.058 0.028 0.067 0.042 0.051 0.04 0.021 0.029

Z15 0.078 0.022 0.106 0.095 0.134 0.162 0.067 0.137 0.051 0.086 0.124 0.059

R41 R42 R43 R44 R45 R46 R47 R48 R49 R50 R51 R52

237


LvoisLna 1 1 1 1 1 1 1 1 1 1 1 1

VoSize - - - 0 1-4 - - 0 0 1-2 3-4 3-4

Vogord 2 2 2 3 3 3 3 4 4 4 4 4

Urb - 0,2 1,3,4 - - - - - - - - -

Comp 1,3 2 2 - - - - - - - - -

Wstat - - - 0,1 0,1 2 - - - - - -

SEC - - - 0,3,1 0,3,1 0,3,1 2 0,1 2,3 - - -

Ncar - - - - - - - - - - 0,2 1

Z16 0.026 0.011 0.014 0.017 0.022 0.056 0.054 0.021 0.059 0.025 0 0.035

Z17 0.026 0.022 0 0 0.022 0.049 0.087 0.011 0.025 0.051 0.028 0.012

Z18 0.026 0.088 0.099 0.043 0.049 0.049 0.013 0.105 0.025 0.051 0.021 0.065

Z19 0.057 0.077 0.042 0.043 0.058 0.028 0.013 0.021 0.025 0.035 0.076 0.047

Z20 0.094 0.055 0.049 0.129 0.054 0.014 0.04 0.084 0.102 0.071 0.152 0.141

Z21 0.036 0.022 0 0.026 0.027 0.014 0.06 0.021 0.017 0.02 0.034 0.035

Z22 0.047 0 0.056 0.026 0 0.07 0.054 0.032 0.017 0.015 0.034 0.029

Z23 0.057 0.033 0.042 0.034 0.018 0.042 0.081 0.011 0.034 0.02 0.021 0.082

Z24 0.062 0.044 0.056 0.017 0.045 0.049 0.034 0 0.034 0.04 0.055 0.035

Z25 0.167 0.044 0.028 0.095 0.062 0.148 0.034 0.042 0.093 0.101 0.09 0.135

N 192 91 142 116 224 142 149 95 118 198 145 170

R41 R42 R43 R44 R45 R46 R47 R48 R49 R50 R51 R52

238

TABLE AIV-3.1 Location Choice Model – Joint Activity – 3 alternatives

Vohtt 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Dmax 0-1 0-1 0-1 0-1 2 2 2 2 3 3 3 3 3 3 3 3 3

NaH 0 1 1 1 0 1 1 1 - - - - - - - - -

Urb - 0,1,2,3 4 4 - - 0,2,1,3 4 - - - - - - - - -

VoSize - - 0-1 2-4 - 0 1-4 1-4 - - - - - - - - -

Aty - - - - - - - - 0,1,10,6,9 0,1,10,6,9 0,1,10,6,9 0,1,10,6,9 2,5,4,7,3 2,5,4,7,3 2,5,4,7,3 2,5,4,7,3 2,5,4,7,3

pAge - - - - - - - - 0,1,2 0,1,2 3,4 - - - - - -

Adur - - - - - - - - 0 1-2 1-2 - - - - - -

Wstat - - - - - - - - - 0 1,2 - - - - - -

NaSize - - - - - - - - - - - - 0-1 0-1 0-1 2-4 2-4

Vogord - - - - - - - - - - - - 0 1-4 1-4 - -

Driver - - - - - - - - - - - - - 0 1 - -

SEC - - - - - - - - - - - - - - - 0,1 2,3

same as previous 0.281 0.477 0.629 0.814 0.163 0.228 0.336 0.547 0.16 0.118 0.035 0.221 0.232 0.205 0.101 0.432 0.262

same as next 0 0 0 0 0.008 0 0 0 0 0 0 0 0 0 0 0 0

other 0.719 0.523 0.371 0.186 0.829 0.772 0.664 0.453 0.84 0.882 0.965 0.779 0.768 0.795 0.899 0.568 0.738

N 128 241 143 97 129 158 232 190 100 153 144 86 142 83 189 125 191


239


Vohtt 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Dmax 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4

NaH - 0 1 1 - - - - - - - - - - - -

VoSize - - - - - - - - 0 1-4 - - - - 0-2 3-4

Aty 8 0,1,10,6,4,5,7 0,1,10,6,4,5,7 0,1,10,6,4,5,7 2,3 8,9 8,9 8,9 0,1,2,10,3,4,8 0,1,2,10,3,4,8 5,7,6,9 - - - - -

pAge - - - - - - - - - - - - 0,2 0,2 1 1

Adur - 0 0 0 0 0 0 0 1 1 1 1 2 2 2 2

Vogord - - 0 1-4 - - 0 1-4 - - - - - - - -

Abt - - - - - 0,5,4 1,2,3 1,2,3 - - - - - - - -

NaisA - - - - - - - - 0 0 0 1 - - - -

Tavail - - - - - - - - - - - - 0-2 3-4 - -

same as previous 0.458 0.065 0.315 0.16 0.284 0.771 0.641 0.458 0.119 0.315 0.083 0 0.039 0.128 0.006 0.048

same as next 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

other 0.542 0.935 0.685 0.84 0.716 0.229 0.359 0.542 0.881 0.685 0.917 1 0.961 0.872 0.994 0.952

N 168 108 124 187 95 118 103 107 118 143 385 82 412 117 346 124

R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33

240


Vohtt 0 0 1-2 1-2 3 3 4 4

Dmax 4 4 - - - - - -

VoSize - - - - 0-1 3-4 - -

pAge 3,4 3,4 - - - - - -

Adur 2 2 - - - - - -

SEC 0,3,2 1 - - - - - -

Pwstat - - 0,1 2 - - - -

VoisA - - - - - - 0 1

same as previous 0.062 0.168 0.359 0.151 0.08 0.248 0.158 0.302

same as next 0 0 0.18 0.049 0.127 0.072 0.036 0.012

other 0.938 0.832 0.461 0.75 0.793 0.68 0.806 0.685

N 208 137 128 172 150 153 222 162

R34 R35 R36 R37 R38 R39 R40 R41

241


Dmax 0 1 1 2 2 2 2 3 3 3 3 3 3 3 4 4 4 4 4 4

NaH -

0 1 0 1 1 -

0 1 1 1 1 1 1 - - - - - -

Dmin - - -

0 0 0 1-4 -

0 0 0 0 1-4 1-4 - -

0 1-4 - -

VoH - - - - -

0 1 - - - - - - - - - - - - -

NaSize - - - - - - - -

0 0 1-4 1-4 - - - - - - - -

Wstat - - - - - - - -

0 1,2 - - - - - - - - - -

Urb - - - - - - - - - -

0 1,4,2,3 - - - - - - - -

Vogord - - - - - - - - - - - -

0 1-4 - - - - - -

Adur - - - - - - - - - - - - - -

0 0 0 0 1 1

LvoisLNa - - - - - - - - - - - - - -

0 1 1 1 0 1

NaH - - - - - - - - - - - - - - -

0 1 1 - -

Comp - - - - - - - - - - - - - - - - - - -

0,1,2,3

Z1 0.116 0.047 0.078 0.025 0.082 0.028 0 0.008 0.027 0.036 0.031 0.03 0.009 0 0.074 0 0.022 0 0.067 0.003

Z2 0.107 0.047 0.064 0.038 0.071 0.103 0.016 0.017 0.036 0.095 0 0.05 0 0.005 0.032 0 0.035 0.006 0.027 0.016

Z3 0.19 0.056 0.054 0.063 0.002 0.084 0.005 0.017 0.116 0.119 0.02 0.026 0 0.005 0.063 0 0.044 0.006 0.027 0.013

Z4 0.116 0.028 0.118 0.051 0.102 0.093 0 0.017 0.116 0.077 0.041 0.017 0 0 0.053 0.008 0.053 0.006 0.027 0.006

Z5 0.091 0.047 0.049 0.025 0.133 0.028 0.005 0.017 0.036 0.012 0.01 0.04 0.005 0.011 0.032 0 0.026 0.006 0.027 0.013

Z6 0.033 0.056 0.093 0.063 0.051 0.056 0.098 0.013 0.04 0 0.031 0.03 0.024 0.016 0.021 0.008 0.044 0.034 0.04 0.006

Z7 0.033 0.028 0.088 0.063 0.071 0.073 0.017 0.042 0.036 0.024 0.031 0.063 0.047 0.054 0.063 0.008 0.04 0.023 0.013 0.006

Z8 0.041 0.075 0.103 0.038 0.041 0.103 0.083 0.029 0.036 0.024 0.051 0.059 0.033 0.054 0.021 0.008 0.026 0.034 0 0.016

Z9 0.05 0.075 0.074 0.051 0.061 0.033 0.057 0.029 0.071 0.012 0.041 0.033 0.047 0.022 0.032 0.008 0.018 0.017 0.027 0.016

Z10 0.05 0.028 0.069 0.013 0.041 0.023 0.062 0.021 0.009 0.048 0.132 0.053 0.005 0.07 0.053 0.008 0.009 0.023 0.067 0.009

Z11 0 0.019 0.034 0.038 0.071 0.061 0.078 0.034 0.027 0.06 0.082 0.059 0.081 0.076 0.053 0.016 0.022 0.051 0.067 0.034

Z12 0.008 0.056 0.02 0.013 0.02 0.061 0.093 0.042 0.045 0.071 0.01 0.083 0.052 0.086 0.011 0.023 0.044 0.017 0.053 0.016

Z13 0 0.065 0.015 0.076 0.031 0.033 0.093 0.05 0.098 0.012 0.031 0.066 0.066 0.081 0.042 0.023 0.022 0.04 0.013 0.006

Z14 0.017 0.037 0.025 0.025 0.061 0.07 0.062 0.055 0.027 0.119 0.061 0.04 0.085 0.076 0.032 0.023 0.009 0.051 0.04 0.022

Z15 0.025 0.028 0.044 0.063 0.031 0.042 0.041 0.029 0.045 0.071 0.112 0.059 0.047 0.135 0.063 0.023 0.048 0.023 0.04 0.022


242


Dmax 0 1 1 2 2 2 2 3 3 3 3 3 3 3 4 4 4 4 4 4

NaH -

0 1 0 1 1 -

0 1 1 1 1 1 1 - - - - - -

Dmin - - -

0 0 0 1-4 -

0 0 0 0 1-4 1-4 - -

0 1-4 - -

VoH - - - - -

0 1 - - - - - - - - - - - - -

NaSize - - - - - - - -

0 0 1-4 1-4 - - - - - - - -

Wstat - - - - - - - -

0 1,2 - - - - - - - - - -

Urb - - - - - - - - - -

0 1,4,2,3 - - - - - - - -

Vogord - - - - - - - - - - - -

0 1-4 - - - - - -

Adur - - - - - - - - - - - - - -

0 0 0 0 1 1

LvoisLNa - - - - - - - - - - - - - -

0 1 1 1 0 1

NaH - - - - - - - - - - - - - - -

0 1 1 - -

Comp - - - - - - - - - - - - - - - - - - -

0,1,2,3

Z16 0.008 0.009 0.005 0.038 0 0.009 0.057 0.059 0.054 0.036 0.041 0.04 0.09 0.054 0 0.062 0.048 0.08 0.08 0.066

Z17 0 0.056 0 0.013 0 0.014 0.026 0.063 0.045 0 0.02 0.063 0.081 0.076 0.042 0.093 0.04 0.074 0.067 0.05

Z18 0 0.037 0.005 0.038 0 0.014 0.031 0.059 0.062 0.036 0.01 0.053 0.09 0.022 0.053 0.07 0.048 0.04 0.053 0.082

Z19 0.017 0.009 0.01 0.025 0.02 0.005 0.01 0.042 0.009 0.06 0.061 0.063 0.076 0.054 0.032 0.054 0.048 0.068 0.04 0.066

Z20 0.017 0.056 0.029 0.025 0 0.033 0.041 0.055 0.018 0.036 0.122 0.063 0.062 0.054 0 0.039 0.07 0.074 0.013 0.088

Z21 0.017 0.028 0 0.063 0 0.005 0.016 0.055 0.009 0 0.01 0.007 0.028 0.011 0.021 0.101 0.088 0.051 0.04 0.094

Z22 0.008 0 0 0.051 0.01 0.009 0.016 0.076 0.018 0.024 0.01 0.017 0.019 0.005 0.011 0.101 0.026 0.04 0.053 0.1

Z23 0.041 0.037 0.005 0.038 0 0.014 0.016 0.055 0 0.024 0 0.007 0.019 0 0.095 0.109 0.066 0.085 0.053 0.082

Z24 0.017 0.047 0.01 0.025 0 0 0.016 0.067 0 0 0.02 0.003 0.005 0 0.042 0.109 0.075 0.091 0.04 0.094

Z25 0 0.047 0.01 0.038 0 0 0.021 0.067 0.018 0.012 0.02 0.007 0.028 0.027 0.063 0.109 0.026 0.062 0.027 0.075

N 121 107 204 79 98 214 193 238 112 84 98 303 211 185 95 129 227 176 75 319


243


Dmax 4 4 4 4 4 4 4 4 4 4 4

Dmin - - - 0 1-4 - - - - - -

Wstat - - 0,2 0,2 0,2 1 - - - - -

Vogord 0-1 2-4 0 0 0 0 1-2 1-2 1-2 1-2 3-4

Adur 1 1 2 2 2 2 2 2 2 2 2

LvoisLNa 1 1 - - - - - - - - -

Comp 4 4 - - - - - - - - -

pAge - - 0,1 2,4,3 2,4,3 - - - - - -

Aty - - - - - - 0,1,10,8,7,4,5,9,6 0,1,10,8,7,4,5,9,6 0,1,10,8,7,4,5,9,6 2,3 -

SEC - - - - - - 0,2,3 0,2,3 1 - -

Abt - - - - - - 0,4,5 1,2,3 - - -

Z1 0.01 0.009 0.006 0.013 0 0 0 0 0 0.006 0.011

Z2 0.01 0 0 0 0 0.014 0 0 0 0.017 0.004

Z3 0 0.018 0.006 0 0 0.014 0 0.01 0 0 0

Z4 0.015 0.018 0.019 0 0 0 0 0 0.047 0.011 0.004

Z5 0 0 0 0 0 0 0.011 0.005 0 0.023 0.011

Z6 0.044 0.018 0 0.013 0.005 0 0 0 0.012 0.023 0.011

Z7 0.024 0.009 0.012 0.013 0.005 0.014 0 0.016 0 0.011 0.011

Z8 0.029 0.045 0 0 0.005 0.007 0.011 0 0.035 0.046 0.03

Z9 0.015 0.018 0 0.013 0.005 0 0 0.01 0.024 0.023 0.023

Z10 0.005 0.027 0 0 0 0 0 0.021 0.035 0.017 0.03

Z11 0.054 0.018 0 0.051 0.011 0.014 0 0.01 0.012 0.029 0.015

Z12 0.02 0.018 0 0 0 0.034 0 0.01 0.024 0.017 0.019

Z13 0.015 0.055 0 0.025 0 0.014 0 0.005 0.047 0.011 0.03

Z14 0.01 0.082 0.012 0 0.011 0.027 0.022 0 0 0.046 0.026

Z15 0.005 0.036 0.012 0.013 0 0.014 0.043 0.01 0 0.069 0.053

R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31

244


Dmax 4 4 4 4 4 4 4 4 4 4 4

Dmin - - - 0 1-4 - - - - - -

Wstat - - 0,2 0,2 0,2 1 - - - - -

Vogord 0-1 2-4 0 0 0 0 1-2 1-2 1-2 1-2 3-4

Adur 1 1 2 2 2 2 2 2 2 2 2

LvoisLNa 1 1 - - - - - - - - -

Comp 4 4 - - - - - - - - -

pAge - - 0,1 2,4,3 2,4,3 - - - - - -

Aty - - - - - - 0,1,10,8,7,4,5,9,6 0,1,10,8,7,4,5,9,6 0,1,10,8,7,4,5,9,6 2,3 -

SEC - - - - - - 0,2,3 0,2,3 1 - -

Abt - - - - - - 0,4,5 1,2,3 - - -

Z16 0.073 0.073 0.049 0.051 0.063 0.021 0.108 0.052 0.059 0.023 0.04

Z17 0.102 0.091 0.019 0.038 0.1 0.041 0.043 0.078 0.012 0.017 0.026

Z18 0.059 0.055 0.049 0.063 0.053 0.021 0.022 0.016 0.024 0.051 0.034

Z19 0.029 0.055 0.031 0.063 0.042 0.068 0 0.031 0.035 0.04 0.064

Z20 0.034 0.109 0.091 0.076 0.074 0.027 0.065 0.021 0.059 0.04 0.14

Z21 0.151 0.055 0.111 0.127 0.153 0.144 0.032 0.14 0.094 0.08 0.068

Z22 0.024 0.055 0.074 0.203 0.074 0.137 0.172 0.135 0.106 0.04 0.064

Z23 0.098 0.036 0.123 0.089 0.121 0.144 0.075 0.119 0.071 0.086 0.087

Z24 0.059 0.055 0.216 0.101 0.105 0.082 0.086 0.176 0.145 0.128 0.087

Z25 0.117 0.045 0.241 0.051 0.174 0.164 0.312 0.135 0.165 0.154 0.155

N 205 110 162 79 190 146 93 193 85 175 265

R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31

245

TABLE AV-1.2 Car Allocation to Non-Work Tour

distm 0 0 0 0 0 0 1 1 1 1 1 1 2 2 2 2 2 3-4

distf 0 0 0 0 0 1-4 0 1 1 1 1 2-4 - - - - - -

ntouf 0 0 0 0 1 - - - - - - - - - - - - -

nshim 0,2 0,2 0,2 1 - - - - - - - - - - - - - -

nNWm 1,4 1,4 2,3 - - - - - - - - - - - - - - -

nNWf 1,4 2,3 - - - - - - - - - - - - - - - -

TTcbF - - - - - - - 0-1 0-1 0-1 2-3 - - - - - - 0-1

AtourF - - - - - - - 2,10,4,3,6,5,7 2,10,4,3,6,5,7 8,9 - - - - - - - -

nNWf - - - - - - - 1 2,4,3 - - - - - - - - -

ntoum - - - - - - - - - - - - 0 0 0 0 1 -

Urban - - - - - - - - - - - - 0,1,2,3 0,1,2,3 4 4 - -

TrAcM - - - - - - - - - - - - 0 1 - - - -

RParkM - - - - - - - - - - - - - - 0 1-4 - -

AtourM - - - - - - - - - - - - - - - - - 2,3,4,5,8

nAcToF - - - - - - - - - - - - - - - - - 1,3

Male 0.348 0.333 0.27 0.514 0.074 0.257 0.37 0.731 0.555 0.425 0.492 0.365 0.553 0.823 0.784 0.691 0.364 0.743

Female 0.054 0.222 0.059 0.051 0.015 0.448 0.093 0.064 0.235 0.087 0.063 0.444 0.107 0.048 0.166 0.103 0.127 0.141

None 0.598 0.444 0.672 0.435 0.912 0.295 0.537 0.205 0.21 0.487 0.445 0.19 0.34 0.129 0.05 0.206 0.509 0.115

N 92 81 204 138 136 105 54 78 119 80 191 63 206 62 199 68 55 269

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18

246

TABLE AV-1.2 Car Allocation to Non-Work Tour

distm 3-4 3-4 3-4 3-4 3-4 3-4 3-4 3-4 3-4 3-4

TTcbF 2-3 2-3 - - - - - - - -

AtourM 2,3,4,5,8 2,3,4,5,8 2,3,4,5,8 2,3,4,5,8 6,7 6,7 6,7 6,7 9,10 9,10

nAcToF 1,3 1,3 2 2 - - - - - -

nleim 0 1 0 1 - - - - - -

RParkM - - - - 0-1 0-1 0-1 2-4 - -

NWdurM - - - - 0-2 3 3 - - -

pAge - - - - - 0,2 1,3,4 - - -

BTM - - - - - - - - 0.1 2,3

Male 0.79 0.619 0.836 0.679 0.739 0.952 0.796 0.648 0.722 0.423

Female 0.065 0.048 0.147 0.198 0.188 0.024 0.185 0.185 0.069 0.115

None 0.145 0.333 0.017 0.123 0.073 0.024 0.019 0.167 0.208 0.462

N 62 63 116 81 165 84 108 54 72 52

R19 R20 R21 R22 R23 R24 R25 R26 R27 R28

247

AUTHOR INDEX

A

Aitken, 10

Anggraini, 95, 115, 154, 168

Antill, 9

Arentze, 3, 4, 27, 28, 44, 46-47, 58,

60-61, 63, 70,80-81, 94, 99-100,

110, 113, 124, 132-133, 142, 160,

178

Atkinson, 22

B

Ben-Akiva, 29

Benson, 70

Bhat, 3, 14, 24, 31, 44, 70, 94, 113

Bianco, 70

Boarnett, 10

Borgers, 10, 11, 16, 19, 21, 22, 44

Bowman, 29, 70

Bradley, 29, 113

Breiman, 79, 99

Brijst, 28

C

Cao, 12

Carrasco, 15, 69, 114

Chai, 12

Chandraskaran, 13, 132

Charypar, 27

Cotton, 9

Crane, 10

Curry, 18

D

Davis, H., 18

Davis, J., 10

Donnelly, 29, 30

Dijst, 27

Dowling, 9

Dueker, 10

E

Elliasberg, 18

Eluru, 31

Emery, 18

Ettema, 9, 10, 12

F

Fagnani, 9

Fosgerau, 29

Friedman, 79, 99

Frick, 27

Fujii, 3, 13, 23, 30, 70

Fujiwara, 21

G

Gliebe, 23, 30, 44, 94, 113, 132

Golob, 8, 10, 113, 132

Gordon, 9

Goulias, 13, 27, 47, 113, 132

Gronau, 8, 10

Gupta, 19

Guo, 3, 70

H

Habib, 114

Hagerstrand, 27

Hanson, S., 9, 10

Hanson, P., 9, 10

Harsanyi, 18

Hickman, 9

Hofman, 10, 11

Huber, 9

Hunt, 8, 69

248

I

Iovanna, 8

Isobe, 29

J

Janssenns, 28

Joh, 27

Johnston, 9

K

Kanaroglou, 23, 44, 94

Kawakami, 29

Kapoen, 27

Kandker, 47

Kass, 79, 99, 142, 160

Keeney, 18

Kikuchi, 30

Kim, 8

Kitamura, 3, 13, 30, 70

Knijn, 10

Kockelman, 2

Koppelmann, 16, 23, 29, 30, 44, 94,

113, 132

Kostyniuk, 13

Krishnamurti, 18

Kruchten, 25

Kumara, 23

Kwan, 9, 10, 27

L

Lawson, 8, 70

Lawton, 70

Lee, 9

Lemp, 2

Lenntrop, 27

Lillidahl, 9

Livne, 19

Lu, 92

Lula, 70

M

McNally, 1, 10, 113, 132

McWethy, 2

Meka, 23

Menasco, 18, 19

Messer, 18

Metcalf, 9

Miller, 15, 25, 44, 45, 46, 69, 70,

94, 114

Molin, 18

Morita, 14

Morris, 9, 10

Munsinger, 18

N

Nagel, 27

Nash, 19

Niemeier, 9, 113

Nurul, 46

O

Olshen, 79, 99

Oppewal, 18

Ortuzar, 1

P

Parkany, 92

Pas, 70

Pendhyala, 23, 30, 70, 113

Petersen, 8, 15, 26, 30, 69

Pinch, 9

Pinjari, 31

Ponje, 10

Pratt, 9

Presser, 9, 10

Pribyl, 27

Q

Quinlan, 79, 99

249

R

Redman, 10

Ren, 8

Rigeaux, 18

Roorda, 15, 25, 45, 46, 69, 70, 94

S

Schwanen, 9, 10

Scott, 23, 44, 94

Senbil, 13

Singell, 9

Sivakumar, 3, 31, 69

Spitze, 9

Srinivasan, 3, 14, 24, 31, 44, 69, 94

Steed, 113

Stone, 79, 99

Stopher, 9

Storey, 9

Strathman, 10

T

Timmermans, 3, 4, 7, 9-11, 16, 18, 19,

21-22, 27-28, 44, 46-47, 58, 60-61,

63, 70, 80-81, 94, 99-100, 110,

113, 124, 132, 133, 142, 154, 160,

168, 178

Turner, 9

V

Vadarevu, 9

Van ark, 19

Vance, 8

Van der lippe, 12

Van hoof, 28

Veldhuisen, 27, 28

Vidacovie, 27

Vovsha, 8, 15, 24, 25, 26, 29, 30, 69-

70, 113

W

Weber, 18

Wen, 16, 29

Wets, 28

White, 9

Willumsen, 1

Winkler, 18

Y

Yamamoto, 30

Z

Zhang, 16, 19, 21, 22, 44

250

SUBJECT INDEX

A

action variable, 79-82, 84-87, 94, 97,

99-101, 104, 107, 123-124, 138-

139, 142, 145-146, 148, 160, 163,

167

activity allocation, 4, 32, 51, 65, 95,

98, 108, 195

activity based, 2-4, 7, 26, 31-32, 44-

46, 48, 51, 65, 70, 93-94, 112, 114,

131-132, 155-156, 171, 190-191,

194-195

activity diary, 44, 47-48, 57, 60, 65,

68, 74, 94, 98, 117, 138, 153, 155,

activity level, 51, 60, 82, 84, 101, 103,

119, 121-123, 137, 139, 141, 147-

148, 150, 169, 173, 186,

activity generation, 2, 26, 31, 95, 189,

195

activity participation, 4, 5, 8, 10, 13,

15-17, 23, 28, 32, 34, 44-45, 48,

51, 65, 93-95, 97-98, 102-105,

107-110, 112-114, 129, 131-132,

189-192, 194, 196, 198

activity scheduling, 32, 44, 47-48, 52,

57, 60, 65, 68, 70, 76, 90, 93-95,

110, 113-115, 130-131, 153-154,

168-169

activity selection, 45, 59, 97, 103-104,

106, 109, 122, 191

activity travel pattern, 2-3, 5, 8, 21, 25,

27-28, 30-31, 45, 113, 171-172,

190-191, 194

activity travel behavior, 13-14, 24, 31,

48, 51, 74, 98, 117, 123, 138, 150,

155, 171, 191, 194-195, 197

activity type, 5, 14, 24, 26-28, 30, 48-

50, 71, 94-95, 97-98, 102-104, 108,

110, 112, 115, 117, 119, 121-122,

125-127, 133, 135, 138, 140, 146-

149, 151, 155-156, 173-174, 182,

184, 189, 191, 193, 196-198

agenda, 2, 3, 43, 113

albatross, 3,4,6, 32, 44-51, 58-59, 61-

62, 65-66, 70-71, 73, 80, 90, 94-95,

98-99, 109-110, 113-114, 117, 123,

127, 129, 131-133, 135-136, 138,

142, 149, 150, 153, 155-157, 169,

171-172, 177-179, 188, 191-196,

198

algorithm, 15, 27-28, 45, 79, 82, 90,

94, 99, 101, 109, 119, 123, 127,

139, 142, 150, 153, 160, 163, 168,

195

attribute, 2, 13, 15, 48-51, 68, 74, 80,

82, 87, 90, 93,98-99, 101, 117, 120,

122, 125, 131, 136-139, 141, 145-

148, 150, 153, 155, 162, 163, 166,

173, 176, 178-179, 187, 192, 194

B

behavior, 1-3, 8

bring/get, 11, 45, 49, 50, 71, 95, 96,

98, 102, 103, 104, 106, 107, 109,

110, 115, 117, 118, 119, 122, 127,

130, 142, 144, 145, 152, 162, 165,

175, 177, 183, 185, 187, 189, 192,

197

C

car allocation, 4, 8, 15, 31, 44, 51-52,

57, 59-60, 66, 68-71, 73-77, 79-80,

251

82-91, 95, 97, 115, 153-155, 157-

159, 161-163, 165-170, 191-192,

194-195

car deficient, 5, 8, 51-52, 60, 66, 68-

70, 88, 90, 153-155, 168-170, 195

car ownership, 8, 15, 50, 69, 118, 125,

132, 139-140, 146, 148, 186

chaid, 5, 28, 44, 60, 77, 79-80, 84, 90,

94, 98-99, 104, 107, 109, 123-124,

127, 131, 142, 145, 150, 153, 159-

160, 165, 168, 195

chi square, 64, 65, 79-81, 88, 99-100,

104, 107, 123, 142, 145-149, 159-

160, 166-167, 172-173, 176

choice facet, 2, 3, 7, 26

computational process model, 44, 71,

94, 142, 153,

condition variable, 45, 52, 68, 73, 79-

84, 87, 89-90, 94, 97-108, 110,

119-125, 127, 136-140, 142, 147,

149-150, 158, 160-163, 167-168,

196

confusion matrix, 86, 89

contingency coefficient, 64, 104, 107,

145-146, 148, 166-167

continuous choice, 60,-61, 63, 65, 112

continuous variable, 47, 58-59, 82,

101, 114, 141, 162

constraint, 2, 4, 12

constraint-based, 27

D

decision making, 1, 2, 4, 5, 7

decision rules, 4, 5, 44, 46-47, 77, 79,

84-85, 90, 94, 104, 107, 109, 114,

124, 125, 127, 136, 146-147, 149,

166, 169, 170, 172, 197

decision tree, 5, 28, 44-45, 47, 52, 57,

60-65, 68, 77, 79-80, 85, 90, 93-95,

97-99, 101-102, 104, 107, 110,

112, 115, 117, 119, 124-127, 132,

136, 143, 146-147, 149, 151, 154,

156, 160-161, 168-169, 172, 174,

192, 196- 198

departure time, 2, 59, 74, 113-114,

135

destination, 1, 2, 7, 8, 26, 59, 74

detour time, 5, 58, 131-133, 136, 196

dimension, 2, 62, 114, 132, 136-137,

189, 192

discrete choice, 2, 47, 60-61, 63-64

discretionary, 9-10, 14, 24, 48, 96,

115, 132, 167, 186-187, 194, 196

duration, 2, 5, 9-10, 12, 14, 23-24, 27-

28, 45, 51-52, 57-59, 61, 71, 74,

77-79, 82-83, 85, 89-90, 95, 97,

101-104, 107-108, 110, 112-115,

118-127, 135, 137-138, 141-142,

147-150, 152, 157, 159, 163-166,

169-170, 192, 194-195, 197-198

E

econometric, 3, 31, 46

episode, 2, 72

escorting, 25, 29, 46

F

forecasting, 1, 7, 113

four-step model, 1, 26, 46, 196

fixed activity, 95

flexible activity, 95

f-statistic, 65, 123, 124, 125, 127, 197

G

generation module, 51, 53, 57

goodness of fit, 21, 22, 63, 90, 93,

110, 146, 149, 151, 154, 169, 173,

189, 192, 195, 197-198

grocery shopping, 5, 16, 114, 191

252

H

hit ratio, 80, 85, 99, 104, 107, 146,

149, 168

household decision making, 3, 4, 5, 7,

8, 15, 23, 26-27, 32, 43-, 45, 46,

51, 57, 65, 96, 113, 132-133, 154,

171, 178, 191-192, 194-195

hold out set, 94

household activity, 3, 27, 28, 51, 106,

109, 110, 140, 155, 195

household member, 3, 4, 8, 13-19, 21-

25, 28-30, 43, 45, 47-48, 51, 59,

69, 74, 98, 112, 114-115, 117, 130,

133, 135, 139, 156, 159, 171, 191-

192, 194, 197

household task, 4, 5, 9, 10, 12, 17, 21,

43, 48, 51-52, 59, 79, 93-95, 97,

102-104, 107-110, 114-115, 134,

156, 186, 196

household level, 7, 30, 31, 46, 51, 66

human decision making, 1, 4, 52

I

impact table, 77, 80, 87, 89-90, 94, 99,

104, 107, 110, 124-125, 142, 146-

149, 153, 155, 160, 166-167, 169,

198

independent activity, 17, 53, 58, 60

individual activity, 8, 27, 30, 194

individual decision making, 113, 154,

190

individual travel pattern, 3, 4

individual level, 14, 30-31, 46, 57, 82,

101, 121, 151, 191

induction method, 61, 69

integration, 6, 53, 57

interdependency, 2-3, 15, 31, 46, 58,

134, 190-191

institutional constraint, 2

intra household interaction, 29, 30

J

joint activity, 4, 5, 13-16, 23-25, 28-

30, 32, 44-45, 48, 51-52, 57, 59,

65, 93, 96, 112-115, 117, 119,

121-122, 124-127, 131-132, 136-

138, 140-141, 143-146, 148-151,

158, 189-192, 194, 196, 198

joint participation, 13, 14, 29-30, 52,

60, 93-96, 112, 114, 127, 129, 134,

141, 156, 193, 195-196

joint decision making, 11, 32, 50-51,

57, 59, 66, 94-95, 129

L

land use, 68, 82, 121, 132, 138, 163

leaf node, 61-65, 80, 84-85, 99, 104,

107, 124, 145, 148, 165-166

leisure, 5, 10-16, 23, 49-50, 71, 96,

102, 104, 113-115, 117-119, 125-

126, 133-134, 141, 143-144, 151,

156-157, 161, 164-165, 167, 169,

174, 182, 184, 186-187, 196-197

likelihood, 10, 63-64, 84, 99

location choice, 5, 30, 52, 58-59, 68-

69, 77, 79-80, 90, 131-133, 143,

148-150, 153, 156, 168, 195-196

long term, 2, 8

M

maintenance activity, 10, 12-13, 16,

23, 29-30, 39, 43, 70, 113, 133

mandatory, 49, 58, 72

microsimulation, 3

mode choice, 67

mon data, 4, 49, 66, 74-75, 78, 99

monotonicity, 87-88, 101

multinomial logit, 11, 15

253

N

non monotonous, 81, 88, 89, 100

non-work activity, 5, 8, 23-24, 49, 52,

57-59, 103, 115, 121-122, 125,

135, 137, 156, 158-159, 162-164,

170, 176

non-work tour, 5, 52, 57, 59-60, 66,

73, 97, 154-157, 159-160, 163-170,

192-193, 196, 198

non-household task, 5, 51, 95

null model, 65, 85

O

observation, 66

operational, 2, 52, 95

origin, 59, 74

P

paradigm, 1

pattern, 58,

person level, 51, 53, 66

postcode, 58, 82, 88, 132, 135-137,

139, 147, 150, 164, 179, 188

prediction, 45-46, 50, 57, 60-61, 64-

65, 80, 86, 99, 107, 110, 124-125,

127, 139, 143, 151, 173-174, 177-

178, 180, 188-190, 193, 197

probability, 8, 10, 12, 27-28, 30, 45,

61-63, 68, 85-86, 88-90, 97, 104,

106-109, 144-145, 154, 156, 168-

170, 174, 198

probabilistic theta, 64

probabilistic assignment, 80, 87

R

route choice, 1, 2

road capacity, 1

rule-based, 3, 5, 15, 27, 44-45, 47, 80,

94, 100, 109, 123-124, 127, 143,

151, 161, 196

resource allocation, 3, 4, 7, 27, 32, 44,

48, 59, 65, 136, 191-192, 195

S

schedule level, 51, 103, 121, 138, 140,

142, 161, 174, 177

sequential, 5, 6, 52,

service, 11-12, 49-50, 91, 95-96, 98,

102, 104, 107-109, 115, 117-118,

133-134, 141, 143-144, 156-157,

161, 164, 174, 182, 184, 186, 196-

197

shared activity, 20, 29

shopping, 5, 11-14, 16-17, 49-50, 71,

95-96, 98, 102, 104, 107-109, 113-

115, 117-119, 126, 134, 136-137,

141, 143-144, 150-151, 156-159,

161, 164-165, 167, 169, 174, 182,

184, 186, 190, 196-197

short-term, 8

skeleton, 28, 46, 47, 51

simulation model, 27

socio-demographic, 51

socio-economic, 1, 13, 69, 79, 82

social activity, 5, 51, 97

situational constraint, 2

space-time constraint, 2, 27-28, 45, 58,

133-134

space-time prism, 58, 134, 136, 141

spatial, 1, 10, 28, 47, 68, 110, 125,

139

split criterion, 64, 65, 123, 143

standard deviation, 66

start time, 28, 45, 52, 57, 58, 59, 61,

71, 97, 98, 112, 113, 114, 115, 117,

119, 122-127, 130, 135, 139, 141-

142, 156-157, 162-163, 169-170,

183, 188, 190, 192, 197-198

T

254

task allocation, 3-5, 8, 10-12, 16,

27, 30-31, 44, 48, 59, 79, 93-94,

97, 102-104, 107, 109-110, 134,

156, 189-190, 192, 194-196, 198

time allocation, 8, 10, 15

time of day, 4, 8, 51, 173-174, 176,

184, 187-188, 198

time use, 3, 16, 20-21, 23, 138

timing, 2, 46, 77, 96

tour-based, 2, 158

tour level, 154, 161, 164, 168, 169,

174, 177, 198

touring, 49-50, 72, 96, 102, 104, 115,

117-118, 125-126, 133-134, 137,

143-144, 151, 156-157, 161, 164,

167, 169, 186, 196-197

trade off, 70

traditional, 1, 2

traffic flow, 1

training set, 61, 63, 85-86, 89, 99, 124,

142, 145, 146, 166-167

transport demand, 1-4, 7, 32, 65, 70,

154, 191-192, 194

transport mode, 1-2, 7-8, 16, 26, 44-

45, 51-52, 57-60, 66, 69, 71, 73-74,

90, 94-98, 103, 117, 120, 122, 134,

138, 154-156, 165, 168-169, 176-

177, 187-189, 191, 195

travel arrangement, 15, 24, 25, 69

travel demand, 1, 3, 5, 7, 14, 32, 51,

65, 69, 93, 114, 132-133, 179,

192-193

travel behavior, 1, 13-14, 24, 31, 48,

51, 98, 113, 133, 139. 154, 156

travel mode, 2, 15, 16, 29, 59, 135

trip-based, 1, 2, 70, 158

trip-chaining, 5, 8, 9, 60, 74

trip-generation, 1, 31

travel pattern, 2, 3, 4, 14, 114, 1

U

urban planning, 1

unit of analysis, 2

utility function, 2, 15, 17-23

utility maximization, 23, 20, 27, 29,

45, 95

V

validation set, 87, 90, 99

variable, 10-16, 20-21, 23-25, 30-31,

45-47, 50, 52, 58-59, 61, 63, 68-71,

73, 79-85, 87-90, 93-94, 97, 99-

105, 108-110, 117, 119-125, 132,

137-143, 147-150, 154, 156, 159,

161-164, 167-170, 173-175, 177,

180, 183, 187-189, 197

W

work activity, 13, 23-24, 51-52, 57-60,

70-71, 73-74, 77-79, 82, 84-86, 88,

90-91, 95-96, 101, 103, 106, 110,

120-122, 127, 134-135, 155, 159,

163-164, 177, 190, 196, 198

work related, 52, 53, 59

work tour, 5, 59

255

List of Publications

International Scientific Journal

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2008). Car Allocation between

Household Heads in Car Deficient Households: A Decision Model. European Journal of Transportation and Infrastructure Research, 8(4), pp. 301-319.

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2009). Continuous Choice Model

of Timing and Duration of Joint Activities. Transportation Research Record: Journal of the Transportation Research Board, No. 2135: Travel Behavior 2009, Volume 2, pp.

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2010). Car Allocation Decisions in

Car-Deficient Households: The Case of Non-Work Tours. Transportmetrica Journal (forthcoming).

Conference Proceedings

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2006). A Model of Within-

Household Travel Activity Decisions Capturing Interactions Between Household

Heads. In J.P van Leeuwen and H.J.P. Timmermans (eds). Progress in Design and Decision Support Systems in Architecture and Urban Planning , Eindhoven

University of Technology, The Netherlands, pp. 19-33.

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2007). Refining Albatross:

Modeling Household Activity Generation and Allocation Decisions Using

Decision Tree induction. In: Proceeding of the 11th WCTR Conference, UC

Berkeley, USA 2007. (CD-ROM, pp. 21).

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2007). Modeling Car Allocation

Decisions in Automobile Deficient Households. In: Proceeding ETC 2007 Conference, Noordwijk, The Netherlands. (CD-ROM, pp. 22).

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2008). Using the Activity-based

ALBATROSS Model to Support Transport Planning in Indonesia. In: Proceeding ISSM 2008, Delft, The Netherlands. (CD-ROM, pp. 85-91).

256

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2008). Modeling Joint Activity

Participation and Household Task Allocation. In: Proceeding AATT 2008 Conference, Greece, Athens. (CD-ROM, pp. 15)

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2009), Gender Roles and Activity-

Travel Patterns. In: Proceedings Household Activity-Travel Behavior Analysis for Urban Policy-Making, JSPS-NOW Workshop, Eindhoven, The Netherland. (On-

line, pp. 31)

Anggraini, R., Arentze. T.A., Timmermans, H.J.P. (2009). Household Location Choice

Models for Independent and Joint Non-Work Activities. In: Proceedings XIII Euro Working Group on Transportation Meeting, Padua, Italy. (CD-ROM, pp. 9)

Anggraini, R., Arentze. T.A., Timmermans, H.J.P., and Feng, T. (2009). Modeling

Household Activity Participation in a Rule-based System of Travel Demand:

Decision of Two Household Heads. In: Proceedings EASTS 09, Surabaya,

Indonesia (On-line, pp. 15)

257

CURRICULUM VITAE

Renni Anggraini was born in Banda Aceh, Indonesia, in September 23, 1971. She

graduated from the Syiah Kuala University, Banda Aceh as a Civil Engineer in 1996.

Soon after the graduation, she became an academic staff in the Civil Engineering

Department of Faculty of Engineering at the same university in 1997. In accordance

with her research interest, she was appointed at the Transportation group. A year later,

she was awarded a competitive fellowship for post-graduate studies under Japanese

government scholarship, namely MONBUSHO. With this award, she further continued

her study at the Department of Civil and Environmental Systems Engineering at

Nagaoka University of Technology, Japan. She graduated as a Master of Engineering

in 2001. During these two years, her major research project concerned about the

application of discrete choice models to simulate individual activity-travel behavior in

Nagaoka city, Japan.

From 2005 – 2009 she was a PhD student at Urban Planning Group at Eindhoven

University of Technology, the Netherlands. Her research still focused on activity-travel

behavior, expanding the individual decision making to household decision making. She

was also exploring an alternative modeling approaches, especially those established on

rule-based models or computer process models.

Her current research interests are in the areas of urban and transport planning, activity-

travel behavior, transport demand management, transport in developing countries, and

various other domains.

Shortly after completing her PhD study, she will be homecoming to the Syiah Kuala

University in Banda Aceh to pursue her career as an academic staff in the Department

of Civil Engineering.

BOUWSTENEN is een publikatiereeks van de Faculteit Bouwkunde, Technische Universiteit Eindhoven.Zij presenteert resultaten van onderzoek en andere aktiviteiten op het vakgebied der Bouwkunde, uitgevoerd in het kader van deze Faculteit.

BOUWSTENEN zijn telefonisch te bestellen op nummer040 - 2472383

KernredaktieMTOZ

Reeds verschenen in de serie BOUWSTENEN

nr 1Elan, a computermodel for building energy design, theory and validationM.H. de WitH.H. DriessenR.M.M. van der Velden

nr 2Kwaliteit, keuzevrijheid en kostenEvaluatie van experiment Klarendal, Arnhemdrs J. SmeetsC. le Nobel, arch. HBOM. Broos, J. Frenken, A. v.d. Sanden

nr 3Crooswijkvan 'bijzonder' naar 'gewoon'drs V. Smitir K. Noort nr 4Staal in de woningbouwir E.J.F. Delsing

nr 5Mathematical theory of stressed skin action in profiled sheeting with various edge conditionsir A.W.A.M.J. v.d. Bogaard

nr 6Hoe berekenbaar en betrouwbaar is de coëfficiënt k in x - ko en x - ks?ir K.B. Lubdrs A.J. Bosch

nr 7Het typologisch gereedschapEen verkennende studie omtrent typologie en omtrent de aanpak typologisch onderzoek J.H. Luiten arch. HBO

nr 8Informatievoorziening en beheerprocessenir A. Nauta / drs J. Smeets (red.)Prof. H. Fassbinder (projectleider)ir A. Proveniers, drs J.v.d. Moosdijk

nr.9Strukturering en verwerking van tijdgegevens voor de uitvoering van bouwwerkenir W.F. Schaeferir P.A. Erkelens

nr 10Stedebouw en de vorming van een speciale wetenschapK. Doevendans

nr 11Informatica en ondersteuning van ruimtelijke besluitvormingdr G.G. van der Meulen

nr 12Staal in de woningbouw, korrosie-bescherming van de begane grondvloerir E.J.F. Delsing

nr 13Een thermisch model voor de berekening van staalplaatbeton- vloeren onder brandomstandighedenir A.F. Hamerlinck

nr 14De wijkgedachte in NederlandGemeenschapsstreven in een stedebouwkundige contextdr ir K. Doevendansdr R. Stolzenburg

nr 15Diaphragm effect of trapezoidally profiled steel sheets. Experimental research into the influence of force applicationir A.W.A.M.W. v.d. Bogaard

nr 16Versterken met spuit-ferrocement.Het mechanische gedrag van met spuit-ferrocement versterkte gewapende betonbalkenir K.B. Lubir M.C.G. van Wanroy

nr 17De tractaten van Jean Nicolas Louis Durandir G. van Zeyl

nr 18Wonen onder een plat dak.Drie opstellen over enkele vooronder-stellingen van de stedebouwdr ir K. Doevendans

nr 19Supporting decision making processesA graphical and interactive analysis of multivariate datadrs W. Adams

nr 20Self-help building productivityA method for improving house building by low-income groups applied to Kenya 1990-2000ir P. A. Erkelens

nr 21De verdeling van woningen: een kwestie van onderhandelendrs V. Smit

nr 22Flexibiliteit en kosten in het ontwerp- proces Een besluitvormingonder-steunend modelir M. Prins

nr 23Spontane nederzettingen begeleidVoorwaarden en criteria in Sri Lankair P.H. Thung

nr 24Fundamentals of the design of bamboo structuresO. Arce-Villalobos

nr 25Concepten van de bouwkundeProf. dr ir M.F.Th. Bax (red.) dr ir H.M.G.J. Trum (red.)

nr 26Meaning of the siteXiaodong Li

nr 27Het woonmilieu op begrip gebrachtJaap Ketelaar

nr 28Urban environment in developing countrieseditors: dr ir Peter A. Erkelens dr George G. van der Meulen

nr 29Stategische plannen voor de stadOnderzoek en planning in drie stedenProf. dr H. Fassbinder (red.)ir H. Rikhof (red.)

nr 30Stedebouwkunde en stadsbestuurir Piet Beekman

nr 31De architectuur van DjennéEen onderzoek naar de historische stad P.C.M. Maas

nr 32Conjoint experiments and retail planningHarmen Oppewal

nr 33Strukturformen Indonesischer Bautechnik Entwicklung methodischer Grundlagen für eine 'konstruktive pattern language' in IndonesienHeinz Frick

nr 34Styles of architectural designingEmpirical research on working styles and personality dispositionsAnton P.M. van Bakel

nr 35Conjoint choice models for urban tourism planning and marketingBenedict Dellaert

nr 36Stedelijke Planvorming als co-produktieProf. dr H. Fassbinder (red.)

nr 37 Design Research in the Netherlandseditors: Prof. dr R.M.Oxman, Prof. dr ir. M.F.Th. Bax,Ir H.H. Achten

nr 38 Communication in the Building IndustryBauke de Vries

nr 39 Optimaal dimensioneren van gelaste plaatliggers

nr 40 Huisvesting en overwinning van armoededr.ir. P.H. Thung en dr.ir. P. Beekman (red.)

nr 41 Urban Habitat: The environmentof tomorrowGeorge G. van der Meulen, Peter A. Erkelens

nr 42A typology of jointsJohn C.M. Olie

nr 43Modeling constraints-based choices for leisure mobility planningMarcus P. Stemerding

nr 44Activity-based travel demand modelingD. Ettema

nr 45Wind-induced pressure fluctuations on building facadesChris Geurts

nr 46Generic RepresentationsHenri Achten

nr 47Johann Santini AichelDirk De Meyer

nr 48Concrete behaviour in multiaxialcompressionErik van Geel

nr 49Modelling site selectionFrank Witlox

nr 50Ecolemma modelFerdinand Beetstra

nr 51Conjoint approaches to developing activity-based modelsDonggen Wang

nr 52On the effectiveness of ventilationAd Roos

nr 53Conjoint modeling approaches for residential group preverencesEric Molin

nr 54Modelling architectural design information by featuresJos van Leeuwen

nr 55A spatial decision support system forthe planning of retail and servicefacilitiesTheo Arentze

nr 56Integrated lighting system assistantEllie de Groot

nr 57Ontwerpend leren, leren ontwerpendr.ir. J.T. Boekholt

nr 58Temporal aspects of theme park choice behavoirAstrid Kemperman

nr 59Ontwerp van een geïndustrialiseerde funderingswijzeFaas Moonen

nr 60Merlin: A decision support system foroutdoor leisure planningManon van Middelkoop

nr 61The aura of modernityJos Bosman (nog niet gepubliceerd)

nr 62Urban Form and Activity-Travel PatternsDaniëlle Snellen

nr 63Design Research in the Netherlands 2000Henri Achten

nr 64Computer Aided DimensionalControl in Building ConstructionRui Wu

nr 65Beyond Sustainable Buildingeditors: Peter A. Erkelens Sander de Jonge August A.M. van Vlietco-editor: Ruth J.G. Verhagen

nr 66Das globalrecyclingfähige HausHans Löfflad

nr 67Cool Schools For Hot SuburbsRené J. Dierkx

nr 68A Bamboo Building Design Decision Support ToolFitri Mardjono

nr 69Driving rain on building envelopesFabien van Mook

nr 70Heating Monumental ChurchesHenk Schellen

nr 71Van Woningverhuurder naar Aanbieder van WoongenotPatrick Dogge

nr 72Moisture transfer properties of coated gypsumEmile Goossens

nr 73Plybamboo Wall-panels for HousingGuillermo E. González-Beltrán

nr 74The Future Site-ProceedingsGer MaasFrans van Gassel

nr 75Radon transport in Autoclaved Aerated ConcreteMichel van der Pal

nr 76The Reliability and Validity of Interactive Virtual Reality Computer ExperimentsAmy Tan

nr 77Measuring Housing Preferences Using Virtual Reality And Belief NetworksMaciej A. Orzechowski

nr 78Computational Representations of Words and Associations in Architectural DesignNicole Segers

nr 79Measuring and Predicting Adaptationin Multidimensional Activity-Travel PatternsChang-Hyeon Joh

nr 80Strategic BriefingFayez Al Hassan

nr 81Well Being in HospitalsSimona Di Cicco

nr 82Solares BauenImplementierungs- und Umsetzungs-aspekte in der Hochschulausbildung in ÖsterreichGerhard Schuster

nr 83Supporting Strategic Design of workplace Environments with Case-Based Reasoning Shauna Mallory-Hill

nr 84ACCEL: a Tool for Supporting Concept Generation in the Early Design PhaseMaxim IvashkovMaxim Ivashkovnr 85Brick-mortar interaction in masonry under compressionAd Vermeltfoort

nr 86 Zelfredzaam WonenGuus van Vliet

nr 87Een ensemble met grootstedelijke allureJos Bosman/Hans Schippers

nr 88On the Computation of Well-Structured Graphic Representations inArchitectural Design Henri Achten

nr 89De Evolutie van een West-Afrikaanse Vernaculaire ArchitectuurWolf Schijns

nr 90ROMBO tactiekChristoph Maria Ravesloot

nr 91External coupling between building energy simulation and computational fluid dynamicsEry Djunaedy

nr 92Design Research in theNetherlands 2005Editors:Henri AchtenKees DorstPieter Jan StappersBauke de Vries

nr 93Ein Modell zur baulichen TransformationJalil H.Saber Zaimian

nr 94Human Lighting DemandsHealthy Lighting in an Office EnvironmentMyriam Aries

nr 95A Spatial Decision Support System for the Provision and Monitoring of Urban GreenspaceClaudia Pelizaro

nr 96Leren CreërenAdri Proveniers

nr 97SimlandscapeRob de Waard

nr 98Design Team CommunicationAd den Otter

nr 99Humaan-EcologischGeoriënteerde WoningbouwJuri Czabanowski

nr 100HambaseMartin de Wit

nr 101Sound Transmission through Pipe Systems and into Building StructuresSusanne Bron - van der Jagt

nr 102Het Bouwkundig ContrapuntJan Francis Boelen

nr 103A Framework for a Multi-Agent Planning Support SystemDick Saarloos

nr 104Bracing Steel Frames with Calcium Silicate Element WallsBright Mweene Ng'andu

nr 105Naar een nieuwe houtskeletbouwF.N.G. De Medts

nr 106Anatomy of DwellingEnno Wiersma(nog niet gepubliceerd)

nr 107Healing ArchitectureEwa Mosiniak(nog niet gepubliceerd)

nr 108Geborgenheiddrs T.E.L. van Pinxteren(nog niet gepubliceerd)

nr 109Modelling Strategic Behaviourin Anticipation of CongestionQi Han

nr 110Reflecties op het WoondomeinFred Sanders

nr 111On Assessment of Wind Comfort by Sand ErosionGabor Dezso

nr 112Bench Heating in Monumental Churches Dionne Limpens-Neilen

nr 113RE. ArchitectureAna Pereira Roders

nr 114Toward Applicable Green ArchitectureUsama El Fiky

nr 115Knowledge Representation UnderInherent Uncertainty in a Multi-AgentSystem for Land Use PlanningLiying Ma

nr 116Integrated Heat Air and MoistureModeling and SimulationJos van Schijndel

nr 117Concrete behaviour in multiaxial compressionJ.P.W. Bongers

nr 118The Image of the Urban LandscapeAna Moya Pellitero

nr 119The Self-Organizing City in VietnamStephanie Geertman

nr 120A Multi-Agent Planning Support System for Assessing Externalities of Urban Form ScenariosRachel Katoshevski-Cavari

nr 121Den Schulbau neu denken, fühlen und wollenUrs Christian Maurer-Dietrich

nr 122Peter Eisenman Theories and PracticesBernhard Kormoss

nr 123User Simulation of Space UtilisationVincent Tabak

nr 124Moisture Transport in Cavity Brick WallA. Aghai(nog niet gepubliceerd)

nr 125In Search of a Complex System ModelOswald Devisch

nr 126Lighting in Work Environment direct Effects of Lighting Level and Spectrum on Pshychophysiological Variables Grazyna Goricka

nr 127Flanking Sound Transmission trough Lightweight Framed Double Leaf WallsStefan Schoenwald

nr 128Bounded Rationality and Spatial- Temporal Pedestrian Shopping BehaviourWei Zhu

nr 129Travel information Impact on activity travel patternZhongwei(nog niet gepubliceerd)

nr 131Allemaal WinnenM.J. Bakker(nog niet gepubliceerd)

nr 132Architectural Cue Model in EvacuationSimulation for Underground Space DesignChengyu Sun

nr 133Uncertainty and sensitivity analysis in building performance simulation fordecision support and design optimizationChristina Hopfe

nr 134Facilitating distributed collaboration in the AEC/FM sector using Semantic Web TechnologiesJakob Beetz

nr 135Circumferentially Adhesive Bonded Glass Panes for Bracing Steel Frame in FaçadesEdwin Huveners

nr 136Circumferentially Adhesive Bonded Glass Panes for Bracing Steel Frame in FaçadesEdwin Huveners

nr 137Nog niet bekend Mariette van Stralen(nog niet gepubliceerd)

nr 138Nog niet bekendJos Smeets(nog niet gepubliceerd)

nr 139Lateral behavior of steel frames with discretely connected precast concrete infill panelsPaul Teeuwen(nog niet gepubliceerd)

nr 140Nog niet bekendPerica Savanovic(nog niet gepubliceerd)

nr 130Co-simulation for performence prediction of innovative integrated mechanical energy systems in buildingsMarija Trcka

household activity-travel behavior: implementation of within

Documents