USF Libraries
USF Digital Collections

Understanding activity engagement and time use patterns in a developing country context

MISSING IMAGE

Material Information

Title:
Understanding activity engagement and time use patterns in a developing country context
Physical Description:
Book
Language:
English
Creator:
Banerjee, Amlan
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Travel time expenditure
Commuter behavior
Inter-agent interaction
Stochastic frontier model
Structural equations model
International comparison
Dissertations, Academic -- Civil Engineering -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Flourishing economy, rapid industrialization and increasing trend of motorization have been shaping societies in the developing countries like India in an unprecedented manner.Infrastructure backlog amid such rapid growth in all imaginable directions has heavily exacerbated the urban transport crisis in these countries by alarming increase in vehicular travel demand, road fatalities, and environmental pollution. To address urban transport challenges, the necessary development and implementation of effective transport planning and policies have generally lagged in the developing countries compared to that seen in the developed countries due to several constraints including resource constraints, knowledge constraints, institutional constraints and so on. However, in the recent past, with the rapid development seen by several emerging economies and the explosive growth in transportation infrastructure investment, there is a growing interest in the development and implementati on of advanced travel demand modeling systems in developing countries. But lack of necessary research and exploration of travel behavior in a developing country context has left very limited knowledge for us to understand the extent of applicability of these advanced theories and methodologies in a different socio-cultural perspective. Assessing the practical relevance of the subject, this research adopts a comprehensive approach to explore the activity engagement pattern and time use behavior from a developing country standpoint. To accomplish this goal, a series of empirical and analytical studies are performed on a household travel survey data set available from Thane Metropolitan Area in India. The study also introduces new concepts and facilitates enhancements of existing modeling methodologies in the field of travel behavior and time use research. The study results provide very insightful findings and plausible interpretations consistent with a developing country perspective reco gnizing a wide spectrum of differences and similarities in activity patterns and time use behavior between a developed and a developing country. Specified model structures are meaningfully able to incorporate various socio-cultural and institutional constraints and reflected sensitivity to the behavioral variability between the contexts suggesting that advanced analytical techniques may be satisfactorily applied on the data set from developing countries which may contribute important ingredients in the development of advanced activity-based model system in the countries like India.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2006.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Amlan Banerjee.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 225 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001913930
oclc - 174958785
usfldc doi - E14-SFE0001729
usfldc handle - e14.1729
System ID:
SFS0026047:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Understanding Activity Engagement and Time Use Patterns in a Developing Country Context by Amlan Banerjee A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Civil and Environmental Engineering College of Engineering University of South Florida Major Professor: Ram M. Pendyala, Ph.D. Steven E. Polzin, Ph.D. Jian J. Lu, Ph.D. Gabriel Picone, Ph.D. Elaine Chang, Ph.D. Date of Approval: May 31, 2006 Keywords: travel time expenditure, commu ter behavior, inter-agent interaction, stochastic frontier model, structural e quations model, international comparison Copyright 2006, Amlan Banerjee

PAGE 2

DEDICATION To my parents...

PAGE 3

ACKNOWLEDGEMENTS First of all, I would like to thank my doctoral committee members Dr. Ram Pendyala (Chair), Dr. Steve Polzin, Dr. John Lu, Dr. Ga briel Picone and Dr. El aine Chang for their insights and suggestions that have immensely contributed in improving the quality of this dissertation. Each of them has been an outst anding mentor. I would al so like to thank Dr. Xuehao Chu for serving as the Chair for my doctoral dissertation defense. In particular, I wish to express my sincere gratitude to my advisor, Dr. Ram Pendyala for his invaluable guidance and c ontinued support for this research work. Without his generous help, this dissertation would not have been po ssible. I was always inspired by his impeccable attitude towards re search and dedication to the profession. I find myself extremely fortunate to have su ch a wonderful person as my supervisor. Sincere thanks to Dr. S. L. Dhingra fr om Indian Institute of Technology Bombay, India for making Thane Household Tr avel Survey data available. I am also very thankful to my colleague s, in particular, Mr. Xin Ye and former USF graduate students Mr. Ram Nehra and Mr. Abdul Pinjari for helping me greatly at various stages of this research effort. I am indebted to the Department of Civil and Environmental Engineering and Center for Urban Transportation Research at USF for providing excellent research environment and facilities. Finally, I would like to exte nd my heartfelt thanks to my parents and my loving wife, Nivedita for their unconditional sacr ifice, support and in spiration during the development of this dissertation.

PAGE 4

i TABLE OF CONTENTS LIST OF TABLES.............................................................................................................iv LIST OF FIGURES.........................................................................................................viii ABSTRACT....................................................................................................................... xi CHAPTER 1 INTRODUCTION.........................................................................................1 1.1 Urban Transport Trends and Ch allenges in Developing Countries................1 1.1.1 Urbanization........................................................................................2 1.1.2 Motorization........................................................................................3 1.1.3 Economic Growth and Migration.......................................................5 1.1.4 Social Inequality.................................................................................6 1.1.5 Trends in Travel Behavior..................................................................7 1.2 Problem Statement..........................................................................................8 1.3 Research Objectives and Scopes...................................................................10 1.4 Outline of the Dissertation............................................................................11 1.5 Summary.......................................................................................................12 CHAPTER 2 STATE OF THE ART IN TRANSPORTATION MODELING.................13 2.1 Introduction...................................................................................................13 2.2 Review of Activity-Bas ed Time Use Research............................................15 2.2.1 Activity Time Allocation..................................................................15 2.2.2 Activity Episode Analysis.................................................................16 2.2.2.1 Activity Episode Duration.................................................17 2.2.2.2 Activity Episode Sequencing.............................................18 2.2.2.3 Activity Frequency.............................................................20 2.3 Time-Space Interaction.................................................................................20 2.4 Inter-Agent Interaction..................................................................................22 2.5 Emerging Trends in Travel Behavi or in Developing Country Context........23 2.5.1 Demographic Shifts..........................................................................23 2.5.2 Growing Challenges in Transit Sector..............................................24 2.5.3 ICT and Changing Travel Behavior..................................................25 2.5.4 Land Use and Travel Behavior Interaction.......................................26 2.6 Summary.......................................................................................................27

PAGE 5

ii CHAPTER 3 EXPLORATORY ANALYSIS OF TRAVEL CHARACTERISTICS AND TIME USE BEHAVIOR............................29 3.1 Introduction to the Study Areas....................................................................29 3.2 Description of Surveys..................................................................................30 3.3 Household Characteristics............................................................................31 3.4 Person Characteristics...................................................................................34 3.5 Person Characteristics by Commuting Status...............................................37 3.6 Person Characteristics of Zero-Trip Makers.................................................42 3.7 Development of Trip Production Rates........................................................44 3.8 Trip Distribution...........................................................................................64 3.9 Time of Day Distribution..............................................................................68 3.10 Modal Split....................................................................................................70 3.10.1 Modal Split Distri bution by Trip Purpose.....................................70 3.10.2 Modal Split Distributi on by Commuting Status by Trip Purpose...........................................................................................85 3.10.3 Modal Split Distribu tion by Household Car Ownership................93 3.11 Activity and Time Use Characteristics.......................................................101 3.11.1 Trip Frequency Analysis..............................................................102 3.11.2 Analysis of Travel Duration.........................................................103 3.11.3 Analysis of Activity Duration......................................................104 3.12 Trip Length Distribution.............................................................................108 3.13 Trip Chaining Analysis...............................................................................110 3.14 Summary and Discussion............................................................................112 CHAPTER 4 EXPLORATION OF TRAV EL TIME FRONTIER AROUND THE WORLD...................................................................................................114 4.1 Travel Time Budget Versus Travel Time Frontier.....................................114 4.2 Modeling Methodology..............................................................................116 4.3 Data Sets.....................................................................................................119 4.4 Model Estimation Results...........................................................................123 4.5 Distribution of Travel Ti me Expenditures and frontier..............................128 4.6 Summary and Discussion............................................................................132 CHAPTER 5 HOW LOW CAN TRAVEL GO...............................................................134 5.1 Introduction to Minimum Travel Time Threshold......................................134 5.2 Modeling Methodology..............................................................................137 5.3 Data Sets and Sample Characteristics.........................................................140 5.4 Model Estimation Results...........................................................................141 5.5 Distribution of Travel Time Expenditures and Cost Frontiers...................147 5.6 Summary and Discussions..........................................................................153 CHAPTER 6 ANALYSIS OF COMMUTI NG LENGTH CHOICE BEHAVIOR.........155 6.1 Introduction.................................................................................................155 6.2 Data Sets.....................................................................................................158 6.3 Descriptive Statistics of the Commuter Segments......................................158

PAGE 6

iii 6.4 Model Estimation Results...........................................................................165 6.5 Summary and Discussion............................................................................172 CHAPTER 7 ACTIVITY-TRAVEL INTERACTION AMONG ADULT HOUSEHOLD MEMBERS....................................................................174 7.1 WithinHousehold Interac tion in Activity Engagement.............................174 7.2 Data Sets and Sample Characteristics.........................................................176 7.3 Activity-Travel Engagement Patterns of Adult Household Members........180 7.4 Model Specification....................................................................................183 7.4.1 Endogenous Variables....................................................................183 7.4.2 Exogenous Variables......................................................................184 7.4.3 Structural Equations Modeling Methodology.................................184 7.4.4 Postulated ActivityTravel Causal Structure..................................186 7.5 Model Estimation Results...........................................................................187 7.5.1 Intra-Person Activity-Travel Interaction.........................................190 7.5.2 Inter-Person Activity-Travel Interaction.........................................191 7.5.3 Intra-Person Activity Interaction....................................................192 7.5.4 Inter-Person Activity Interaction....................................................192 7.5.5 Intra-Person Travel Interaction.......................................................193 7.5.6 Inter-Person Travel Interaction.......................................................194 7.5.7 Effects of Demographic and Socio-Economic Attributes...............194 7.6 Summary and Discussion............................................................................199 CHAPTER 8 SUMMARY AND CONCLUSIONS........................................................201 CHAPTER 9 FUTURE RECOMMENDATIONS AND RESERACH GUIDELINES..........................................................................................211 9.1 Data Needs..................................................................................................211 9.2 Exploration of Temporal Dynamics in Travel Time Frontier in an International Context..................................................................................212 9.3 Modeling Positive Utility of Travel from the Utility Maximization Framework..................................................................................................212 REFERENCES................................................................................................................214 APPENDICES.................................................................................................................224 Appendix 1: Formulation of Producti on Frontier Model with Logarithmic Transformation of Dependent Variable..........................................225 ABOUT THE AUTHOR.......................................................................................End Page

PAGE 7

iv LIST OF TABLES Table 3.1 Household Characteristics.........................................................................32 Table 3.2 Person Characteristics................................................................................36 Table 3.3 Commuter Characteristics..........................................................................38 Table 3.4 Non-Commuter Characteristics.................................................................41 Table 3.5 Person Characteristics of Zero-Trip Makers..............................................43 Table 3.6 Trip Purpose Definition.............................................................................45 Table 3.7 Household Trip Produc tion Rate by Household Size and Household Car Ownership: India...............................................................46 Table 3.8 Household Trip Produc tion Rate by Household Size and Household Car Ownership: USA...............................................................47 Table 3.9 Person Trip Production Rate by Household Size and Household Car Ownership: India.................................................................................49 Table 3.10 Person Trip Production Rate by Household Size and Household Car Ownership: USA.................................................................................50 Table 3.11 Trip Production Rates of Pe rsons with Zero Income: India......................52 Table 3.12 Trip Production Rates of Persons with Low Income: India.......................53 Table 3.13 Trip Production Rates of Pe rsons with Low Household Income: USA............................................................................................................54 Table 3.14 Trip Production Rates of Pe rsons with Medium Income: India................56 Table 3.15 Trip Production Rate of Pers ons with Medium Household Income: USA............................................................................................................57 Table 3.16 Trip Production Rate of Persons with High Income: India........................59 Table 3.17 Trip Production Rate of Pe rsons with High Household Income: USA............................................................................................................60 Table 3.18 Household Trip Producti on Rates by Household Size and Household Two-Wheeler Ownership: India..............................................62 Table 3.19 Person Trip Production Rate s by Household Size and Household Two-Wheeler Ownership: India................................................................63

PAGE 8

v Table 3.20 Mode Definitions.......................................................................................71 Table 3.21 Modal Split Distribu tion of All Trips: India..............................................73 Table 3.22 Modal Split Distri bution of All Trips: USA..............................................74 Table 3.23 Modal Split Distribu tion of HBW Trips: India..........................................75 Table 3.24 Modal Split Distri bution of HBW Trips: USA..........................................76 Table 3.25 Modal Split Distribu tion of HBS Trips: India...........................................77 Table 3.26 Modal Split Distri bution of HBS Trips: USA............................................78 Table 3.27 Modal Split Distributi on of HBSocRec Trips: India.................................79 Table 3.28 Modal Split Distributi on of HBSocRec Trips: USA..................................80 Table 3.29 Modal Split Distribu tion of HBO Trips: India...........................................81 Table 3.30 Modal Split Distri bution of HBO Trips: USA...........................................82 Table 3.31 Modal Split Distribu tion of NHBW Trips: India.......................................83 Table 3.32 Modal Split Distribu tion of NHBW Trips: USA.......................................83 Table 3.33 Modal Split Distribu tion of NHBO Trips: India........................................84 Table 3.34 Modal Split Distribu tion of NHBO Trips: USA........................................84 Table 3.35 Modal Split Distribution of Commuter Sample by Trip Purpose: India...........................................................................................................87 Table 3.36 Modal Split Distribution of Commuter Sample by Trip Purpose: USA............................................................................................................88 Table 3.37 Modal Split Distribution of Non-commuter Sample by Trip Purpose: India............................................................................................89 Table 3.38 Modal Split Distribution of Non-Commuter Sample by Trip Purpose: USA.............................................................................................90 Table 3.39 Modal Split Distribution of School-Age Children Sample (Age <16 years) by Trip Purpose: India.............................................................91 Table 3.40 Modal Split Distribution of School-Age Children Sample (Age <16 years) by Trip Purpose: USA..............................................................92 Table 3.41 Modal Split Dist ribution of Zero-Car H ousehold Members by Trip Purpose: India............................................................................................95 Table 3.42 Modal Split Distribution of Non-Zero-Car Household Members by Trip Purpose: India....................................................................................96 Table 3.43 Modal Split Dist ribution of Zero-Car H ousehold Members by Trip Purpose: USA.............................................................................................97 Table 3.44 Modal Split Distribution of One-Car Household Members by Trip Purpose: USA.............................................................................................98

PAGE 9

vi Table 3.45 Modal Split Distribution of Two-Car Household Members by Trip Purpose: USA.............................................................................................99 Table 3.46 Modal Split Distribution of Three or More Car Household Members by Trip Purpose: USA..............................................................100 Table 3.47 Activity and Time Use Characte ristics of Mobile Samples: India..........106 Table 3.48 Activity and Time Use Charac teristics of Mobile Samples: USA...........107 Table 3.49 Average Trip Lengt h Distribution by Purpose.........................................108 Table 3.50 Non-Work Activity Engagement Patterns by Commuters Around the World.................................................................................................111 Table 4.1 Household Characteris tics (US, Swiss, India).........................................120 Table 4.2 Person Characteristics of M obile Adults (US, Swiss, India)...................122 Table 4.3 Average Daily Travel Duration by Purpose of Mobile Adults................123 Table 4.4 Stochastic Frontier Models of Travel Time Frontier: USA.....................125 Table 4.5 Stochastic Frontier Models of Travel Time Frontier: Swiss....................126 Table 4.6 Stochastic Frontier Models of Travel Time Frontier: India.....................127 Table 4.7 International Comparison of Average Travel Time Expenditures and Average Estimated Travel Time Frontiers........................................131 Table 5.1 Comparison of Daily Travel Duration by Purpose of Mobile Adults......141 Table 5.2 Stochastic Frontier Models of Minimum Required Travel Time: USA..........................................................................................................142 Table 5.3 Stochastic Frontier Models of Minimum Required Travel Time: Swiss........................................................................................................144 Table 5.4 Stochastic Frontier Models of Minimum Required Travel Time: India.........................................................................................................146 Table 5.5 International Comparison of Average Travel Time Expenditures and Average Expected Minimum Re quired Travel Time Frontiers........153 Table 6.1 Person Characteristic s of Commuter Groups: USA.................................162 Table 6.2 Person Characteristic s of Commuter Groups: Swiss...............................163 Table 6.3 Person Characteristic s of Commuter Groups: India................................164 Table 6.4 Commute Time Choice Model: USA......................................................166 Table 6.5 Commute Time Choice Model: Swiss.....................................................169 Table 6.6 Commute Time C hoice Model: Thane, India..........................................170 Table 7.1 Demographic Characteristics of Two-Worker Households.....................177 Table 7.2 Person Characteristics of the Adult Household Members.......................178

PAGE 10

vii Table 7.3 Activity-Travel Patterns of the Adult Household Members....................182 Table 7.4 Structural Equations Mode l Estimation Results (Causal Effects between Endogenous Variables)..............................................................197 Table 7.5 Structural Equations Model Estimation Results (Causal Effects of Exogenous Variables on Endogenous Variables)....................................198

PAGE 11

viii LIST OF FIGURES Figure 1.1 Population Growths in India and US...........................................................3 Figure 1.2 Rising Motorcycle Ownership in India........................................................4 Figure 1.3 Rising Car Ownership in India....................................................................4 Figure 1.4 Annual Per-Capita Income in India.............................................................6 Figure 2.1 Simplified Representation of Time-Space Interaction Using the Prism Concept............................................................................................22 Figure 3.1 Trip Distribution by Pu rpose at Trip Origin: India....................................64 Figure 3.2 Trip Distribution by Pu rpose at Trip Origin: USA....................................65 Figure 3.3 Trip Distribution by Purpose at Trip Destination: India............................66 Figure 3.4 Trip Distribution by Pur pose at Trip Destination: USA............................66 Figure 3.5 Trip Distribution by Purpose: India...........................................................67 Figure 3.6 Trip Distribution by Purpose: USA...........................................................67 Figure 3.7 Time of Day Dist ribution by Purpose: India..............................................69 Figure 3.8 Time of Day Distribution by Purpose: USA..............................................69 Figure 3.9 Modal Split Distribu tion of All Trips: India..............................................73 Figure 3.10 Modal Split Distri bution of All Trips: USA..............................................74 Figure 3.11 Modal Split Distribu tion of HBW Trips: India..........................................75 Figure 3.12 Modal Split Distri bution of HBW Trips: USA..........................................76 Figure 3.13 Modal Split Distribu tion of HBS Trips: India...........................................77 Figure 3.14 Modal Split Distri bution of HBS Trips: USA............................................78 Figure 3.15 Modal Split Distributi on of HBSocRec Trips: India.................................79 Figure 3.16 Modal Split Distributi on of HBSocRec Trips: USA..................................80 Figure 3.17 Modal Split Distribu tion of HBO Trips: India...........................................81 Figure 3.18 Modal Split Distri bution of HBO Trips: USA...........................................82 Figure 3.19 Modal Split Distribution of Commuter Sample by Trip Purpose: India...........................................................................................................87

PAGE 12

ix Figure 3.20 Modal Split Distributions of Commuter Sample by Trip Purpose: USA............................................................................................................88 Figure 3.21 Modal Split Distribution of Non-Commuter Sample by Trip Purpose: India............................................................................................89 Figure 3.22 Modal Split Distributions of Non-Commuter Sample by Trip Purpose: USA.............................................................................................90 Figure 3.23 Modal Split Distribution of School-Age Children Sample (Age <16 years) by Trip Purpose: India.............................................................91 Figure 3.24 Modal Split Distribution of School-Age Children Sample (Age <16 years) by Trip Purpose: USA..............................................................92 Figure 3.25 Modal Split Dist ribution of Zero-Car H ousehold Members by Trip Purpose: India............................................................................................95 Figure 3.26 Modal Split Distribution of Non-Zero-Car Household Members by Trip Purpose: India....................................................................................96 Figure 3.27 Modal Split Distributions of Zero-Car Household Members by Trip Purpose: USA.....................................................................................97 Figure 3.28 Modal Split Distributions of One-Car Household Members by Trip Purpose: USA.............................................................................................98 Figure 3.29 Modal Split Distributions of Two-Car Household Members by Trip Purpose: USA.............................................................................................99 Figure 3.30 Modal Split Distributions of Three or More Car Household Members by Trip Purpose: USA..............................................................100 Figure 3.31 Trip Length Distri bution by Purpose: India.............................................109 Figure 3.32 Trip Length Distri bution of Purpose: USA..............................................109 Figure 3.33 Demonstration of Trips, Tours and Trip Chain.......................................110 Figure 4.1 Distribution of Travel Time Expenditures and Estimated Frontiers: USA Mobile Commuters and Non-Commuters.......................................128 Figure 4.2 Distribution of Travel Time Expenditures and Estimated Frontiers: Swiss Mobile Commuter s and Non-Commuters.....................................129 Figure 4.3 Distribution of Travel Time Expenditures and Estimated Frontiers: India Mobile Commuter s and Non-Commuters......................................130 Figure 5.1 Distributions of Travel Ti me Expenditures and Expected Minimum Required Travel Time Frontiers: USA Mobile Commuters and Non-Commuters.......................................................................................149 Figure 5.2 Distributions of Travel Ti me Expenditures and Expected Minimum Required Travel Time Frontiers: Swiss Mobile Commuters and Non-Commuters.......................................................................................150

PAGE 13

x Figure 5.3 Distributions of Travel Ti me Expenditures and Expected Minimum Required Travel Time Frontiers: India Mobile Commuters and Non-Commuters.......................................................................................151 Figure 7.1 Postulated ActivityTravel Causal Structure...........................................188

PAGE 14

xi UNDERSTANDING ACTIVITY ENGAGEMENT AND TIME USE PATTERNS IN A DEVELOPING COUNTRY CONTEXT Amlan Banerjee ABSTRACT Flourishing economy, rapid industr ialization and increasing tr end of motorization have been shaping societies in th e developing countries like Indi a in an unprecedented manner. Infrastructure backlog amid such rapid growth in all imaginable directions has heavily exacerbated the urban transport crisis in thes e countries by alarming increase in vehicular travel demand, road fatalities, and environmental pollution. To address urban transport challenges, the necessary development and implementation of effective transport planning and policies have genera lly lagged in the developing countries compared to that seen in the developed countries due to seve ral constraints including resource constraints, knowledge constraints, institutional constraint s and so on. However, in the recent past, with the rapid development seen by seve ral emerging economies and the explosive growth in transportation infrastructure inve stment, there is a grow ing interest in the development and implementation of advanced travel demand modeling systems in developing countries. But lack of necessary research and explor ation of travel behavior in a developing country context has left very limited knowledge for us to understand the extent of applicability of th ese advanced theories and met hodologies in a different sociocultural perspective. Assessing th e practical relevance of the s ubject, this research adopts a comprehensive approach to explore the activity engagement pattern and time use behavior from a developing country standpoi nt. To accomplish this goal, a series of empirical and analytical studi es are performed on a househol d travel survey data set available from Thane Metropolitan Area in Indi a. The study also introduces new concepts

PAGE 15

xii and facilitates enhancements of existing m odeling methodologies in the field of travel behavior and time use research. The study resu lts provide very insi ghtful findings and plausible interpretations consistent with a developing country pe rspective recognizing a wide spectrum of differences and similarities in activity patterns and time use behavior between a developed and a developing c ountry. Specified model structures are meaningfully able to incorporate various so cio-cultural and institutional constraints and reflected sensitivity to the behavioral variab ility between the contexts suggesting that advanced analytical techniques may be sa tisfactorily applied on the data set from developing countries which may contribute im portant ingredients in the development of advanced activity-based model system in the countries like India.

PAGE 16

1 CHAPTER 1 INTRODUCTION 1.1 Urban Transport Trends and Challenges in Developing Countries Most developing nations around the world sh are common challenges associated with operation and management of their overstresse d transportation infr astructures. Rapidly increasing motorization in the mega cities and growing population pressure are creating huge infrastructure backlog, escalating envi ronmental problems, imposing hindrance on economic developments, and serious implicati ons on energy shortage and global climate change. The influences of rapid urbanization, changing lifestyle of the people and institutional constraints have al so been revealing to be significantly critical in the context of transportation challenges in the developing world. As it can be imagined, the nature of tr ansportation challenges and its impacts on societies in the developing countries are ve ry different than that observed in the developed world. Countries like China and Indi a are perhaps setting th e best examples of how flourishing economy is transforming the so cieties of the devel oping nations in an unprecedented pace. Transportation crisis of many major cities of these countries are heavily exacerbated by rapid economic growth amid low per-capita income, rising motorization, inferior transportation infras tructure, uncoordinated land use pattern and primitive transit services. Particularly, the imbalance between skyrocketing motor vehicle ownership and poor transportation infrastructure are leading to alarming increases in road fatalities and injuries, traffic congestion, en vironmental pollution, loss of productivity and high rate of energy consumption. If th ese current trends continue, increasing contributions of the developing countries to greenhouse gas emission and energy use will far offset the modest reductions achieved by the developed world (Pucher et al., 2005). Even though the developing nations are substantially diverse among themselves in terms

PAGE 17

2 these nations have several issues in common that contribute to th e severity of their transport problems (Pucher et al., 2005). India has been chosen in th e context of this present st udy considering the fact that Indian cities share a range of critical issues that the other developing countries are facing related to current trends in urban development and urban trav el patterns. Some of these common issues are discussed further in the Indi an context to provide a sense of the extent and nature of the problems the country is experiencing. 1.1.1 Urbanization Constant population growth with increasing trend of urbanization and economic growth are the most important factors common in India and other deve loping countries. Along with relentless populat ion growth at the national level, the total urban population of India has grown over the past three decades, ri sing from 109 million in 1971 to 160 million in 1981 (+47 percent), 217 million in 1991 (+36 percent), and 285 million in 2001 (+31 percent) (Office of the regist rar General of India, 2001; Pucher et al., 2005). The three mega polis Mumbai (Bombay), Kolkata (Calcu tta) and Delhi are at the leading position with 16.2 million, 13.2 million and 12.8 million inhabitants respectively. Figure 1.1 shows a trend of much greater rate of populat ion growth compared to US with about 1.1 billion people recorded in 2005 in India (Office of the register General of India, 2001). As a result of such enormous population pressure in the Indian cities, space scarcity and lack of land-use planning and controls, rampant sprawled developments have expanded over the years in all directions, far beyond old city boundaries into distant suburban fringe (Pucher et al. 2005). Governme nt policies have actually encouraged such suburban settlements in India in order to decongest city centers. However, unplanned and scattered commercial and residential settle ments in outlying areas not only lacked necessary infrastructural amenities, but also created dire consequences for existing transportation services as well. Many of such suburban sprawls are taken place along some major circumferential highways and have increasingly cr eated transportation problems like congestion, degradation of transit services and so on. It has further led to

PAGE 18

increased commuting length and time, discourage public transportation use and induce desire in people to own automobiles. 3 010020030040050060019891990199119921993199419951996199719981999200020012002200320042005YearPopulation (in million) 9001200 1100 1000 700800 India US Figure 1.1 Population GrowSource: Office of the Registrar Ge ths in India and US neral of India, 2004; United States Census Bureau g societies The growths in motorized two-wheelers and passenger cars in India over the last decade are shown in Figure 1.2 and Figure 1.3. Figure 1.2 shows ownership of two-wheelers in India has maintained continued increase at a constant rate of 2 vehicles per 1,000 persons per year over the last fifteen years (1987 to 2002). Two-wheelers in India are extremely attractive to middle class people as it is affordable and provides 1.1.2 Motorization Motorization is transforming cities and as well as rural areas of the developing world in an unprecedented rate. Resembling western societies, automobile has been becoming a symbol of modern urban culture and obsession of the people in the developinas owning a car renders mobility, status and social freedom. Over the last few years, the most dramatic development in the transportation sector in a developing country like Indiais the striking growth of private automobiles, especially car and low-cost scooters and motorcycles. India has become the worlds fastest-growing car markets, with about a million cars being sold in each year.

PAGE 19

4convenience and flexibility. Although the growth of two-wheeler fleet is considered to be the major contributor in the process of motorization, the recent growth in car ownership has gained significant attention as well. Figure 1.3 suggests that the number of passenger cars per 1,000 people in India is doubled between 1991 and 2002 indicating the trend of continuous growth in the subsequent years. 04 8 of 12 Moto 16202428323640441976197819801982198419861990199219941996199820002002Numberrcycles per 1,000 People 1988 Year Figure 1.2 Rising Motorcycle Ownership in India Source: Ministry of Road Transport and Highways, India, 2003 012345678199119921993199419951996199719981999200020012002YearNumber of Passenger Cars per 1,000 People Figure 1.3 Rising Car Ownership in India Source: Ministry of Road Transport and Highways, India, 2003

PAGE 20

5 d gnificant as it expands job and educational opportunities and boa sts rural economy by s. Be sides these desired benefits, the negative he d e, many Indian mega-cities including Bangalore and Mumbai, have been xperiencing extreme challenges in investment and business expansion in and around the city in recent years due to poor urban pla nning coupled with rapid growth of motorized transport that has swamped the capacity of the transportation infrastructure of the city. Many of such cities in India and also other developing count ries, with a fraction of car ownership of the United States, now experi ence the worst traffic congestion, pollution and road accidents than exist in the United States (Sperling and Claussen, 2004). 1.1.3 Economic Growth and Migration Another distinguishing transformation happening in the major Indian cities is rapid industrialization and subsequent economic prosperity. Although per-capita income in India is very low compared to developed countries like United States (In 2002, US: $30,906 vs. India: $2,700), still India shows remark able increase in per-capita income in recen e more than doubled in last twenty years from USD 1,200 pe r year in 1982 to USD 2,700 per year in 2002. Motorization in cities has great benefits and highly valued, especially when mobility is closely linked with economic deve lopment (Ng and Scipper, 2005). It is also seen to be a means of improvi ng quality of life by bringing more freedom, flexibility an social status in ones life. The economic and social benefits of motorization are also si providing improved access to rural market consequences of motorization have started becoming prominently visible as well. In addition to the fact that motorization in th e countries like India a nd China, which bear more than one-third of worlds population toge ther will be the largest consumers of t worlds fossil fuel supplies in near future increased congestion, pollution, safety hazards and in many cases, detrimental effects on economic development and society have starte revealing as the acute rami fications of motorization in the developing world. For instanc e t years. Figure 1.4 suggests that per-capit a income in India has becom

PAGE 21

250Pecome 175020002250250027503000 (USD) 500750100012501500r-capita In 0 1971197319751977197919811983198519871989199119931995199719992001Year Figure 1.4 Annual Per-Capita Income in India Source: Organization for Economic Cooperation and Development (OECD) Growing economy has created new job opportunities and as a result of that the cities attract cheap labor force from rural areas in India like magnet. The newly built highways are carrying thousands of people in and out of the cities everyday. Allured by the thriving lifestyle of metropolis and potential access to employment, poor people are migrating to cities abandoning their rural life behind. These rural migrants often compromise to make their livelihood on meager earnings and strained resources, which offer the m a grim standard of living. ciety respond to these changes equally. Economic successes have further extended the distance Informal and unorganized settlement is a common problem in any major Indian city and growing travel demand generated by this elusive migrant population imposes additional burden for the over-stressed transportation infrastructure. 1.1.4 Social Inequality Capitalism and globalization have convulsed India in an unprecedented rate of change. India is already one of the fastest growing economies and most rapidly evolving soin the world (Waldman, 2005). Shining highways, foreign cars are all the manifestation of the radical changes that has been reshaping Indian society. But the society does not 6

PAGE 22

7 cent from 36 percent, yet th e poor seem poorer than ever (Ministry of inance, India, 2002; Waldman, 2005). The problem of overall low per-capita income is income dist ribution in the population. The wealthiest os, d growth in moto r vehicles have benefited people who can afford p t ravel Behavior Econom t ety r between rich and poor. Despite the fact that since 1991, India population of poor has dropped to 26 per F compounded by extremely skewed tenth of the population typically earns over half of total na tional income (Vasconcell 2001). Rapid motorization has led to inequalities in transport mobility and accessibility in India. New highways an rivate cars and the disadvantaged poor and middle class suffer from the overcrowded public transport and worseni ng transportation problems in cities. The concentration of wealth among economic and political elite has distorted transpor policies alike in all de veloping countries (Pucher et al. 2005). Despite the fact that nonmotorized mode share (50 per cent) is still the highest in Indian cities, government policies and investments are heavily focused on serving the needs of auto-owners, while the traveling conditions of pedestrians and bi cyclists have significan tly deteriorated by the increasing number road accidents and pollution caused by motor vehicles over the past few years. 1.1.5 Trends in T ic growth and technological advancem ents are reshaping peoples lifestyle, choices and preferences in every direction of their life. Occurrence of simultaneous changes in consumer behavior and their attitudinal preferen ces are obvious in the contex of transportation as well. Consumer preferences vary across society and culture. In general, Indian soci is characterized by traditional gender roles, where labor pa rticipation rate among female is significantly lower than that of male and females are primarily responsible fo undertaking the major share of household oblig ations; family ties are very strong with average household size 4.5; and auto-ownersh ip and household income level are low. Traditionally, these constraints have defined tr avel patterns and preferences of the Indian people. Due to unavailability of private vehi cles, people are appeared to be captive to

PAGE 23

8 of low di sposable income coupled with lack of opportunities; and out-of-home activity particip ation has been typically very limited to ehold ch nding e transport hcost so ore re using non-motorized or public transportation; engagement in discretionary recreational activities are very low because among females. During the past decades, economic and t echnological growth has contributed the relaxation of many constrai nts in peoples life. Trend to ward adopting some of the ways of the West, particularly among young ge neration; substantial shifts in hous structure especially in urban areas; trend of higher educa tional attainment among female and eventually joining labor force; increas ing per-capita income and sense of social recognition attached with owning a car; more and more availability of recreational facilities are all collectively leading to ch ange in peoples atti tude and activity engagement patterns. However, virtually no kno wledge is available so far about how su dynamic transformation in the Indian society could bring consecutive changes in travel behavior patterns. Therefore, more attention is needed to be paid to understand changing consumer behavior and preferences in regard to transportation systems. Understa local factors influencing consum er practices will be keys to direct future transportation policies and planning options. 1.2 Problem Statement In general, the development and implemen tation of effective pl anning and transport policies have been lagged in the developing coun tries like India to deal with th problems that they are experiencing. The cost of infrastructure development and deployment of advanced transportation technolog ies are not often viab le for the countries with constrained financial resources and lack of institutional flexibility. Therefore, hig lutions in the developing countries should be avoided and need to develop accessible land-use patterns and implement smart growth practices that encourage efficient planning. Consideration of the infl uence of local conditions, needs and m emphasis on consumer behavior are needed to be included in the transportation planning practices. Significant efforts should be mobilized in the developing countries to explo the feasibility of the development and implementation of transportation demand

PAGE 24

9 e r el ior ta constraints, and inst itutional constraints. In the recent past, with the rapid development s een by several emerging economies and the xplosive growth in tran sportation infrastructure investment there is a growing interest in f advanced travel demand modeling systems in in very come management strategies (TDM) and transportation control measures (TCM) that promot variety of mobility management practices l eading to reduced level of congestion and ai pollution by reducing car usage. In most deve loped countries, the focus of transportation planning has shifted away from capacity expa nsion to that of operation, management, and optimization of existing capacity. This shift in planning emphasis has motivated trav behavior researchers to be concerned with relationsh ips and trade-offs among individuals time expenditures, travel, and activities. It is envisioned that travel behav models based on an understanding of peoples time use patterns offer a robust framework for analyzing the impacts of alternative tr ansportation policies and control measures In developing countries like India, the development and implementation of advanced travel demand modeling systems has generally lagged that seen in developed countries. There are several reasons for th is, including resource constraints, staff constraints, knowledge constraint s, da e the development and implementation o developing countries. But it is still unknown about th e transferability of these theories and methodologies that have been developed in the devel oped world to the developing world. The time is ripe to explore the feas ibility of applying these methods different socio-cultural and transportation system contexts. Recent development of activity-based modeling approach has increasingly focused on the analysis of activitytravel patterns and time use behavior of i ndividual, primarily because traditional tripbased models are inadequate to incorporate underlying behavioral patterns of individuals and the interaction among their activities and trip making patterns, which is highly essential to evaluate and analyze travel demand management strategies. Therefore, understanding individual travel behavior from the activity-b ased perspective has be highly critical in the development of futu re demand management strategies in the developing country context.

PAGE 25

10 f t and implem deling ural f onstraints that people encounter in making their daily activity-travel decisions. d ountries, this research offers rich comparisons of activit y patterns and time use behavr Evide ext chapter), it is r can be analyz g the scope developing c ness of the wholes in nature and ing the model de ies, this research me use and activity ngagement patterns and discuss about their possible implications on future 1.3 Research Objectives and Scope Considering the growing need of improving tr ansportation facilities and development o effective planning strategies in the developing countries, there is a growing need to adop new paradigm of advanced demand forecasting models. Development entation of activity-based model sy stems necessitates be tter understanding of individual activity-travel patt erns and time use behavior. Mo tivated by the relevance of the subject in the current context, the focus of this research is to present a comprehensive analysis of activity engagement and time use patterns in a developing country perspective based on the household travel survey data se t from Thane in India in 2001. The study includes wide range of exploratory analyses between various market segments stratified by their socio-economic and demographic char acteristics to capture the behavioral heterogeneity that may exist in the populat ion. Separate models of activity-travel engagement and time use behavior are also estimated applying econometric mo methodologies for various socio-economic groups accounting poten tial struct differences in their behavior and decision ma king. The models incorporate the effects o various c Considering the contextual differences that are likely to exist between developing an developed c io across diverse sociocultural and geographical contexts. nce from the rich body of travel behavi or literature (dis cussed in the n conceivable that numerous aspects of activity-travel engagement behavio ed with respect to time use beha vior of individual traveler, but considerin ofthis research, some key areas that are perceived to be extremely relevant in a ountry context will be focused still maintaining the comprehensive reearch effort. Most of the analyses conducted in this rese arch are empirical d avanced econometric modeling methodologies are applied in accomplish velopment process. In additi on to introduce new modeling methodolog exhaustively explores be havioral aspects of ti e

PAGE 26

11 s in a of commuting time choice beha vior in the international context Analysis of activity-travel interactions among household adults and its effects on inter-person a nd within-person activity-travel engagements decisions in a developing country context pter uting nd ssed in the e transportation policy analysis and planning pr ocesses in a very different socio-cultural context. The main objectives of this research can be broadly classified into the following categories: Exploratory analysis of ac tivity-travel engagement and time use pattern developing country context and si multaneous comparison between a developing and a developed country to explore the effects of contextual differences in individual activity-travel engagement decision making Understanding travel time expenditu re around the world by introducing the notion of travel time frontier and minimum travel time threshold Analysis 1.4 Outline of the Dissertation The state-of-the-art practices in travel beha vior research and the recent developments in activity-based modeling approaches are discusse d in the next chapter. The third cha offers a comprehensive exploratory analys is of demographic, socio-economic and activity-travel characteristics in a developing country context. Expl oration of the notion of travel time frontiers and minimum trav el time threshold around the world will be presented in the forth and fifth chapter resp ectively. A detailed analysis of comm length choice behavior in an international co ntext will be offered in the sixth chapter a an analysis of activity-travel interaction among household adu lts will be discu seventh chapter. A concluding discussion and research summary will be included in th eighth chapter and finally, the ninth chapter will offer recommendations and directions for future research.

PAGE 27

12 .5 Summary Recent economic growth and trend in motori tion and urbanization in the developing countries like China and India have resulted transportation crisis by alarming increase in congestion, fatalities and environmental problems. In general, implementation and development of effective planning and transp rt policies have been generally lagged in the developing countries. Constraints on the availa bility of financial resources to maintain and expand the existing infrastructure and c oncerns about the environmental impacts of transportation investments have led to th e consideration of alternative mobility management options such as travel dema nd management and smart growth practices. However, simultaneous developments of advanced modeling techniques and understanding of individual activity and tim e use patterns from a developing country standpoint are also necessary to evaluate the complex nature of user responses to such planning strategies. Understanding the growing need in the development and implementation of advanced travel demand modeling systems in developing countries, this research adopts a comprehensive appro ach to explore the activity engagement and time use behavior in a developing country perspective. 1 za o

PAGE 28

13 CHAPTER 2 OF THE ART IN TR ANSPORTATION MODELING ntry cal vels of traffic congestion, vehicul and and e to evaluate alternative transportation infrastructure provisions, which are mainly shifted STATE The previous chapter has introduced the tr ansportation challenges that the developing countries are typically e xperiencing due to overwhelm ing growth of population, economic activities, motorization and urbanization. The chapter has also addressed the fact that the development and implementation of effective transport policies and planning options to tackle such enorm ous problems have generally lagged in the developing world that seen in the developed world. Having th e research problem stated, the need for advanced behavior-based modeling approaches is acknowledged in a developing cou context and then research objectives a nd scopes are outlined. The current chapter discusses about the recent deve lopments in activity and tine use studies and state-of-theart approaches in travel behavior modeling. The emerging trends in travel behavior in India and similar developing countries are identified and scope of applying advanced analytical techniques in modeling travel behavior is also discussed. 2.1 Introduction The need for efficient transportation and land use system has never been more criti than it is today due to increasing concer ns over high le ar emissions, the sustainability of growth patterns and travel, and the related adverse impacts on regional and nationa l productivity (Bhat and Lawton, 2000). The purpose of travel demand models lies in its ability to forecast the transportation dem both from the aspects of users and transportation system attri butes that aid planners policy makers to efficiently operate the tran sportation infrastructu re. Constraints on th availability of financial resources to main tain and expand the existing infrastructure and concerns about the environmental impacts of transportation invest ments have led

PAGE 29

14 due s g the variation of tast e and choices in the population, but also has the potential to identify ch is om the eir heref ore, the studies of activity engagement and time use patterns en able us to enrich our understanding of the ior and that wi ll lead to increased 9). As activ ity-travel engagement behavior including (1) patial and temporal constraints on activity and travel choice, (2) scheduling and d space, (3) interactions between activity and travel towards mobilizing demand management policie s. The complex nature of responses to such demand management strategies and rapid changes in activity and travel patterns to substantial shift in household structure a nd socio-demographic characteristics in the population have led to the need of incorporating realisti c representation of decisionmaking behavior in the travel demand mode ls. Behavioral framework not only allow capturin the differential quality of transpor tation services associated with different segments in the population (Bhat and Lawton, 2000). The conceptual development of activity-b ased approach of analyzing travel behavior is virtually resulted from the limitati ons of tradition trip-based models, whi based on simplistic assumptions that are unable to examine the complex behavioral responses of the demand management actions. In the past decades, the attention of transportation community has largely shifted towards the paradigm of activity-based modeling approaches, which are based on the id ea that travel demand is derived fr necessity to participate in spatially sepa rated activities. Give n that, the approach explicitly recognizes that indi viduals activity-travel patterns are a manifestation of th decision to allocate time to various activities during a day. T complexity and variability of individual travel behav capability of forecasting travel demand and evaluating planning options (Pendyala 2003). Time use study has gained significant importance in activity and travel behavior research because of recognizing the fact that individuals activity-travel patterns are a result of their time use decisions within a continuous time domain (B hat et al. 199 time is a finite resource, an individual al locates time to various activities during the course of a day according to his/her needs and preferences. Such exp licit recognition of underlying complexities in i ndividual time allocation behavior has allowed time use studies to determine many aspects of s sequencing of activities in time an

PAGE 30

15 ment ivity categor man, ries s based on household/ individual characteristics. In these studies, the activity episode ain. gsaeter (1983) and Juster (1985). decisions, and interactions be tween individuals, and (4) ro le played by members of a household in accomplishing household activities and tasks. 2.2 Review of Activity-Based Time Use Research Many past and recent studies on travel beha vior research have demonstrated and conceptualized individual activity scheduling and travel pattern as a dynamic, complex and adaptive system. The research effort in this area is very dive rse and the develop of activity and time use studies can be classified into two br oad categories: 1) act time allocation studies; 2) activity episode analysis. Studies carried out under the y of activity episode an alysis can be further classi fied into following groups of analyses: activity episode duration, activity sequencing, ac tivity timing and scheduling, activity episode genera tion and scheduling and activity frequency (Bhat and Koppel 1999; Pendyala and Goulias, 2002). 2.2.1 Activity Time Allocation The activity time allocation studies have prim arily focused on the analysis of daily time use and allocation behavior of a household or an individual to various activity catego or purpose s are not considered. In other words, activity time allocation studies do not focus on the contextual aspects of activities, e.g. time-of-day of activities, activity scheduling and sequencing in the continuous temporal domain, location of the activities, frequency and duration of activities, and the decision of joint activity participation with other members. However, these studies can cap ture the time allocation and trade-offs associated with time allocation to various activities types in a continuous time dom The evidence of empirical studies of activ ity time allocation was first seen in time use literatures back in 1970s. Some cross-sectional and longitudinal cross-country studies of time allocation patterns of individuals were conducte d by Szalai (1972), Harvey (1986), Lingsom & Ellin Chaplin (1974) studied the effects of stage in lifecycle, race and status on household time allocation behavior on weekdays and weekends. Later in the

PAGE 31

16 on the literature on individual time allocation behavior. In the udy conducted by Kitamura and Fujii (1996) on individual activity participation and me discretionary activities indicates that me een inome and out-of-home. A separate array of research efforts (Golob and McNally, 1995; Koppelman and ownsend, 1987; Stopher, 1995; l u, 1997) is carried out wi th the development of ch to an alyze the interactions in time allocated to transportation literature, Jones et al. (1983) and Kostyniuk and Kitamura (1986) studied and confirmed the significant effect of lif ecycle stages on household time allocati behavior. Besides household time allocation behavi or, Kitamura and Fujii (1996), Bhat and Misra (1998) largely enri ched st time allocation decision to in-home and out-of-ho being employed and having long work commute have negative effect on allocation ti on out-of-home discretionary activities. Again, the same tendency found for older individuals and individuals be longing to a large household si ze to allocate time to out-ofhome discretionary act ivities. Furthermore, Bhat and Mi sra (1998) modeled discretionary time allocation behavior between weekdays and weekends in addi tion to betw h T structural equations modeling approa out-of-home activities and travel mainly in three activity categories: 1) work; 2) maintenance and 3) discretionary between members in a household. These studies explicitly account for the inte r-individual interaction effects in time allocation behavior among household members. 2.2.2 Activity Episode Analysis The term activity episode is referred to discrete activity partic ipation. Therefore, a number of episodes of the same type or purpose over some time unit (e.g. a day or a week) collectively termed as an activity (Bhat and Koppelman, 1999). Activity episode analysis incorporates a series of analyses th at emphasize on the contextual attributes of an activity such as activity time, sequenci ng, scheduling, joint pa rticipation etc.

PAGE 32

17 2.2.2.1 e the y al. and gs a large househ y d on th e same modeling methodology. The results of the stud Activity Episode Duration Past studies on activity episode duration have primarily dealt with modeling individual activity episodes duratio ns by purpose or category. Episode duration models have been traditionally based on hazard-based duration models and Tobit models that explain th context of single activity episode. Hazard-base d duration model in this context provides a powerful tool in understanding the factors th at influence individual activity episode durations and the probability that a certain ac tivity will be terminated given that a certain duration that elapsed. Under such modeling fr amework it is possible to incorporate the effect of interdependence among activity ep isodes and the timing of activities under consideration that the end of one activity ep isode reflects the beginning of the subsequent activity episode in the continuous time domain. However, is sues associated with activity scheduling, sequencing and time allocation are difficult to capture in episode duration modeling series and hence those are tack led applying different modeling methodolog (Bhat, 2000; Pendyala and Goulias, 2002). In early 90s, Mannering and his colleague s (Mannering et al. 1992; Kim et 1993) studied activity episode duration between successive participation in in-home out-ofhome activity episodes applying Cox proportional hazard model. Their findin suggested that older individua ls, unemployed and an indivi dual belonging to old are likely to have longer home-stay episode compared to others. A later stud by Neimeier and Morita (1996) was accomplishe d to analyze the duration of out-of-home activity episodes associated with maintena nce-related shopping, pe rsonal business, and free time activities for workers base y suggested that males and females appear to have same amount of activity episodes for their personal bus iness and free time activity episodes, but females are more likely to have longer maintenance-related s hopping episode than me n particularly when such activities are pursued duri ng the return to home from work. In the same context, Bhat (1996) developed a non-parametric ba seline hazard model of shopping activity episodes during the evening commute. Unlike Cox proportional hazard model, Bhats model incorporates unobserved heterogene ity in durations using a nonparametric distribution. The covariance structure in the model is capable of accommodating an

PAGE 33

18 travel time duration to the activity episode. Duration is modeled sing a linear regression structure. ics of i ndividual behavior, the necessity of the t e mmodates the variatio n ticipation .2.2.2 Activity Episode Sequencing e th e sequence in which the activity episodes good b (2000) developed a simultaneous model of household activity partic ipation and trip chain generation. individuals work schedule characteristics, the work durat ion characteristics of that individuals spouse, travel mode to work and socio-demographic attributes. Notably in contrast to the duration models, Hamed and Mannering (1993) and Bhat (2001) introduced discrete-continuous fram ework in analyzing activity episode type, episode duration, and u Considering the temporal dynam analysis of the multi-day data for better policy actions and understanding activity-travel pattern has been realiz ed long time ago. Over the last tw o decades, number of studies has contributed in understanding multiday activity-travel behavior either by examining the extent of interpersonal and intr apersonal variations or day-to-d ay variations in the contex of daily work activities (Bhat, 2001; Hans on and Huff, 1988; Hirsh et al., 1986; Ma and Goulias, 1997; Pas and Koppelman, 1987; Pas and Sunder, 1995). Recently, Bhat and Srinivasan (2005) developed a unifying mu ltivariate hazard model to examine th participation of individuals and their dependence in partic ipating in various non-work activities over a multiweek period. The fl exible model structure acco n in interepisode duration due to unobserved individualspecific factors, variatio within the spells of the same individual and also considers the joint nature of par in various activities. 2 The studies of activity sequencing examin are linked in a continuous time domain. The st udies of trip chaining patterns are the examples that belong to this category because trip chaining is simp ly a manifestation of activity sequencing decisions (Pendyala and G oulias, 2002). In late 70s, Adler and BenAkiva (1979) developed a theoretical and empi rical model of trip chaining behavior and later Kitamura (1984) incorporated the concept of trip chaining in the development of destination choice model. Also, Timm erman et al. (1992) conducted a study on pedestrian trip chaining behavior and r ecently Golo

PAGE 34

19 e of ing and Scheduling ctivity episode timing and scheduling mode ls focus on identifying when a certain y of literatures, hazard-based duration niques pt of joint model structure of activ ity episode generation and scheduling is the ost recent development in activity time use research. Two approaches have been t of activity episode genera tion and scheduling modeling y S) TGW) e generati on and allocation of From the evidence of literature, the st udy of activity episode sequencing can b further distinguished into two categorie s. The first category focuses on episode scheduling and the second type concerns a bout the development of the joint model activity episode sche duling and generation. Activity Episode Tim A activity or trip will be pursued. In the current bod models, time-of-day period-based discrete c hoice models, and heuristic algorithms have been adopted to analyze activ ity episode scheduling behavior. These modeling tech do not capture time use decisions in particular ; rather they examine the role of time in activity-travel behavior by accounting the fact that activity sc heduling behavior itself is temporal in nature (Pen dyala and Goulias, 2002). Most of the activity epis ode scheduling models are computerized production system that attempt to capture the decisionmaking process of individuals. These models are capable of incorporati ng effects of interaction among household members and individual activity travel choices. SCHE DULER (Garling et al., 1989), SMASH (Ettema et al., 1993), AMOS (Kitamura et al., 1996) are the examples of some existing activity scheduling models. Activity Episode Generation and Scheduling The conce m undertaken to the developmen system within the context of continuous time domain. The first approach proposed b Kitamura and Fujji (1996) is the Prism-Constrained Activity-Travel Simulator (PCAT and the second is the Comprehensive Activit y-Travel Generation for Workers (CA model systems proposed by Bhat and Singh (2000) Bhat (1999), and Srinivasan and Bhat (2004) also examined the effect of househol d interaction (conside ring both substitution and companionship effects) and resource cons traints on th

PAGE 35

20 shoppin is greater ing .2.2.3 Activity Frequency he literature have mainly dealt with the frequency of ive the time f ible in a given day due to limited tim e availability and limited speed of travel on the tran g activities in the hous ehold. They found significant imp acts of those factors in household shopping activity generation and allo cation behavior. The gender roles are significant in the allocation of shopping respons ibilities; solo shopping duration for females than males. However, joint shopping duration is longer than solo shopp duration. 2 Activity frequency studies in t occurrence of various types of activities. Count data mode ls, e.g. Poisson or negat binomial, discrete choice models and orde red response models are widely used in modeling activity frequency. Activity frequency models do not explicitly capture dimension; rather they are exclusively focu sed on the number of occurrences of various activities, regardless of the durat ions of the episodes. Lately, Bhat and his colleagues (Popur i and Bhat, 2003; Bhat and Srinivasan, 2005) have contributed some e fforts in activity frequency st udies. Popuri and Bhat (2003) have proposed a joint model of homebased telecommuting choice and weekly telecommuting frequency. Bhat and Srinivas an (2005) examined the frequency o participation of individuals in out-of-h ome non-work activities over the weekend by estimating a multivariate mixed ordered response logit model. 2.3 Time-Space Interaction A growing body of literature has explicitly recognized the role of space and time dimensions in shaping activity and travel pa tterns of individuals (Bhat and Koppelman, 1999; Pendyala, 2003). The role of time in travel behavior modeling has become extremely important by the fact that time is a finite resource to every individual and is consumed in undertaking travel and various sp atially separated activit ies. Therefore, an individual can pursue only a finite number of activities by traveling to a set of poss destinations sportation network. Then, it is quite implicit that the spatial dimension is very closely related to the temporal dimension as the distance traveled and the set of

PAGE 36

21 e idely taken ced o roaches ed int A ual at work in compliance with work schedu thout by the ore, increase in speed of travel would allow the time space prism to be increased in size and the individual can ndertake more activities, spend more time at the same activities, or visit destinations rther away. On the other hand, if speed of travel were to d ecrease (say, due to creased congestion), then the prism shrinks and the individual is more constrained with spect to activity engagement and locations that can be visited. Thus, the time-space rism concept provides a framework for anal yzing the induced (or suppressed) travel ffects of capacity incr eases (or decreases). destination visited are constr ained by scheduling and time av ailability. Such constraints define a spatio-temporal action space for ever y individual within which he/she can pursu daily activities. The temporal and spatial aspects of these activities tend to impose constraints on an individuals daily activity-travel engagement pattern. It is now w recognized that human activity and travel patterns may be c onsidered as being under within time-space prisms, which represent spat io-temporal constraints that are influen by socio-economic, demographic and travel characteristics (Hgerstrand, 1970; Kond and Kitamura, 1987; Miller, 1991). Accurate representation of time-space prisms has gained added importance in the context of th e emergence of microsimulation app of travel demand forecasting, where activity -travel patterns of each individual are simulated. A simplified representation of a typical ti me-space prism of an individual deriv from Pendyala and Bhat (2004) is presented here. According to the Figure 2.1, po represents the earliest possible time point for an individual to leave home (to go to work), possibly due to the need to ta ke care of household obligations and/or the desire to sleep until a certain time prior to starting the work day. Similarly, the time point B represents the latest possible arrival time for an indi vid les. The prism shown in the figure thus, is the representation of a time-space continuum in which an individual can undert ake various activities and travel wi violating the time constraints. The spatial boundaries (or constraints) that dictate the range of destinations (activity locations) th at the individual can vi sit are governed speed of travel, v, in the figure. This value, in turn, is directly dependent on the transportation system characteristics (level of service). Theref u fa in re p e

PAGE 37

22Many recent studies in this area have cused on the modeling and representation of time space prism vertices or boundaries to understand about how individual perceptions of 2000; Pndyalt al., 2004). Understanding time-space ave direct contribution to the development of models of ity to l in mong household members may occur in several ways. Household member may allocate d fo constraints influence his/her activity-travel patterns (Kitamura, et al., e a, et al., 2002; Yamamoto, e interactions are considered to h dynamics of activity and travel decision making processes. Time B A Figure 2.1 Simplified Representation of Time-Space Interaction Using the Prism Concept 2.4 Inter-Agent Interaction Activity-based model systems are becoming increasingly sophisticated in their abilincorporate a variety of household interactions and constraints that influence individuaactivity-travel patterns. Over the past few years, significant advances have been madeunderstanding in the nature of household interactions and its role in explaining activity participation of and travel-activity decisions made by household members. Interactions Work Home Space a tasks among one another, make joint activities, and depend on one another for undertaking activities and travel (particularly in case of children who depend on adults for their transport). As interactions among household members are undoubtedly important determinants of individual activity-travel behavior, an understanding of such interactionsand task allocation behavior is critical to the development of activity-based travel deman

PAGE 38

23 among adult household memb ers with respect to activity and travel engage nt, there ds embracing comprehensive ctivity-based model systems to replace the traditional four-step trip-based method. approach has been already proving invaluable in evaluating based he s n modeling systems. Even though most mode ls incorporate hous ehold-level socioeconomic variables as explanatory factors, they may not be sufficient to explicitly account for the range of possible household inte ractions and task allocations that may take place. There have been several studies in the recent past aimed at exploring and modeling interactions ment (Golob and McNally, 19 97; Meka et al., 2001; Simma and Axhausen, 2001). 2.5 Emerging Trends in Travel Behavior in Developing Country Context Realizing growing complexity in travel behavi or and changing travel environme is an increased interest in many developi ng countries towar a Implementation of this new transport policies and forecasting travel de mand in the developed world. Activityapproach is rapidly gaining momentum in the demand modeling profession. However, t development and implementation of such advanced modeling methods has seriously lagged in the developing nations while these countries are experiencing the same change as developed countries in terms of people s life style, travel behavior, demographic structure ad so on. Some emerging issues in activity and travel behavior are discu ssed in this sectio in the Indian context. These tr ends are considered to have potential implications on future transportation policies and extremely pertinent in the context of the development and implementation of advanced modeling approaches for passenger travel demand forecasting in India or similar developing countries. 2.5.1 Demographic Shifts The demographic trend of Indian populati on is typically iden tified by population explosion coupled with extremely differentia l growth rate between impoverished rural and modern urban areas. Economic bettermen t has been creating over-sized metropolitan areas with overwhelmingly dense settlements in and around the cities. The larger a city, the greater is its proportion of migrants and the more cosmopolitan is its population mix.

PAGE 39

24 in the een birth ing ated rging market segments such as mobile retired and female workers be critical in the development of future transport policies in India because the their pected to be potenti ally different from or ed ping f People with diverse background and lifestyle have their di fferent activity and travel behavior and that have been resulting increasingly complex urban travel pattern Indian cities. Over the past few decades, th e remarkable increase in the gap betw and death rates in India has resulted increa sing population pressure with a new elderly cohort. Travel behavior of su ch newly retired community is likely to be very different than their predecessor, because they are ac customed with more active and modern life style similar to their counterparts in th e developed world. Changing household structure is also an emerging transformation in th e demographic trend in India. Shrink household size in the urban areas with incr easing number of workers in the household and growing household income is epidemic acro ss urban areas in India. Elevated rate of labor force participation and education level among females have primarily aggrav this situation. The eme will activity-travel patterns and decision-making are ex traditional commuters like male workers. Female workers are likely to face the constraints and household obligations in th eir daily course that are not typically encountered by male workers due to the promin ent gender role in the Indian society. F instance, pattern of mode choice among females is likely to be very different than males because of their responsibilities related to childcare and household maintenance and increasing concerns about safety that tend to make females more reliant on personaliz means of transportation. Mobilizing more efforts to foster development and implementation of activity-based approach in the Indian context has become immensely pertinent in identifying such behavioral hete rogeneities and constrai nts underlying in the population and understand their growing role s on shaping peoples activity-travel engagement patterns. 2.5.2 Growing Challenges in Transit Sector Adequate provision of public transportation is a key element in maintaining social integration, economic development and envir onmental sustainability in the develo country context. In recent years in many Indi an cities, trend towards increased use o

PAGE 40

25 ugh transit market is exp ected to drop significantly among high and medium till ental stainability. Now, efficient operation of public transporta tion requires deep s, attitudes, va lues, awareness and level of aphic ifiable en te chnology and travel behavior has bee personal vehicles has threatened productive e fficiency of transit operation by degrading level of service of ur ban roadways resulted from congest ion, slower speed in the network and so on. Altho income market groups over the comi ng years due to their growing reliance on personal motor vehicles, maintaining improved quality of public transport service is s extremely important to meet future growth of travel demand and attract more people to use transit than personal automobiles for the sake of fuel shortage and environm su understanding of the users need s, preference tolerance that collectively shap es their decision to use alternative mode s of transportation The potential relevance of such information in the growth of transit market merits investigation about attitudinal preferences that can onl y be gathered by exploring consumers activity engagement and trip making behavior, which includes desired activity participation location, de sired time of day of travel by trip purpose, acceptable limits of travel time and cost, and other at tributes like socio-economic and demogr characteristics. Variables representing attit udinal preferences are not al ways easily quant and may not be directly derivable from traditional travel surveys. There is a growing interest in the application of advanced analytical techniques to travel behavior research that allows estimation of such unobservable aspects of travelers behavior. 2.5.3 ICT and Changing Travel Behavior Rapid advancements and innovations in information and telecommunication technology (ICT) have been transforming todays soci ety in an unprecedented manner. The profound impact of ICT has accentuated changes in peoples life style in developing countries including India as well. The close relationship betwe n a subject of interest to the travel behavior researchers in the developed world. Rapidly increasing adoption and market pe netration of technology have also been impacting activity and travel char acteristics in the count ries like India in recent years.

PAGE 41

26 an generation and wide use of mobile technologies have facilitated young people to schedule and ex ecute their ac tivities and commerce and trend of 24-7 and ng ing he e re an ted developments and residential lo cation choice patterns can be e xpected to have significant The use of cell phones, personal computer and internet has brought about many fundamental changes in activity and travel pa tterns especially from a scheduling and execution standpoint. For instance, as one can speculate, the tre nd of embracing such technologies is predominant among young trips through real-time pla nning. Similarly, advent of ebusiness and employment establishments have added more flexibil ity in scheduling participation of peoples daily activities. Shopping, gaming, hotel/car /flight re servation, work, banking and personal communication can be accomplished at home without havi to travel to physically separated activity locations. However, diversification of work or personal business arrangements, shopp and, recreational activities have resulted substa ntial complexity in activity-trave l patterns by eliminating many conventional trips and ge nerating more unconventional trips at t same time. Literature has recognized the relationship between technology and travel behavior is inextricably linke d to an understanding of the ti me-space interaction and tim use patterns of individuals (Pendyala and Bhat, 2004). Deeper understanding of the substitution and complementary effects of ICT on travel behavior is cr itical for analyzing and forecasting travel demand accurately. 2.5.4 Land Use and Travel Behavior Interaction Urban form and land use development ha ve profound influence on peoples travel behavior. Understanding inter actions between land use and transportation is even mo critical in the developing countries like India, where popul ation growth, rural to urb migration, economic development and rapid urbanization have been shaping land use patterns in major cities dramatically. Exam ining the impacts of land-use on travel behavior is crucial to designing non-motorized /transit-friendly transportation systems in the developing countries. Degrading living standard, skyrocketing housing cost, and increased adoption of personal automobile are some of the facts that have contribu growing trend of suburbanization along the frin ge of major Indian cities. Such land use

PAGE 42

27 city ke ple more reliant on their pe rsonal vehicles as these areas are poorly served by public here is a growing debate in the developing countries like US regarding ty recognizing the role of me dimension on individual activity-travel patte rns. The studies of activity engagement nd time use patterns have enabled to enrich our understanding of the complexity and ariability of individual travel behavior and that has led to increased capability of orecasting travel demand and evaluating pl anning options. However, the development nd implementation of activity-based methods have generally lagged in the developing ountries that seen in the de veloped world over the past deca des. But the emerging trends changing demographic composition, life styl e and technological advancements have been reshaping travel behavi or and activity patterns of the people in the developing countries like India in a dramatic wa y. Therefore, understanding the growing effect on commuter travel beha vior. Uncontrolled suburban se ttlements with growing boundaries are highly likely to cause significant increase in commuting length and ma peo transportation. T the effects of the natural and built environment on peoples trav el behavior in the light of the growing concern about the health a nd well-being of people (Handy et al., 2002). Researchers argued that subur ban development patterns, separation of residences, businesses, and employment centers, absence of pedestrian and bicycle facilities, inabili to serve outlying areas with re liable and high-frequency transit service, and the absence of grid-pattern street networks are all cont ributing factors to high levels of automobile dependency and inactive life st yle and consequently, epidem ic health problems like obesity (Pendyala and Bhat, 2004). Similar problems are quickly emerging in the societies of the developing nations like India as well and in th is regard, it is important to understand the cause and effect of relations hips underlying the influence of land use developments on travel behavior and choi ces from a developing country standpoint. 2.6 Summary Studies on activity engagement and time use be havior have gained significant momentum in the travel behavior profe ssion due to the emergence of activity-based approaches of travel behavior modeling. Activity-based a pproaches have contri buted considerable improvements over traditional tr ip-based methods by explicitly ti a v f a c in

PAGE 43

28 complexities in peoples travel behavior andctivity engagement patterns in accordance with these social transformati ons have becom extremely important in the context of the development and implementation of advanced modeling approaches for passenger travel demand forecasting in India or similar developing countries. a e

PAGE 44

29 HAPTER 3 d as a e is ties in India located in the state of Maharashtra. It is one of the major u as resulted migration of thousands of people from different parts of I ndia over the past years; this has led to an crease in sl um settlements, which are poorly served by The city spread over 50 sq-miles is expect ed to hold a population of 3 million by 2031 C EXPLORATORY ANALYSIS OF TRAVEL CHARACTERISTICS AND TIME USE BEHAVIOR The previous chapter has offered a review of the conceptual developments in travel behavior research and state of the art advances in tran sportation modeling techniques. Emerging trends in travel behavior from a developing country context have been addressed and the scopes of applying advan ced modeling methodologie s are discusse well. The current chapter includes a compar ative exploratory analysis of travel characteristics and time use behavior based on datasets from a developing country and developed country. 3.1 Introduction to the Study Areas This study builds upon a household travel survey sample available from the Thane metropolitan area situated on the west coast of India (near Bombay/Mumbai). Than one of the most vibrant ci rban conglomerations of Mumbai (formerly Bombay) Metropolitan Region (MMR) and is well connected to all parts of the country by rail and road. The city's proximity to Mumbai, the commercial capital of India and location at the geographical center of MMR has given tremendous impetus to the growth of industries in and around Thane. Job opportunities created by rapid indus trialization h acute housing shortage and in transport infrastructure and services. The population of Thane city as per 1991 census was 0.85 million and the city with a presen t population of about 1.55 million has the distinction of having its population doubled ev ery decade during the last five decades.

PAGE 45

30 S in terms of trends in population and economic activities in rn ion with a population deny and travel surveys as t nts are demographic, and other characteristics of households and persons. However, there are (web: District of Thane). The current population density of Thane is about 12,500 persons per square mile. A second data set used in the study is based on the Florida sample of the 2001 National Household Travel Survey conducted in the United States. Florida is one of the fastest growing states in the U ecet years. The current population of Florida is about 17.5 mill sit 300 persons per square miles (US Census Bureau). 3.2 Description of Surveys Two data sets, one from a developing count ry and another from a developed country are adopted in this study to conduct a comparative analysis of trav el characteristics and time use behavior between two different geographical areas. It is believed in the literature that the variability in travel behavior between different geographical contexts may be explained by exploring the differences in trav el characteristics, act ivity engagement time use patterns (Gangrade et al., 2000). The da ta sets are derived fr om household in which respondent samples provide d detailed trip information for a 24-hour period. The two surveys are: 2001 Household Travel Survey of the City of Thane, India 2001 National Household Travel Survey of the United States, Florida Sample While NHTS, Florida sample constitutes a survey at the regional level (i.e. the state of Florida), the survey from India is from a single metropolitan area in India. Unfortunately there is no regional or national tr avel survey in India that can be used for this analysis. However, as the City of Thane is a rather re presentative metropolitan area of India, it w considered suitable to serve as the developi ng country context for this study. Moreover, i was a large sample survey including a sa mple of 3,505 households and therefore is considered large enough from a model deve lopment and estimation perspective. There are differences and similarities be tween the surveys that should be noted here. Both the travel surveys are based on th e trip-diary format in which responde asked to provide detailed information a bout trips in addition to socio-economic,

PAGE 46

31 -assisted telephone interview (CATI) survey where avel data is retrieved for the househol d over the phone. Finally, the 2001 Thane, India survey where field workers actually visited households a low nces, the ousehold, person, and travel charact ne (unweighted) includes 3,016 individual s residing in 1,437 households and the corresponding number of total trips reported is 12,110. The India sample is typically characte rized by larger household size, higher number of children in the household and re markably low level of vehicle ownership relative to the US sample. Th e average household size is 4.12 in the India sample. This is substantially higher than average household size found in the US where average household size is typically in the 2.5 persons per household range. About two-thirds of the households in the Thane sample have four or more persons in the household. About 40 percent of the Indian households have at least one child below the age of 16 years while only 23 percent of the households in th e US sample have at least one child. The working status distributions of the house hold members are evident of some clear differences with respect to the survey administration method. The 2001 NHTS is a combination of mail-out/computer tr survey is a face-to-face in-person and interviewed people in their homes to retr ieve travel data. In a developing country context such as India, it is quite common to conduct face-to-face interviews involving large number of field workers because of the poor telecommunication systems, literacy rates, and the desire to obtain high response rates. Despite these differe survey data offer rather standard inform ation regarding h eristics and appear worthy of use in an international study of this nature. However, the author can not rule out the po ssibility that differences in results, model estimates, and findings among the three data sets may, in part, be due to differences in survey administration methods. 3.3 Household Characteristics Table 3.1 offers a detailed listing of household ch aracteristics of the data sets. The Tha data set includes 14,428 individuals residi ng in 3,505 households and 28,603 is total number of trips reported by the respo ndents, while the NHTS, Florida sample

PAGE 47

32 Table 3.1 Household Characteristics Thane, India NHTS, Florida Characteristic Statistics Statistics Sample Size (Unweighted) 3505 1437 Sample Size (Weighted) NA 6,179,588 Household Size 4.12 2.36 1 2.0% 28.53% 2 12.2% 36.06% 3 19.8% 16.06% 4 29.5% 13.23% 5+ 36.5% 6.12% Number of Children (Age 15 or below) 0.65 0.37 0 59.6% 76.99% 1 21.5% 11.79% 2 14.1% 6.84% 3+ 4.8% 3.53% Missing 0% 0.85% Number of Workers 1.34 1.18 0 9.3% 28.5% 1 57.3% 33.9% 2 25.8% 30.7% 3+ 7.6% 6.9% Number of Bicycles 0.19 0.77 0 82.1% 57.74% 1 16.9% 18.35% 2+ 1.0% 23.87% Missing 0% 0.04% Two-wheel vehicles (scooter + motorcycle) 0.22 NA 0 80.2% NA 1 17.9% NA 2+ 2.9% NA Auto Ownership 0.06 1.73 0 auto 94.7% 5.43% 1 auto 4.9% 40.69% 2 autos 0.4% 36.94% 3 or more autos 0% 16.94%

PAGE 48

33 Table 3.1 (Continued) Thane, India NHTS, Florida Characteristic Statistics Statistics Status of Home Ownership Own 79.3% 69.5% Rent 17.6% 30.0% Govt./Company provided house 5.0% 0.5% Number of Drivers 0.1 1.72 0 91.2% 3.9% 1 7.4% 35.3% 2 1.0% 48.3% 3+ 0.4% 12.5% Annual Household Income (USD) Missing NA 9.04% < $5,000 NA 2.42% $5,000 $9,999 NA 7.34% $10,000 $14,999 NA 6.87% $15,000 $19,999 NA 7.95% $20,000 $24,999 NA 6.19% $25,000 $29,999 NA 9.31% $30,000 $34,999 NA 4.15% $35,000 $39,999 NA 8.65% $40,000 $44,999 NA 3.48% $45,000 $49,999 NA 6.09% $50,000 $54,999 NA 3.25% $55,000 $59,999 NA 5.59% $60,000 $64,999 NA 1.72% $65,000 $69,999 NA 3.41% $70,000 $74,999 NA 1.60% $75,000 $79,999 NA 2.54% $80,000 $99,999 NA 4.57% $100,000 NA 5.84% differences between the Indian and US househol ds. Majority of the Indian households are one-worker households (57 percent) while the percentage of households with two or more workers in the US sample is higher than the India sample. This finding is quite obvious in the Indian society, where typically the male adult is the only earning member in the household and the female members are most lik ely to stay at home and bear household responsibilities. As expected, vehicle owners hip levels are very low with 95 percent of

PAGE 49

34 the households having zero cars in India samp le, whereas US sample shows an average two automobiles per household. Similarly, th e presence of licensed drivers in the household shows remarkable difference between the samples. However, 21 percent of households in the India sample have at leas t one two-wheeler. In the India sample, the percentage of households that own a home (80 percent) is 10 percent higher than the US (70 percent). Unfortunately, the information about household income is not available in the India sample. However, in the US samp le, about 40 percent of the households falls into low income category ( $30,000), about 30 percent falls into medium income category ($31,000 $60,000) and the correspondi ng percentage for high income category is about 20 percent. 3.4 Person Characteristics A comparison of person characteristics between the two data sets is given in Table 3.2. Comparison of the gender split shows that the pe rcentage of male (55 percent) is higher in the Indian sample, whereas the percentage of female (53 percent) is higher in the US sample. Average age of the India sample is lo wer than the US sample. With respect to the occupational distribution, that majority of individuals have reporte d their profession in sales or service, whereas 15 pe rcent of individuals have reported sales/se rvice and 18 percent of individuals have re ported their occupation as admini strative in the US sample. A significant percent of indivi duals fall into the categories such as student, homemaker and retired or unemployed at a total of 65 pe rcent in the India sample. Most likely, the other category that shares 50 percent of all occupations in the NHTS sample is the corresponding representation of these groups in the US context. NHTS 2001 does not differentiate between these categories. Hi gher percentage of individuals holds high school or college degrees in the US sample. Nearly 62 percent of individual reported no income in the India sample. The corresponding income information at the person level is not available in the US sample, however, majo rity of the individuals in the US sample belong to medium income households. Substantial presence of non-workers like students, homemakers and retired/unemployed persons in the India sample may be the reason behind the presence of such a large percent of zero-income people in the sample. Once

PAGE 50

35 again, a very small percent of i ndividuals is licensed to driv e in India relative to a vast majority of individuals is licensed to drive in the US. In general, the differences found in the comparison of household and person characteristics between the two survey sa mples are consistent with expectations. Furthermore, it was felt necessary to inve stigate the socio-economic and demographic characteristics by commuting status of indi viduals in understanding the exogenous factors behind the variability in their travel behavior.

PAGE 51

36 Table 3.2 Person Characteristics Thane, India NHTS, Florida Characteristics Statistics Statistics Sample Size (unweighted) 14428 3016 Sample Size (weighted) NA 1,45,29,896 Gender Male 55% 47.2% Female 45% 52.8% Average Age (in years) 31 39 < 5 0.9% 5.74% 5-15 16.5% 13.69% 16-21 15.7% 7.01% 22-49 50.6% 38.26% 50-64 12.2% 16.33% 65+ 3.9% 18.02% missing 0.2% 0.95% Occupation Sales/Service 23.8% 15.1% Farmer/Laborer 2.8% 8.1% Business/Professional 8.9% 18.3% Administrative NA 5.6% Other NA 52.9% Student 26.4% NA Homemaker 25.5% NA Retired/Unemployed 12.6% NA Highest Education Level Illiterate 8.0% NA Less than high school 63.2% 11.0% High school graduate 11.5% 25.1% College graduate 17.3% 18.9% Other (non-degree) NA 24.7% missing 0% 20.3%

PAGE 52

37 Table 3.2 (Continued) Thane, India NHTS, Florida Characteristics Statistics Statistics Monthly Personal Income (1 USD = Rs. 44) No income 62.2% NA Upto Rs. 2,000 6.1% NA Rs. 2,001Rs. 5,000 16.1% NA Rs. 5,001Rs. 10,000 9.7% NA Rs. 10,001Rs. 15,000 4.3% NA Rs. 15,001Rs. 20,000 0.5% NA Rs. 20,001 Rs. 30,000 0.1% NA Rs. 30,001 Rs. 40,000 0.1% NA Rs. 40,001+ 0.0% NA missing 0.9% NA Annual Household Income < $5,000 NA 1.9 $5,000 $14,999 NA 8.9 $15,000 $24,999 NA 8.9 $25,000 $34,999 NA 13.3 $35,000 $44,999 NA 13.4 $45,000 $54,999 NA 11.5 $55,000 $64,999 NA 10.7 $65,000 $74,999 NA 9.2 $75,000 $99,999 NA 5.2 >$100,000 NA 11.5 missing NA 5.5 Drivers License No license 91.8% 5.3% Two-wheeler 4.7% NA Auto 2.5% 94.7% 3.5 Person Characteristics by Commuting Status Comparisons of person characteristics between the commuter and non-commuter samples from both the data sets are presented in Table 3.3 and Table 3.4 respectively. Commuter and non-commuters samples are further stratified by gender status to understand the differences in socio-economic characteristics between males and females.

PAGE 53

38 Table 3.3 Commuter Characteristics Thane, India NHTS, FL Characteristics All Male Female All Male Female Sample Size (unweighted) 4646 3960 686 819 430 389 Sample Size (weighted) NA NA NA 46,94,974 24,41,294 22,53,680 Average Age (in years) 36 36 35 40 41 39 18-21 6.1% 6.0% 6.3% 8.1% 8.4% 7.7% 22-49 78.2% 77.3% 83.7% 66.1% 63.3% 69.1% 50-64 15.0% 15.9% 9.8% 21.4% 23.0% 19.6% 65+ 0.7% 0.8% 0.3% 4.5% 5.3% 3.6% Occupation Sales/Service 70.6% 69.4% 77.6% 29.6% 27.6% 31.8% Farmer/Laborer 6.9% 7.0% 6.4% 16.1% 26.1% 5.2% Business/Professional 21.5% 23.0% 13.0% 38.0% 40.7% 35.1% Administrative NA NA NA 13.6% 2.9% 25.1% Other NA NA NA 2.7% 2.7% 2.7% Student 0.3% 0.2% 0.4% NA NA NA Homemaker 0.4% 0.1% 2.3% NA NA NA Retired/Unemployed 0.3% 0.3% 0.3% NA NA NA Education Level Illiterate 5.1% 4.5% 8.6% NA NA NA Less than high school 53.3% 57.1% 31.6% 8.7% 10.5% 6.7% High school graduate 12.3% 12.8% 9.3% 26.8% 28.6% 24.8% College graduate 29.3% 25.6% 50.4% 25.7% 32.9% 45.9% Other (non-degree) NA NA NA 38.8% 28.0% 22.6% Monthly Personal Income No income 0.3% 0.1% 1.6% NA NA NA Upto Rs. 2,000 14.5% 13.9% 17.8% NA NA NA Rs. 2,001Rs. 5,000 42.7% 43.4% 38.9% NA NA NA Rs. 5,001Rs. 10,000 28.1% 27.7% 30.6% NA NA NA Rs. 10,001Rs. 15,000 10.0% 10.8% 5.8% NA NA NA Rs. 15,001Rs. 20,000 1.2% 1.3% 1.0% NA NA NA Rs. 20,001 Rs. 30,000 0.3% 0.3% 0.1% NA NA NA Rs. 30,001 Rs. 40,000 0.3% 0.3% 0.1% NA NA NA Rs. 40,001+ 0.1% 0.1% 0% NA NA NA missing 2.4% 2.1% 3.9% NA NA NA

PAGE 54

39 Table 3.3 (Continued) Thane, India NHTS, FL Characteristics All Male Female All Male Female Annual Household Income < $5,000 NA NA NA 0.6% 0.8% 0.3% $5,000 $14,999 NA NA NA 10.8% 15.0% 6.2% $15,000 $24,999 NA NA NA 5.7% 2.3% 9.3% $25,000 $34,999 NA NA NA 9.5% 8.7% 10.3% $35,000 $44,999 NA NA NA 12.5% 9.2% 16.0% $45,000 $54,999 NA NA NA 10.7% 12.6% 8.8% $55,000 $64,999 NA NA NA 13.3% 14.0% 12.5% $65,000 $74,999 NA NA NA 11.2% 11.6% 10.7% $75,000 $99,999 NA NA NA 7.1% 6.8% 7.5% >$100,000 NA NA NA 13.8% 15.0% 12.6% missing NA NA NA 4.9% 4.1% 5.8% Drivers License No license 82.1% 80.5% 91.8% 4.1% 3.6% 4.5% Two-wheeler 11.7% 12.8% 5.7% NA NA NA Auto 6.1% 6.8% 2.5% 95.9% 96.4% 95.5% Commuter and non-commuter samples are derived from the market segment: mobile adults. Mobile adults are those indi viduals who reported at least one trip on the travel survey day and are 18 years of age or ab ove. All mobile adults who made at least one work trip or work-related business trip on the travel survey day were treated as commuters and all others were treated as non-commuters. All the i ndividuals who did not report any trip at all are categorized as zero-trip makers (see Table 3.5). Table 3.3 offers a detailed look at the pers on socioeconomic characteristics of the commuter sample by gender status. The ge nder splits between both Indian and US commuter samples are quite interesting. Consis tent with contemporary Indian society, a large proportion of the commuters are male at 85 percent while in the US context the gender split is close to 50 percent. These fi ndings can be explained by the fact that the labor force participation rate among females in India is still quite low compared to the US context. The age profile shows that the av erage Indian commuters are younger than the US commuters. Although the average age of th e different groups is quite similar, the distributions are quite different. As expect ed, a large proportion of workers fall in the 22-

PAGE 55

40 49 year range in both survey samples. With respect to educational level, majority of the commuters in the India sample do not have college degree as opposed to the US sample. However quite remarkably, 50 percent of fe male commuters in the India sample are found to have college degree. This is consiste nt with expectations; if females are working in the Indian context, it is generally becau se they are well-educat ed and are putting the education to use. Most Indian commute rs report incomes in the Rs. 2,001 Rs. 10,000 range on a monthly basis. Low level of pers onal income is prev alent among the Indian commuters which otherwise indicate very low standard of living compared to the developed world. The occupational distribution shows that the majority of the Indian commuters are service workers while in the US majority are professi onals, which is also quite consistent with the educational profile s of both samples. However, percentage of females in administrative job is significantly higher than males in the US sample. As expected, a very small percent of individuals is licensed to driv e in India relative to a vast majority of individuals is licensed to drive in the US. The person characteristics of non-commu ters are provided in Table 3.4. The majority of non-commuters are male in the I ndia sample, but female in the US sample. Many non-working females in the India sample were found to have zero trips associated with their records either due to legitimate zer o trip making behavior or not responding to the survey. With respect to age profile, it is found that non-com muters are young studentage oriented in the India sample while th e non-commuters in the US sample are more retirement age oriented. With respect to occupational distribution, non-commuters in the India sample show considerable differences between males and females. More than 50 percent male commuters are students and rest s are mainly retired or unemployed people in the India sample. More than 50 percent of female non-commuters are homemakers. No male non-commuter indicates his occupation as homemaker indicating the presence of a rather strong gender role in the household. A vast major ity of Indian non-commuters have reported zero monthly income, which is also consistent with the occupational distribution. Some male non-wo rkers report income, possibly due to retirement income. Again as expected, very few individuals have drivers license in the India sample as opposed to the US sample.

PAGE 56

41 Table 3.4 Non-Commuter Characteristics Thane NHTS, FL Characteristics All Male Female All Male Female Sample Size (unweighted) 1677 900 777 1633 690 943 Sample Size (weighted) NA NA NA 81,06,725 34,55,990 46,50,735 Average Age (in years) 31 31 30 52 51 53 18-21 43.1% 50.0% 35.1% 5.1% 5.6% 4.7% 22-49 40.8% 29.0% 54.6% 40.5% 41.7% 39.6% 50-64 10.4% 12.8% 7.6% 21.5% 20.5% 22.2% 65+ 5.7% 8.2% 2.7% 33.0% 32.2% 33.5% Occupation Sales/Service 2.8% 4.3% 1.0% 12.2% 13.8% 11.0% Farmer/Laborer 1.4% 2.0% 0.6% 6.7% 14.3% 1.1% Business/Professional 5.8% 9.2% 1.9% 14.9% 17.1% 13.3% Administrative NA NA NA 3.3% 0.8% 5.2% Other NA NA NA 62.8% 54.0% 69.4% Student 47.6% 56.0% 37.8% NA NA NA Homemaker 25.1% 0.8% 53.3% NA NA NA Retired/Unemployed 17.3% 27.7% 5.3% NA NA NA Education Level Illiterate 4.9% 2.4% 7.9% NA NA NA Less than high school 43.8% 42.4% 45.4% 13.0% 13.7% 12.6% High school graduate 24.2% 27.2% 20.7% 35.0% 33.0% 36.6% College graduate 27.0% 27.9% 26.0% 23.4% 26.0% 21.6% Other (non-degree) NA NA NA 28.6% 27.3% 29.2% Monthly Personal Income No income 82.6% 72.4% 94.5% NA NA NA Upto Rs. 2,000 4.9% 7.3% 2.1% NA NA NA Rs. 2,001Rs. 5,000 7.8% 12.8% 1.9% NA NA NA Rs. 5,001Rs. 10,000 2.4% 3.7% 0.9% NA NA NA Rs. 10,001Rs. 15,000 2.0% 3.2% 0.5% NA NA NA Rs. 15,001Rs. 20,000 0.3% 0.4% 0.1% NA NA NA Rs. 20,001 Rs. 30,000 0% 0% 0% NA NA NA Rs. 30,001 Rs. 40,000 0% 0% 0% NA NA NA Rs. 40,001+ 0.1% 0.1% 0% NA NA NA

PAGE 57

42 Table 3.4 (Continued) Thane NHTS, FL Characteristics All Male Female All Male Female Annual Household Income < $5,000 NA NA NA 2.7% 1.6% 3.6% $5,000 $14,999 NA NA NA 7.2% 6.7% 7.5% $15,000 $24,999 NA NA NA 10.3% 7.9% 12.0% $25,000 $34,999 NA NA NA 14.4% 16.5% 12.7% $35,000 $44,999 NA NA NA 13.7% 13.5% 13.8% $45,000 $54,999 NA NA NA 12.4% 13.7% 11.5% $55,000 $64,999 NA NA NA 9.7% 9.6% 9.8% $65,000 $74,999 NA NA NA 7.8% 9.5% 6.6% $75,000 $99,999 NA NA NA 4.1% 4.3% 3.9% >$100,000 NA NA NA 9.9% 8.9% 10.6% missing NA NA NA 7.9% 7.7% 8.0% Drivers License No license 93% 89.9% 97.6% 6.9% 4.9% 8.5% Two-wheeler 5% 7.0% 1.7% NA NA NA Auto 2% 3.1% 0.6% 93.1% 95.1% 91.5% 3.6 Person Characteristics of Zero-Trip Makers Table 3.5 provides a glimpse of the demographic characteristics of the zero-trip makers in the Thane and NHTS samples. The percentage of zero trip makers among all the survey respondents is substantially high er in the Indian context (40%) relative to the US context (13%). Females are present at higher percenta ges than males in both cases and with much greater majority in the India sample. This fi nding is quite consiste nt with occupational distribution in the Indian c ontext where homemakers share a significant percentage in the sample. The age-profiles of both the samples s uggest that zero-trip makers are mostly at their middle or retirement age groups. However, the percentage of indi viduals in their 65 years or higher age group is much higher in the US sample relative to the India sample. As expected, majority of zero-trip makers do not have college education in both contexts. Again, a very few minority of individuals are li censed to drive in India as opposed to the US context.

PAGE 58

43 Table 3.5 Person Characteristics of Zero-Trip Makers Thane, India NHTS, Florida Characteristics Statistics Statistics (weighted) Sample Size (unweighted) 5263 401 Sample Size (weighted) NA 2,060,716 Gender Male 27.7% 41.2% Female 72.3% 58.8% Average Age (in years) 38 44 < 5 0.3% 11.1% 5-15 3.2% 11.8% 16-21 14.3% 5.3% 22-49 56.8% 23.9% 50-64 16.8% 16.4% 65+ 8.4% 31.6% missing 0.2% 0% Occupation Sales/Service 1.6% 8.8% Farmer/Laborer 0.7% 2.7% Business/Professional 3.4% 8.2% Administrative NA 2.3% Other NA 78.0% Student 4.2% NA Homemaker 61.5% NA Retired/Unemployed 28.6% NA Highest Education Level Illiterate 15.0% NA Less than high school 63.4% 18.7% High school graduate 9.3% 24.8% College graduate 12.3% 9.9% Other (non-degree) NA 20.9% Missing 0% 25.7% Drivers License No license 97.9% 8.5% Two-wheeler 1.1% NA Auto 0.9% 91.5% Auto Ownership Car Two-wheeler 0.5% 1.0% NA NA

PAGE 59

44 Table 3.5 (Continued) Thane, India NHTS, Florida Characteristics Statistics Statistics (weighted) Monthly Personal Income (1 USD = Rs. 44) No income 91.0% NA Upto Rs. 2,000 2.0% NA Rs. 2,001Rs. 5,000 3.6% NA Rs. 5,001Rs. 10,000 1.0% NA Rs. 10,001Rs. 15,000 2.2% NA Rs. 15,001Rs. 20,000 0.2% NA Rs. 20,001 Rs. 30,000 0% NA Rs. 30,001 Rs. 40,000 0% NA Rs. 40,001+ 0% NA missing 0% NA Annual Household Income < $5,000 NA 3.5% $5,000 $14,999 NA 5.8% $15,000 $24,999 NA 12.5% $25,000 $34,999 NA 12.4% $35,000 $44,999 NA 11.9% $45,000 $54,999 NA 13.7% $55,000 $64,999 NA 13.0% $65,000 $74,999 NA 7.4% $75,000 $99,999 NA 1.9% >$100,000 NA 9.1% missing NA 9.0% 3.7 Development of Trip Production Rates This section offers the cross-classification ta bles of trip producti on rates at the household and personal levels. The trip production tables consist of mean trip rates for each trip purpose and further resolving the trip rates based on household or personal characteristics. Such tabulations are create d for the India sample and compared with similar trip production tabl es based on NHTS, FL sample where applicable. The characteristics used in this process are the num ber of vehicles (In cas e of India, cars and two-wheelers are treated separa tely), including categories 0, 1+ (categories 0, 1, 2, >2 used for US sample), and the household size, including categories 1, 2, 3, 4, and 5+

PAGE 60

45 (where 5+ indicates 5 or more household me mbers). Additional tables are generated based on personal/household income levels as well. Table 3.6 provides the definitions used in this analysis for categorizing the trip purposes based on survey responses. Table 3.6 Trip Purpose Definitions Trip Purpose Origin Descriptor* Destination Descriptor* HBA: Home-Based All Trips Home All Purpose HBW: Home-Based Work Home Work HBS: Home-Based Shopping Home Shopping/Errands HBSC: Home-Based School Home School Home Meal Home Social Home Recreational HBSocRec: Home-Based Social/Recreational Home Friend/Relatives Home HBO: Home-Based Other Home Other HBU: Home-Based Unknown Home Unknown NHBA: Non-Home-Based All All but Home All but Home NHBW: Non-Home-Based Work All but Home Work NHBO: Non-Home-Based-Other All but Home/Work All but Home/Work Trip origin and destination descriptor can be revers ed and still be considered as the same trip purpose Table 3.7 and Table 3.8 presents the household trips production rates by household size and household car ownership fo r India and US sample respectively. As expected, household trip rates in general te nd to increase consistent ly with household size and vehicle ownership for both contexts. Aver age rate of total trips reported by the US households are much greater compared to th at reported by the Indi an household (Average total trip rate: US 8.18 vs. India 5.31). Ho wever, home-based work and home-based school trips produced by the Indian households are much higher than the US households while average home-based shopping, socio/recreational, and non-home-based trip rates reported by the Indian households are much lower than the US households.

PAGE 61

46 Table 3.7 Household Trip Production Rate by Household Size and Household Car Ownership: India Household Size Trip Purpose Household Car Ownership 1 2 3 4 5+ Weighted Average 0 68 412 663 970 1211 3324 1+ 1 17 36 63 64 181 Cell Counts Total 69 429 699 1033 1275 3505 0 2.09 2.74 4.15 5.37 6.90 5.29 1+ 4.00 2.82 4.75 5.76 7.08 5.74 All Trips Total 2.12 2.75 4.18 5.39 6.90 5.31 0 1.38 1.94 2.41 2.64 3.22 2.69 1+ 4.00 2.12 2.83 2.79 3.67 3.06 HBW Total 1.42 1.94 2.43 2.65 3.24 2.71 0 0.03 0.08 1.16 2.17 3.14 2.02 1+ 0.00 0.00 1.28 2.43 2.92 2.13 HBSCH Total 0.03 0.07 1.16 2.19 3.12 2.02 0 0.24 0.33 0.26 0.23 0.21 0.24 1+ 0.00 0.35 0.28 0.16 0.16 0.20 HBS Total 0.23 0.33 0.26 0.23 0.21 0.24 0 0.32 0.20 0.16 0.16 0.15 0.16 1+ 0.00 0.24 0.22 0.16 0.19 0.19 HBSocRec Total 0.32 0.21 0.17 0.16 0.15 0.17 0 0.12 0.16 0.13 0.13 0.15 0.14 1+ 0.00 0.12 0.03 0.13 0.06 0.08 HBO Total 0.12 0.16 0.13 0.13 0.14 0.14 0 0.00 0.02 0.02 0.03 0.02 0.02 1+ 0.00 0.00 0.08 0.06 0.08 0.07 NHBW Total 0.00 0.02 0.03 0.03 0.02 0.02 0 0.00 0.00 0.00 0.01 0.00 0.00 1+ 0.00 0.00 0.00 0.03 0.00 0.01 NHBO Total 0.00 0.00 0.00 0.01 0.00 0.01

PAGE 62

47 Table 3.8 Household Trip Production Rate by Household Size and Household Car Ownership: USA Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 219315 46127 47672 20628 1610 335352 1 1377529 681192 229975 153785 71959 2514440 2 114709 1229334 388027 383050 167511 2282631 >2 51385 271431 326625 260328 137395 1047164 Cell Counts Total 1762938 2228084 992299 817791 378475 6179587 0 2.16 4.69 7.74 2.37 22 3.41 1 4.21 7.33 6.08 11.02 13.13 5.90 2 4.27 7.55 10.25 15.21 16.52 9.79 >2 2.77 7.24 10.71 15.05 19.54 11.66 All Trips Total 3.92 7.38 9.32 14.05 17.00 8.18 0 0.14 0.64 1.28 0.00 3.00 0.38 1 0.45 0.42 0.72 0.69 1.67 0.52 2 0.85 0.87 1.12 1.45 1.52 1.05 >2 0.90 1.29 1.72 2.00 2.44 1.73 HBW Total 0.45 0.78 1.23 1.44 1.89 0.91 0 0.85 1.59 3.86 0.79 0.00 1.37 1 0.98 1.59 0.85 2.72 0.93 1.24 2 0.73 1.57 1.72 2.78 2.35 1.82 >2 0.88 0.75 1.86 2.33 2.45 1.72 HBS Total 0.95 1.48 1.67 2.58 2.11 1.54 0 0.23 0.47 1.33 0.79 7.00 0.49 1 0.81 1.89 0.89 1.73 1.43 1.18 2 1.17 1.68 1.78 3.36 2.73 2.03 >2 1.00 2.08 1.69 2.69 3.05 2.18 HBSocRec Total 0.76 1.77 1.52 2.78 2.62 1.63 0 0.12 0.00 0.32 0.00 4.00 0.14 1 0.03 0.18 0.79 0.76 2.56 0.26 2 0.00 0.16 0.65 0.91 1.79 0.48 >2 0.00 0.16 0.47 1.05 2.05 0.72 HBSCH Total 0.04 0.16 0.61 0.90 2.04 0.41 0 0.29 0.58 0.00 0.00 2.00 0.28 1 0.56 0.92 1.19 2.27 3.31 0.90 2 0.44 0.84 1.73 2.98 3.94 1.56 >2 0.00 0.53 1.12 2.13 3.78 1.51 HBO Total 0.50 0.82 1.32 2.50 3.76 1.21

PAGE 63

48 Table 3.8 (Continued) Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 0.00 0.00 0.00 0.00 1.00 0.00 1 0.00 0.01 0.01 0.00 0.03 0.00 2 0.00 0.01 0.00 0.02 0.00 0.01 >2 0.00 0.00 0.00 0.00 0.05 0.01 HBU Total 0.00 0.01 0.00 0.01 0.03 0.01 0 0.07 0.23 0.47 0.00 1.00 0.15 1 0.45 0.46 0.50 0.60 1.02 0.48 2 0.58 0.65 1.05 1.02 1.44 0.83 >2 0.00 0.87 1.31 1.66 1.84 1.29 NHBW Total 0.40 0.61 0.98 1.12 1.50 0.73 0 0.46 1.19 0.47 0.79 4.00 0.60 1 0.93 1.86 1.12 2.26 2.18 1.32 2 0.51 1.78 2.20 2.70 2.75 2.01 >2 0.00 1.57 2.54 3.19 3.88 2.50 NHBO Total 0.82 1.77 1.98 2.72 3.06 1.74 Person trip production rates by household car ownership and household size for the India and US samples are presente d in Table 3.9 and Table 3.10 respectively. According to Table 3.9, person trip rates in the India sample gene rally decrease with household size and increase with car ownership while in the US context, person trip rates tend to increase with household vehicle owne rship but no such clear relationship is found between person trip rate a nd household size. However, it is found that person trip production rates in single or tw o-person households with car av ailability are consistently higher than any other households unit in both samples. Average person trip rates in the US sample are quite higher than the India sample as expected except home-based work and home-based school trips rates, on average are higher in the Indian context. Very few respondents reported non-home-based tr ips in the India sample.

PAGE 64

49 Table 3.9 Person Trip Production Rate by Household Size and Household Car Ownership: India Household Size Trip Purpose Household Car Ownership 1 2 3 4 5+ Weighted Average 0 68 822 1977 3855 6933 13655 1+ 1 34 108 252 378 773 Cell Counts Total 69 856 2085 4107 7311 14428 0 2.09 1.37 1.38 1.34 1.21 1.29 1+ 4.00 1.41 1.58 1.44 1.20 1.34 All Trips Total 2.12 1.37 1.39 1.35 1.21 1.29 0 1.38 0.97 0.80 0.66 0.57 0.66 1+ 4.00 1.06 0.94 0.70 0.62 0.72 HBW Total 1.42 0.97 0.81 0.66 0.57 0.66 0 0.03 0.04 0.38 0.55 0.55 0.49 1+ 0.00 0.00 0.43 0.61 0.49 0.50 HBSCH Total 0.03 0.04 0.39 0.55 0.55 0.49 0 0.24 0.17 0.09 0.06 0.04 0.06 1+ 0.00 0.18 0.09 0.04 0.03 0.05 HBS Total 0.23 0.17 0.09 0.06 0.04 0.06 0 0.32 0.10 0.06 0.04 0.03 0.04 1+ 0.00 0.12 0.07 0.04 0.03 0.04 HBSocRec Total 0.32 0.10 0.06 0.04 0.03 0.04 0 0.12 0.08 0.04 0.03 0.03 0.03 1+ 0.00 0.06 0.01 0.03 0.01 0.02 HBO Total 0.12 0.08 0.04 0.03 0.03 0.03 0 0.00 0.01 0.01 0.01 0.00 0.01 1+ 0.00 0.00 0.03 0.02 0.01 0.02 NHBW Total 0.00 0.01 0.01 0.01 0.00 0.01 0 0.00 0.00 0.00 0.00 0.00 0.00 1+ 0.00 0.00 0.00 0.01 0.00 0.00 NHBO Total 0.00 0.00 0.00 0.00 0.00 0.00

PAGE 65

50 Table 3.10 Person Trip Production Rate by Household Size and Household Car Ownership: USA Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 171885 98048 178710 34683 7840 491166 1 1160148 1410323 677895 585189 319303 4152858 2 98456 2562465 1236428 1616439 850395 6364183 >2 45074 541002 1008189 1109241 818184 3521690 Cell Counts Total 1475563 4611838 3101222 3345552 1995722 14529897 0 3.11 0.53 3.03 3.00 4.66 2.81 1 4.24 3.46 2.58 4.22 2.82 3.54 2 4.85 4.26 3.75 4.11 4.02 4.10 >2 2.57 4.33 3.97 4.24 4.17 4.14 All Trips Total 4.14 4.05 3.63 4.16 3.94 3.97 0 0.07 0.53 0.21 0.00 0.71 0.21 1 0.36 0.26 0.17 0.33 0.20 0.27 2 1.08 0.49 0.46 0.30 0.33 0.42 >2 0.31 0.75 0.69 0.62 0.43 0.61 HBW Total 0.40 0.46 0.50 0.43 0.36 0.44 0 0.85 0.00 2.14 1.00 0.00 1.37 1 1.03 0.82 0.27 1.00 0.05 0.73 2 1.02 0.93 0.61 0.67 0.62 0.76 >2 1.69 0.48 0.66 0.70 0.56 0.63 HBS Total 1.04 0.84 0.65 0.72 0.52 0.73 0 0.47 0.00 0.45 1.00 1.55 0.49 1 0.71 0.61 0.39 0.49 0.30 0.56 2 0.82 0.94 0.49 1.08 0.72 0.85 >2 0.58 1.19 0.69 0.80 0.59 0.76 HBSocRec Total 0.69 0.88 0.55 0.90 0.62 0.75 0 0.00 0.00 0.24 0.00 0.84 0.15 1 0.01 0.12 0.24 0.30 0.84 0.20 2 0.00 0.07 0.24 0.31 0.47 0.22 >2 0.00 0.00 0.16 0.32 0.46 0.26 HBSCH Total 0.01 0.07 0.21 0.31 0.51 0.23 0 0.46 0.00 0.00 0.00 0.33 0.13 1 0.72 0.57 0.77 0.90 0.87 0.71 2 0.56 0.44 0.72 0.86 0.85 0.66 >2 0.00 0.37 0.44 0.49 0.77 0.52 HBO Total 0.66 0.45 0.58 0.70 0.82 0.61

PAGE 66

51 Table 3.10 (Continued) Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 0.00 0.00 0.00 0.00 0.18 0.01 1 0.00 0.00 0.01 0.00 0.00 0.00 2 0.00 0.01 0.00 0.01 0.00 0.01 >2 0.00 0.00 0.00 0.00 0.02 0.00 HBU Total 0.00 0.01 0.00 0.00 0.01 0.00 0 0.33 0.00 0.00 0.00 0.24 0.09 1 0.52 0.35 0.22 0.37 0.06 0.35 2 0.82 0.40 0.40 0.19 0.31 0.34 >2 0.00 0.67 0.55 0.38 0.40 0.47 NHBW Total 0.51 0.42 0.42 0.29 0.32 0.38 0 0.94 0.00 0.00 1.00 0.82 0.36 1 0.89 0.74 0.52 0.81 0.49 0.73 2 0.54 0.99 0.82 0.70 0.73 0.85 >2 0.00 0.87 0.78 0.91 0.94 0.86 NHBO Total 0.83 0.91 0.72 0.80 0.79 0.82 Table 3.11 through 3.17 provides person tr ip production rates but further resolving the trip rates based on personal (in the Indian cont ext) or household income (in the US context). Overall comparison between different income groups suggests that average trip rates increase with income in both survey samples. Home-based educational trip rates are generally higher among low inco me groups. This finding is quite expected because presence of students is likely to be predominant in the zero or low income range.

PAGE 67

52 Table 3.11 Trip Production Rates of Persons with Zero Income: India Household Size Trip Purpose Household Car Ownership 1 2 3 4 5+ Weighted Average 0 2 358 1087 2408 4682 8537 1+ 0 12 50 143 230 435 Cell Counts Total 2 370 1137 2551 4912 8972 0 2.00 0.57 0.93 1.02 0.91 0.93 1+ 0.00 0.50 1.24 1.15 0.83 0.97 All Trips Total 2.00 0.57 0.95 1.03 0.91 0.94 0 0.00 0.00 0.00 0.00 0.00 0.00 1+ 0.00 0.00 0.04 0.00 0.01 0.01 HBW Total 0.00 0.00 0.00 0.00 0.00 0.00 0 1.00 0.09 0.69 0.87 0.81 0.78 1+ 0.00 0.00 0.92 1.05 0.77 0.86 HBSCH Total 1.00 0.09 0.70 0.88 0.80 0.78 0 0.00 0.31 0.12 0.08 0.05 0.08 1+ 0.00 0.33 0.12 0.05 0.03 0.05 HBS Total 0.00 0.31 0.12 0.08 0.05 0.08 0 1.00 0.09 0.06 0.04 0.03 0.04 1+ 0.00 0.17 0.16 0.01 0.02 0.04 HBSocRec Total 1.00 0.10 0.07 0.04 0.02 0.04 0 0.00 0.07 0.06 0.03 0.03 0.04 1+ 0.00 0.00 0.00 0.03 0.00 0.01 HBO Total 0.00 0.07 0.06 0.03 0.03 0.03 0 0.00 0.00 0.00 0.00 0.00 0.00 1+ 0.00 0.00 0.00 0.00 0.00 0.00 NHBW Total 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 0.00 0.00 0.00 0.00 0.00 1+ 0.00 0.00 0.00 0.01 0.00 0.00 NHBO Total 0.00 0.00 0.00 0.00 0.00 0.00

PAGE 68

53 Table 3.12 Trip Production Rates of Persons with Low Income: India Household Size Trip Purpose Household Car Ownership 1 2 3 4 5+ Weighted Average 0 48 305 529 832 1429 3143 1+ 0 0 4 13 45 62 Cell Counts Total 48 305 533 845 1474 3205 0 2.04 1.96 1.95 1.88 1.81 1.87 1+ 0.00 0.00 2.00 1.54 1.64 1.65 All Trips Total 2.04 1.96 1.95 1.88 1.81 1.87 0 1.38 1.64 1.76 1.73 1.69 1.70 1+ 0.00 0.00 2.00 1.31 1.27 1.32 HBW Total 1.38 1.64 1.76 1.72 1.68 1.70 0 0.00 0.00 0.01 0.01 0.02 0.01 1+ 0.00 0.00 0.00 0.00 0.13 0.10 HBSCH Total 0.00 0.00 0.01 0.01 0.02 0.01 0 0.17 0.07 0.07 0.03 0.02 0.04 1+ 0.00 0.00 0.00 0.00 0.09 0.06 HBS Total 0.17 0.07 0.07 0.03 0.02 0.04 0 0.33 0.12 0.05 0.05 0.03 0.05 1+ 0.00 0.00 0.00 0.00 0.13 0.10 HBSocRec Total 0.33 0.12 0.05 0.05 0.04 0.06 0 0.17 0.10 0.03 0.03 0.03 0.04 1+ 0.00 0.00 0.00 0.15 0.00 0.03 HBO Total 0.17 0.10 0.03 0.03 0.03 0.04 0 0.00 0.02 0.02 0.02 0.01 0.02 1+ 0.00 0.00 0.00 0.08 0.02 0.03 NHBW Total 0.00 0.02 0.02 0.02 0.01 0.02 0 0.00 0.00 0.00 0.00 0.00 0.00 1+ 0.00 0.00 0.00 0.00 0.00 0.00 NHBO Total 0.00 0.00 0.00 0.00 0.00 0.00

PAGE 69

54 Table 3.13 Trip Production Rates of P ersons with Low Household Income: USA Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 132528 63653 178710 16684 7840 399415 1 590585 676522 348664 408360 263694 2287825 2 25139 740356 298112 453427 296084 1813118 >2 25203 107383 168134 263432 282641 846793 Cell Counts Total 773455 1587914 993620 1141903 850259 5347151 0 2.29 2.65 2.69 0 4.66 2.48 1 4.18 3.91 2.23 2.85 3.38 3.47 2 2.92 3.79 2.91 3.56 3.39 3.51 >2 2.25 3.72 3.72 3.91 3.74 3.74 All Trips Total 3.76 3.79 2.76 3.34 3.51 3.45 0 0.13 0.41 0.52 0.00 0.71 0.35 1 0.49 0.41 0.27 0.17 0.52 0.38 2 0.68 0.29 0.42 0.52 0.44 0.40 >2 1.22 0.45 0.39 0.35 0.74 0.53 HBW Total 0.45 0.36 0.38 0.35 0.57 0.41 0 0.98 1.30 1.18 0.00 0.00 1.06 1 1.15 0.75 0.37 0.89 0.26 0.76 2 0.03 0.83 0.49 0.46 0.33 0.59 >2 1.03 0.50 0.82 0.44 0.46 0.55 HBS Total 1.08 0.79 0.63 0.60 0.35 0.69 0 0.23 0.19 0.45 0.00 1.55 0.34 1 0.70 0.97 0.24 0.50 0.27 0.63 2 1.55 0.79 0.53 0.63 0.22 0.63 >2 0.00 1.56 0.15 0.97 0.43 0.67 HBSocRec Total 0.63 0.90 0.35 0.65 0.32 0.61 0 0.14 0.00 0.13 0.00 0.84 0.12 1 0.06 0.11 0.37 0.12 0.65 0.20 2 0.00 0.20 0.18 0.44 0.61 0.32 >2 0.00 0.08 0.39 0.12 0.33 0.24 HBSCH Total 0.07 0.14 0.28 0.25 0.53 0.24 0 0.29 0.17 0.00 0.00 0.33 0.13 1 0.62 0.53 0.30 0.41 0.83 0.53 2 0.03 0.56 0.55 0.84 0.94 0.68 >2 0.00 0.40 0.36 0.72 0.80 0.61 HBO Total 0.53 0.52 0.33 0.65 0.85 0.57

PAGE 70

55 Table 3.13 (Continued) Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 0.00 0.00 0.00 0.00 0.18 0.00 1 0.00 0.00 0.00 0.00 0.01 0.00 2 0.00 0.00 0.01 0.02 0.00 0.01 >2 0.00 0.00 0.00 0.00 0.00 0.00 HBU Total 0.00 0.00 0.00 0.01 0.00 0.00 0 0.00 0.20 0.20 0.00 0.24 0.13 1 0.30 0.28 0.17 0.10 0.29 0.24 2 0.60 0.24 0.16 0.05 0.32 0.20 >2 0.00 0.10 0.19 0.40 0.35 0.29 NHBW Total 0.25 0.24 0.18 0.15 0.32 0.22 0 0.52 0.39 0.20 0.00 0.82 0.34 1 0.85 0.86 0.50 0.66 0.55 0.73 2 0.03 0.88 0.57 0.61 0.53 0.70 >2 0.00 0.64 1.42 0.91 0.63 0.85 NHBO Total 0.74 0.84 0.62 0.69 0.57 0.71

PAGE 71

56 Table 3.14 Trip Production Rates of Persons with Medium Income: India Household Size Trip Purpose Household Car Ownership 1 2 3 4 5+ Weighted Average 0 17 149 340 568 719 1793 1+ 1 17 46 72 85 221 Cell Counts Total 18 166 386 640 804 2014 0 2.24 2.03 1.91 1.88 1.85 1.89 1+ 4.00 1.88 1.85 1.81 1.80 1.83 All Trips Total 2.33 2.02 1.90 1.87 1.84 1.88 0 1.53 1.86 1.80 1.79 1.80 1.80 1+ 4.00 1.65 1.65 1.60 1.67 1.65 HBW Total 1.67 1.84 1.78 1.77 1.78 1.78 0 0.00 0.00 0.01 0.01 0.00 0.01 1+ 0.00 0.00 0.00 0.00 0.04 0.01 HBSCH Total 0.00 0.00 0.01 0.01 0.01 0.01 0 0.47 0.03 0.02 0.01 0.01 0.02 1+ 0.00 0.00 0.09 0.04 0.00 0.03 HBS Total 0.44 0.02 0.03 0.01 0.01 0.02 0 0.24 0.05 0.04 0.03 0.02 0.03 1+ 0.00 0.12 0.00 0.11 0.02 0.05 HBSocRec Total 0.22 0.06 0.04 0.04 0.02 0.03 0 0.00 0.07 0.02 0.02 0.01 0.02 1+ 0.00 0.12 0.02 0.01 0.02 0.03 HBO Total 0.00 0.07 0.02 0.02 0.01 0.02 0 0.00 0.02 0.01 0.01 0.01 0.01 1+ 0.00 0.00 0.07 0.03 0.05 0.04 NHBW Total 0.00 0.02 0.02 0.02 0.01 0.01 0 0.00 0.00 0.00 0.00 0.00 0.00 1+ 0.00 0.00 0.00 0.01 0.00 0.00 NHBO Total 0.00 0.00 0.00 0.00 0.00 0.00

PAGE 72

57 Table 3.15 Trip Production Rates of Pers ons with Medium Household Income: USA Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 0 0 0 0 0 0 1 279175 342644 162891 119480 55609 959799 2 36605 627321 330835 349533 184253 1528547 >2 12832 151112 118564 332920 164403 779831 Cell Counts Total 328612 1121077 612290 801933 404265 3268177 0 NA NA NA NA NA NA 1 4.45 4.41 2.32 4.68 3.28 4.04 2 4.06 4.06 3.43 5 3.52 4.07 >2 3.29 2.94 3.24 3.63 4.57 3.64 HBA Total 4.37 4.02 3.1 4.39 3.92 3.96 0 NA NA NA NA NA NA 1 0.48 0.17 0.30 0.03 0.21 0.27 2 0.76 0.62 0.37 0.54 0.33 0.52 >2 0.54 0.81 0.66 0.35 0.35 0.49 HBW Total 0.51 0.51 0.41 0.39 0.32 0.44 0 NA NA NA NA NA NA 1 0.87 1.13 0.29 0.59 0.16 0.79 2 1.34 0.88 0.51 0.95 0.61 0.79 >2 0.00 0.47 1.08 0.61 0.90 0.70 HBS Total 0.89 0.90 0.56 0.75 0.67 0.77 0 NA NA NA NA NA NA 1 0.98 1.02 0.21 0.61 0.69 0.80 2 0.880.670.520.91 0.55 0.69 >2 2.750.500.220.54 0.57 0.53 HBSocRec Total 1.030.760.380.71 0.58 0.68 0 NANANANA NA NA 1 0.000.100.370.68 0.64 0.22 2 0.000.070.220.08 0.59 0.17 >2 0.000.240.140.36 0.91 0.41 HBSCH Total 0.000.100.240.28 0.73 0.24 0 NANANANA NA NA 1 0.430.540.471.71 0.97 0.67 2 0.160.470.971.05 0.53 0.71 >2 0.000.070.220.49 0.75 0.41 HBO Total 0.380.440.690.92 0.68 0.63

PAGE 73

58 Table 3.15 (Continued) Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 NANANANA NA NA 1 0.000.020.020.00 0.00 0.01 2 0.000.030.000.00 0.00 0.01 >2 0.000.000.000.00 0.06 0.01 HBU Total 0.000.020.010.00 0.03 0.01 0 NANANANA NA NA 1 0.660.410.470.19 0.21 0.45 2 0.720.380.180.47 0.25 0.35 >2 0.000.410.560.36 0.33 0.39 NHBW Total 0.640.390.330.38 0.28 0.39 0 NANANANA NA NA 1 1.041.040.190.88 0.39 0.84 2 0.200.940.651.00 0.66 0.84 >2 0.000.440.370.94 0.71 0.69 NHBO Total 0.910.900.480.95 0.64 0.80

PAGE 74

59 Table 3.16 Trip Production Rates of Persons with High Income: India Household Size Trip Purpose Household Car Ownership 1 2 3 4 5+ Weighted Average 0 16122021 60 1+ 0582014 47 Cell Counts Total 111204035 107 0 2.002.002.001.601.52 1.70 1+ NA2.002.002.001.86 1.96 All Trips Total 2.002.002.001.801.66 1.81 0 2.001.332.001.401.43 1.53 1+ NA1.602.001.901.71 1.83 HBW Total 2.001.452.001.651.54 1.66 0 0.000.000.000.000.10 0.03 1+ 0.000.000.000.100.00 0.04 HBSCH Total 0.000.000.000.050.06 0.04 0 0.000.000.000.000.00 0.00 1+ NA0.400.000.000.00 0.04 HBS Total 0.000.180.000.000.00 0.02 0 0.000.670.000.100.00 0.10 1+ NA0.000.000.000.00 0.00 HBSocRec Total 0.000.360.000.050.00 0.06 0 0.000.000.000.100.00 0.03 1+ 0.000.000.000.000.14 0.04 HBO Total 0.000.000.000.050.06 0.04 0 0.000.000.000.000.00 0.00 1+ NA0.000.000.000.00 0.00 NHBW Total 0.000.000.000.000.00 0.00 0 0.000.000.000.000.00 0.00 1+ NA0.000.000.000.00 0.00 NHBO Total 0.000.000.000.000.00 0.00

PAGE 75

60 Table 3.17 Trip Production Rates of P ersons with High Household Income: USA Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 3918000 0 3918 1 1737172599776346257348 0 554504 2 32574993126557355782116 337085 2702256 >2 7039242910611037495103 371140 1727229 Cell Counts Total 217248149601312318541334567 708225 4987907 0 6.82NANANA NA 6.82 1 4.934.224.965.32 NA 4.64 2 5.694.794.424.36 4.68 4.59 >2 2.615.244.394.59 4.08 4.49 HBA Total 5.004.764.434.48 4.37 4.56 0 0.83NANANA NA 0.83 1 0.600.100.330.99 NA 0.38 2 1.210.580.440.38 0.31 0.47 >2 0.000.850.690.83 0.47 0.70 HBW Total 0.680.540.560.57 0.40 0.54 0 0.00NANANA NA 0.00 1 0.850.750.141.55 NA 0.79 2 0.390.930.710.87 0.79 0.84 >2 1.740.460.650.70 0.37 0.58 HBS Total 0.800.820.650.84 0.57 0.75 0 0.17NANANA NA 0.17 1 0.791.211.060.99 NA 1.04 2 1.131.210.731.09 1.08 1.06 >2 0.871.490.900.81 0.82 0.94 HBSocRec Total 0.831.250.830.98 0.95 1.02 0 0.00NANANA NA 0.00 1 0.000.160.310.12 NA 0.12 2 0.000.040.320.25 0.32 0.19 >2 0.000.010.170.34 0.30 0.23 HBSCH Total 0.000.050.250.28 0.31 0.20 0 0.00NANANA NA 0.00 1 0.550.552.090.37 NA 0.70 2 1.190.400.560.74 1.06 0.62 >2 0.000.360.480.57 0.71 0.54 HBO Total 0.620.420.600.66 0.88 0.60

PAGE 76

61 Table 3.17 (Continued) Household Size Trip Purpose Household Vehicle Ownership 1 2 3 4 5+ Weighted Average 0 0.00NANANA NA 0.00 1 0.000.000.000.00 NA 0.00 2 0.000.000.000.00 0.00 0.00 >2 0.000.000.000.00 0.00 0.00 HBU Total 0.000.000.000.00 0.00 0.00 0 4.15NANANA NA 4.15 1 0.960.260.000.86 NA 0.51 2 0.530.540.630.34 0.30 0.47 >2 0.000.710.580.52 0.40 0.54 NHBW Total 0.920.520.570.43 0.35 0.50 0 1.66NANANA NA 1.66 1 1.181.211.030.45 NA 1.10 2 1.251.081.030.69 0.80 0.92 >2 0.001.350.920.80 1.00 0.96 NHBO Total 1.161.140.980.72 0.90 0.96 Table 3.18 and Table 3.19 show household and person trip ra tes respectively by household size and household two-wheeler ownership for the Indian sample. Comparison between the effects of car and two-wheeler ownership in the household/person trip patterns suggests that there are no significant di fferences in trip rates due to availability of car or two-wheelers in th e Indian households.

PAGE 77

62 Table 3.18 Household Trip Production Ra tes by Household Size and Household Two-Wheeler Ownership: India Household Size Trip Purpose Household Two-Wheeler Ownership 1 2 3 4 5+ Weighted Average 0 65368577839 1054 2903 1+ 461122194 221 602 Cell Counts Total 694296991033 1275 3505 0 2.122.704.115.33 6.88 5.25 1+ 2.003.034.515.65 7.02 5.63 All Trips Total 2.122.754.185.39 6.90 5.31 0 1.451.912.392.56 3.11 2.62 1+ 1.002.152.633.04 3.84 3.15 HBW Total 1.421.942.432.65 3.24 2.71 0 0.030.081.122.22 3.24 2.05 1+ 0.000.071.352.05 2.60 1.90 HBSCH Total 0.030.071.162.19 3.12 2.02 0 0.220.320.270.23 0.21 0.24 1+ 0.500.390.200.20 0.21 0.22 HBS Total 0.230.330.260.23 0.21 0.24 0 0.340.220.170.15 0.14 0.16 1+ 0.000.130.160.17 0.21 0.18 HBSocRec Total 0.320.210.170.16 0.15 0.17 0 0.090.140.130.13 0.15 0.14 1+ 0.500.230.110.12 0.13 0.13 HBO Total 0.120.160.130.13 0.14 0.14 0 0.000.020.020.03 0.02 0.02 1+ 0.000.020.050.05 0.03 0.04 NHBW Total 0.000.020.030.03 0.02 0.02 0 0.000.000.010.01 0.00 0.00 1+ 0.000.020.000.02 0.00 0.01 NHBO Total 0.000.000.000.01 0.00 0.01

PAGE 78

63 Table 3.19 Person Trip Production Rates by Household Size and Household TwoWheeler Ownership: India Household Size Trip Purpose Household TwoWheeler Ownership 1 2 3 4 5+ Weighted Average 0 65734172233286027 11876 1+ 41223637791284 2552 Cell Counts Total 69856208541077311 14428 0 2.121.351.371.341.21 1.28 1+ 2.001.521.501.411.21 1.33 All Trips Total 2.121.371.391.351.21 1.29 0 1.450.960.800.640.55 0.64 1+ 1.001.070.880.760.66 0.74 HBW Total 1.420.970.810.660.57 0.66 0 0.030.040.370.560.57 0.50 1+ 0.000.030.450.510.45 0.45 HBSCH Total 0.030.040.390.550.55 0.49 0 0.220.160.090.060.04 0.06 1+ 0.500.200.070.050.04 0.05 HBS Total 0.230.170.090.060.04 0.06 0 0.340.110.060.040.02 0.04 1+ 0.000.070.050.040.04 0.04 HBSocRec Total 0.320.100.060.040.03 0.04 0 0.090.070.040.030.03 0.03 1+ 0.500.110.040.030.02 0.03 HBO Total 0.120.080.040.030.03 0.03 0 0.000.010.010.010.00 0.01 1+ 0.000.010.020.010.00 0.01 NHBW Total 0.000.010.010.010.00 0.01 0 0.000.000.000.000.00 0.00 1+ 0.000.010.000.010.00 0.00 NHBO Total 0.000.000.000.000.00 0.00

PAGE 79

643.8 Trip Distribution This section provides a discussion about distribution of trips by purpose at the trip origin and the trip destination. Trip distribution analyses are performed on the Indian sample as well as the US sample. The comparison of trip distribution patterns between the data sets shed new lights on the differences in trip making patterns between a developed and a developing country context. Figure 3.1 and Figure 3.2 show the trip distribution patterns of the India and the US sample respectively based on the trip purpose at origin. Comparison between the figures suggests that 50 percent of all daily trips start at home in the Indian context while 37 percent of all daily trips originate at home in the US context. It is also found that the percent of trips originating at work or school are significantly higher in case of India compared to the US sample. However, the percentages of trips originating at shopping, social and recreational, and other activity locations are quite high in the US sample as opposed to the India sample. Social/Recreational2%Other1%Shopping2%School19%Work/Work-related business26%Home50% Figure 3.1 Trip Distribution by Purpose at Trip Origin: India

PAGE 80

Social/Recreational16%Other13%Shopping19%School4%Work/Work-related business11%Home37% Figure 3.2 Trip Distribution by Purpose at Trip Origin: USA Figure 3.3 and Figure 3.4 offer the trip distribution patterns at destination of the two data sets. Trip distribution patterns found at the destination are similar to the patterns at origin. Indian trip patterns are typically characterized by higher percentages of trips ending at home, work and school and much lower percent of trips are ending at the destinations with purposes such as shopping, social/recreational and other relative to the US sample. 65

PAGE 81

Home50%Work/Work-related Business26%School19%Shopping2%Other1%Social/Recreational2% Figure 3.3 Trip Distribution by Purpose at Trip Destination: India Home35%Work/Work-related Business11%School4%Shopping19%Other14%Social/Recreational17% Figure 3.4 Trip Distribution by Purpose at Trip Destination: USA Figure 3.5 and Figure 3.6 provides a glimpse of the percent share of trips by purpose of all daily trips reported by the respondents in the surveys. Figure 3.5 suggests that majority of trips reported in the India sample are home-based work trips and home-based educational trips. Home-based shopping, social/recreational and non-home-based 66

PAGE 82

67trips in the Indian context share in combination only about 10 percent of all reported trips while in the US sample, home-based shopping, social, other, and non-home-based trips share a significant portion of all daily trips as opposed to the India sample. HBO2.6%NHBW0.5%HBSCH38.1%HBSocRec3.1%HBS4.5%HBW51.1%NHBO0.1% Figure 3.5 Trip Distribution by Purpose: India NHBW9.0%NHBO20.9%HBO20.5%HBSocRec19.7%HBS18.4%HBW11.5% Figure 3.6 Trip Distribution by Purpose: USA Preliminary analysis of the trip distribution patterns discussed above provides some insights about the differences in trip making patterns in the Indian and US contexts.

PAGE 83

68 The comparative analysis indicates that travel ers in India make just about two trips per day by mainly traveling between home and work or school. Participation in other out-ofhome non-work activities is stil l very limited in India while in the US context, it is found that participation in non-work activities is substantially high as evidenced from large percentages of home-based non-work and non-home based trips. 3.9 Time of Day Distribution Time of day distribution of daily trips is a vital component of tr ansportation planning process. The departure time for a particular trip is an important travel characteristic that indicates why and when a trip is made, which apparently is the manifestation of peoples activity scheduling behavior. Even with wide variation in people s activity scheduling behavior in the population, ther e are significant patterns that emerge in the demand for transportation. These patterns are primarily dependent on various trip making purposes that occur at specific times of a day. Figure 3.7 and Figure 3.8 present the time of day distributi on by trip purpose based on the India and the US data sets respectively. The comparison between the two figures sheds light on many similarities and differences in tempor al aspects of trip making patterns between the two countries. Comp arison between the distributions of trip beginning time of all reported trips in both samples suggest that the trip scheduling pattern in the Indian context is remarkably di fferent than the US context. In the Indian context, it is seen that majority of trips are mainly scheduled dur ing three distinct time periods of a day i.e. morning peak hours, mid-day peak hours and evening peak hours based on various trip purposes. But in the US sample, it is found that the number of trips starts peaking up during the morning peak hours and the tr end continues throughout the day until it reaches its highest point at about 5 p.m. Time of day distributions of HBW trips are consistently bi-modal in nature in both contexts with their prominent peak s during morning and evening rush hours. Distribution of HBO trips is similar to HBW in both samples. The morning and evening

PAGE 84

0246810121416182022240123456789101112131415161718192021222324Time of Day (hr)Percent of Trips All HBW HBS HBSR HBO NHBW NHBO Figure 3.7 Time of Day Distribution by Trip Purpose: India 012345678910111213141501234567891011121314151617181920212223Time of Day (hr)Percent of Trips All HBW HBS HBSR HBO NHBW NHBO Figure 3.8 Time of Day Distribution by Trip Purpose: USA 69

PAGE 85

70 peaks of the HBO distribution may be explai ned by trip chaining on the way to and from work. High percent of home-based school trips may also be a potential contributor to the morning and evening peaks. The sharp midday spike is an additional distinguishing characteristic of the HBO distribution in th e India sample, which basically indicates a mid-day surge of returning ho me trips of students from school. Notably, the distribution of home-based school trips is not generated separately; rather thes e trips are coded as HBO trips in this analysis. Majority of the HBS trips are ma de in the evening hours in the Indian context while HBS trips in the US peak at 10 a.m. and continue into the evening hours. As expected, social and recreationa l (HBSocRec) trips ar e prevalent in both samples during the after-work evening hours. NHBW trips peak during the mid-day hours which are typically characterized by the surge in eat-meal trips during the lunch hours. Other non-home-based trips (NHBO) ar e found to occur after the morning peak and continue for the remainder of the day in the US sample while NHBO trips peak into mid-day and evening hours in the Indian context. 3.10 Modal Split The section offers a comprehensive analys is of survey respondents modal split distribution. The mode choice behavior is in vestigated for both India and US survey respondents by trip purpose, commuting stat us and household car ownership. This study has derived a different set of modes by recodi ng and aggregating the mode types actually reported in the surveys. The definitions adopt ed in this study are reported in Table 3.20. 3.10.1 Modal Split Distribution by Trip Purpose The tables and their corresponding figures included in this section provide the modal split distributions of the Indian and US sample s by trip purpose and time of day. Comparison across the distributions highlights many pronounced differences in mode choice behavior between the counties. Travel ers in India are found to be heavily reliant on public transportation and non-motorized modes while in the US context, people are heavily auto-oriented. This can be e xplained by differences in auto availability and land use

PAGE 86

71 patterns in India and the US. Dense and mixed land use patterns in the Indian urban areas result significant usage of non-motorized modes that typical ly serve shorter-distance Table 3.20 Mode Definitions Mode Definition SOV Single Occupant Vehicle HOV2 High Occupant Vehicle with Occupancy 2 HOV>2 High Occupant Vehicle with Occupancy > 2 Transit City/Public Bus Non-motorized Walk + Bicycle Auto-rickshaw Motorized Three-wheeler Taxi seen in India trips. And, people in India generally captiv e about using public transportation due to unavailability of personal automobiles. On the other hand, disperse land use patterns are common in majority of the US cities and that has made people so dependent private form of transportation and 90 pe rcent of trips are made by automobiles. Modal split distribution of work trips in the US sample indicates that SOV consistently dominates over all other m odes with an average share of 85 percent throughout the day for both home-based and nonhome-based work trips. However, the shares of HOV2, HOV>2 and non-motorized modes are found to increase significantly by 40 to 50 percent compared to HBW trips for shopping, social/recreational, and other trip purposes. These patterns can be expl ained by the fact that maintenance and discretionary activities are jointly undertaken by multiple household members as opposed to the solo nature of work activity partic ipation. The temporal na ture of modal split distribution suggests that the share of high occupant vehicl es increases during the PM peak and evening hours, which is quite expected as the am ount of discretionary travel increase after regular work hours in a weekday. With respect to the India sample, work tr ips are almost equally shared by rail, bus and non-motorized modes. Shopping trips, which mainly include short-distance household errands, are typically served by wa lking and bicycle. But remarkably, the shares of auto-rickshaw and rail are found to have modestly increased by about 10 to 15 percent for social/recreational and other tr ips compared that found in shopping trips. These findings suggest that increase in phys ical distance between home and recreational

PAGE 87

72 or social activity locations may result in grea ter transit usage and desire to participate in those activities jointly with other househol d members increase the tendency of using auto-rickshaw, which provides greater privacy relative to bus/rail and also a cheaper alternative compared to a conve ntional taxi. Auto rickshaws are also considered as major feeder service to major transi t lines in many Indian cities.

PAGE 88

73Table 3.21 Modal Split Distribution of All Trips: India AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent Walk 42.2747.8946.0331.39 30.11 Car 0.851.361.082.73 1.27 Two-wheeler 3.054.544.126.27 4.58 Bicycle 3.022.332.462.46 2.21 Bus 22.7719.7919.4120.78 21.46 Rickshaw 5.037.926.427.50 5.34 Rail 22.0914.9119.7327.96 33.84 Carpool 0.730.860.450.54 0.85 Other 0.170.350.290.27 0.25 Missing 0.020.040.030.11 0.08 Total 100.00100.00100.00100.00 100.00 Sample size 4133763737871867 1179 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00 pm10:00 pm 6:00 amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips Walk Car Scooter/motorcycle Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.9 Modal Split Distributions of All Trips: India

PAGE 89

74Table 3.22 Modal Split Distribution of All Trips: USA AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:3018:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent SOV 45.2343.9438.9728.07 45.49 HOV2 23.5928.3431.8432.94 25.90 HOV>2 14.7116.6717.8526.89 18.36 Transit 1.801.410.460.57 0.55 School 5.641.630.050.00 0.00 Nonmotorized 8.007.209.6910.00 7.90 Other 0.680.741.071.53 1.54 Missing 0.340.060.070.00 0.25 Total 100.00100.00100.00100.00 100.00 Sample size 751652128716174107691757563110 2557442 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00 pm10:00 pm 6:00 amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips SOV HOV 2 HOV > 2 Transit School Bus Non-motorized Other Missing Figure 3.10 Modal Split Distributions of All Trips: USA

PAGE 90

75Table 3.23 Modal Split Distribution of HBW Trips: India AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent Walk 26.5932.0625.4826.54 22.72 Car 1.302.651.973.16 1.69 Two-wheeler 4.909.176.827.25 5.62 Bicycle 3.953.163.362.73 2.47 Bus 25.1121.2524.0920.09 23.17 Rickshaw 4.857.594.645.60 4.84 Rail 32.8823.2233.1033.86 39.26 Carpool 0.130.290.050.43 0.11 Other 0.220.580.480.36 0.11 Missing 0.040.030.000.00 0.00 Total 100.00100.00100.00100.00 100.00 Sample size 2226309718761394 889 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00 pm10:00 pm 6:00 amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips Walk Car Scooter/motorcycle Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.11 Modal Split Distributions of HBW Trips: India

PAGE 91

76Table 3.24 Modal Split Distribution of HBW Trips: USA AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent SOV 79.5984.0980.1980.14 84.50 HOV2 12.095.4110.066.70 10.23 HOV>2 2.871.653.491.54 5.27 Transit 2.844.783.0811.35 0.00 School 0.000.000.000.00 0.00 Nonmotorized 1.563.562.760.27 0.00 Other 0.480.260.000.00 0.00 Missing 0.550.250.420.00 0.00 Total 100.00100.00100.00100.00 100.00 Sample size 205644121191001424495361799 632302 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00pm10:00 pm 6:00amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips SOV HOV 2 HOV > 2 Transit School Bus Non-motorized Other Missing Figure 3.12 Modal Split Distributions of HBW Trips: USA

PAGE 92

77Table 3.25 Modal Split Distribution of HBS Trips: India AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent Walk 73.9049.7072.8065.80 58.80 Car 0.000.000.001.10 0.00 Two-wheeler 0.002.631.720.50 0.00 Bicycle 0.001.320.690.00 23.50 Bus 8.7020.0713.1020.30 17.70 Rickshaw 8.7019.089.6611.20 0.00 Rail 8.706.581.720.50 0.00 Carpool 0.000.660.340.50 0.00 Other 0.000.000.000.10 0.00 Missing 0.000.000.000.00 0.00 Total 100.00100.03100.04100.00 100.00 Sample size 46304290187 17 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00 pm10:00 pm 6:00 amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips Walk Car Scooter/motorcycle Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.13 Modal Split Distributions of HBS Trips: India

PAGE 93

78Table 3.26 Modal Split Distribution of HBS Trips: USA AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent SOV 69.8250.8546.0830.89 51.13 HOV2 16.5331.3637.2829.84 34.42 HOV>2 8.9311.2810.9136.80 9.01 Transit 0.001.650.000.00 0.00 School 0.000.000.000.00 0.00 Nonmotorized 4.714.255.232.46 5.44 Other 0.000.610.500.00 0.00 Missing 0.000.000.000.00 0.00 Total 100.00100.00100.00100.00 100.00 Sample size 517120630198720802671461482 124108 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00 pm10:00 pm 6:00 amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips SOV HOV 2 HOV > 2 Transit School Bus Non-motorized Other Missing Figure 3.14 Modal Split Distributions of HBS Trips: USA

PAGE 94

79Table 3.27 Modal Split Distribution of HBSocRec Trips: India AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent Walk 35.1024.9045.0023.08 10.70 Car 0.002.720.01.71 0.00 Two-wheeler 8.203.505.006.84 3.57 Bicycle 0.000.780.710.85 0.0 Bus 29.7024.1217.1426.50 25.00 Rickshaw 0.0016.7322.8623.93 17.86 Rail 21.6024.126.4313.68 28.57 Carpool 5.401.171.432.56 7.14 Other 0.001.950.710.00 7.14 Missing 0.000.000.710.85 0.02 Total 100.00100.00100.00100.00 100.00 Sample size 37257140117 28 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00 pm10:00 pm 6:00 amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips Walk Car Scooter/motorcycle Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.15 Modal Split Distributions of HBSocRec Trips: India

PAGE 95

80Table 3.28 Modal Split Distribution of HBSocRec Trips: USA AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent SOV 29.3125.6021.2121.28 27.44 HOV2 24.1927.9131.0434.69 27.50 HOV>2 5.9421.5319.0620.59 28.11 Transit 5.210.810.160.08 0.00 School 0.000.000.000.00 0.00 Nonmotorized 31.9721.4325.8419.61 14.76 Other 3.372.722.693.74 2.18 Missing 0.000.000.000.00 0.00 Total 100.00100.00100.00100.00 100.00 Sample size 787221415849326544482827857 830484 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00pm10:00 pm 6:00amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips SOV HOV 2 HOV > 2 Transit School Bus Non-motorized Other Missing Figure 3.16 Modal Split Distributions of HBSocRec Trips: USA

PAGE 96

81Table 3.29 Modal Split Distribution of HBO Trips: India AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent Walk 33.3029.8441.6722.90 39.30 Car 0.000.783.10.00 0.00 Two-wheeler 8.892.336.254.17 0.00 Bicycle 2.221.163.138.35 0.00 Bus 28.8922.8711.4618.75 14.30 Rickshaw 6.7024.4219.7918.75 17.90 Rail 20.0017.8314.5827.08 28.50 Carpool 0.000.780.000.00 0.00 Other 0.000.000.000.00 0.00 Missing 0.000.000.000.00 0.00 Total 100.00100.00100.00100.00 100.00 Sample size 452589648 28 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00 pm10:00 pm 6:00 amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips Walk Car Scooter/motorcycle Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.17 Modal Split Distributions of HBO Trips: India

PAGE 97

82Table 3.30 Modal Split Distribution of HBO Trips: USA AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18:30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent SOV 17.8429.6822.2530.37 40.59 HOV2 32.6930.5134.8739.70 39.03 HOV>2 26.1625.5234.7423.50 14.78 Transit 0.952.180.000.00 0.00 School 12.386.100.140.00 0.00 Nonmotorized 9.195.867.156.44 4.35 Other 0.510.160.760.00 0.00 Missing 0.290.000.090.00 1.25 Total 100.00100.00100.00100.00 100.00 Sample size 1513333399996215670941032298 463419 0%10%20%30%40%50%60%70%80%90%100%6:00 am 8:30 am8:30 am 4:00 pm4:00 pm 6:30 pm6:30 pm 10:00pm10:00 pm 6:00amAM PeakMid DayPM PeakEveningOver Night Time of DayPercent Trips SOV HOV 2 HOV > 2 Transit School Bus Non-motorized Other Missing Figure 3.18 Modal Split Distributions of HBO Trips: USA

PAGE 98

83 Table 3.31 Modal Split Distribution of NHBW Trips: India AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18: 30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent Walk 20.0027.0835.7053.33 33.30 Car 0.0012.500.06.67 66.70 Two-wheeler 0.0018.7514.306.67 0.00 Bicycle 0.006.2514.306.67 0.00 Bus 40.0014.587.106.67 0.00 Rickshaw 20.004.1714.306.67 0.00 Rail 20.0016.6714.3013.33 0.00 Carpool 0.000.000.000.00 0.00 Other 0.000.000.000.00 0.00 Missing 0.000.000.000.00 0.00 Total 100.00100.00100.00100.00 100.00 S am ple size 5481415 3 Table 3.32 Modal Split Distribution of NHBW Trips: USA AM Peak Mid Day PM Peak Evening Over Night 6:00-8:30 8:30-16:00 4:00-18: 30 18:30-22:00 22:00-6:00 Mode Percent Percent Percent Percent Percent SOV 78.9674.2580.8059.39 57.04 HOV2 7.4413.9715.8711.68 20.67 HOV>2 13.006.522.6110.51 12.82 Transit 0.000.330.000.00 9.47 School 0.110.220.000.00 0.00 Nonmotorized 0.494.130.3015.36 0.00 Other 0.000.250.423.06 0.00 Missing 0.000.320.000.00 0.00 Total 100.00100.00100.00100.00 100.00 Sample size 6327783367913749219222715 149441

PAGE 99

84 Table 3.33 Modal Split Distribution of NHBO Trips: India AM Peak Mid Day PM Peak Evening Over Night Mode 6:00-8:30 8:30-16:00 4:00-18: 30 18:30-22:00 22:00-6:00 Walk 100.0033.3050.0025.00 100.00 Car 0.000.000.050.00 0.00 Two-wheeler 0.000.0016.7025.00 0.00 Bicycle 0.000.000.000.00 0.00 Bus 0.0016.7016.700.00 0.00 Rickshaw 0.000.0016.600.00 0.00 Rail 0.0033.300.000.00 0.00 Carpool 0.000.000.000.00 0.00 Other 0.0016.700.000.00 0.00 Missing 0.000.000.000.00 0.00 Total 100.00100.00100.00100.00 100.00 S a m p l e s i z e 1664 1 Table 3.34 Modal Split Distribution of NHBO Trips: USA AM Peak Mid Day PM Peak Evening Over Night Mode 6:00-8:30 8:30-16:00 4:00-18: 30 18:30-22:00 22:00-6:00 SOV 24.6233.4025.7519.75 13.66 HOV2 40.5237.5046.1836.95 30.75 HOV>2 18.2921.0823.3939.35 31.81 Transit 1.290.540.080.00 0.00 School 10.391.820.150.00 0.00 Nonmotorized 3.975.153.623.75 16.66 Other 0.000.510.830.20 7.12 Missing 0.920.000.000.00 0.00 Total 100.00100.00100.00100.00 100.00 Sample size 640171744790119512261595473 299870 Note: The corresponding graphs of these tables are not provided because the sample sizes are too small in the Thane distributions.

PAGE 100

85 3.10.2 Modal Split Distribution by Commu ting Status by Trip Purpose Modal split distribution for commuter a nd non-commuter samples by trip purpose are investigated in this section for the two data sets. The modal split distributions obtained for the Indian commuter sample are presented by Table 3.35 and Figure 3.19. The corresponding distributions for the US co mmuter sample are shown in Table 3.36 and Figure 3.20. It was felt that travel behavior of school going ch ildren are likely to be quite different than adults because children ofte n do not have the freedom to choose their travel patterns, have less flexibility with re spect to their travel options and decisions. Therefore, the modal split distributions of sc hool-age (< 16 years) children samples are analyzed and compared in this section as well. It should be noted that school-age children samples are defined in the Indian and US contexts with a litt le difference. In the Indian context, all individuals who re ported at least one school trip but did not report any work trip on the survey day and with 15 years of ag e or less are treated as school-age children and in the US sample, everyone who falls le ss than 16 years of age is treated as the school-age children sample. It is considered that such definitional differences would not affect the interest of this research too mu ch and will nevertheless offer useful insights into the potential differences in mode choice behavior between the two student samples. The mode choice patterns of the commuter sample of India are shown in Table 3.35 and the corresponding graphical distributi on is shown in Figure 3.19. Majority of the work trips reported by Indian commuters are primarily served by walk (28%), bus (23%) and rail (30%) and to a lesser extent by two-wheelers like scooters and motorcycles (7%). Shares of bus and rail are consistently pr evalent in all trip pu rposes. Though the number of non-work trips reported by Indian commuters are very limited stil l the mode choice behavior for non-work trip purposes are quite different than work mode choice pattern. It can be see in the Table 3.35 that the percentage of walk mode is si gnificantly dropped to 10 percent for home-based social or recreat ional trips while share of auto-rickshaw increases to 28 percent for HBS trips and 17 percent for HBSocRec trips. The percentages of HBO and NHBW trips made by personal vehicles like bicycles, twowheelers and cars are higher relative to other trip categories.

PAGE 101

86 As expected, the percent trips made by cars are dominant in all trip purposes in the US context. Large percents of HBW a nd NHBW trips are made by SOV at the range of about 80 percent and 75 pe rcent respectively. Shares of HBV and HOV>2 increase significantly for non-work trips. Non-motorized mode has its highest share at about 21 percent for home-based social and recreational trips. Table 3.37 provides modal split distribu tion for Indian non-commuter sample. Comparison of mode choice patterns betw een Indian commuter and non-commuter samples show that about 45 percent of all trips reported by non-commuters are made by walk while only 27 percent walk trips reporte d by commuters. Similarly, the percent of all reported trips made by auto-ricksha w is about 10 percent higher for the noncommuters than the commuter sample. However, transit shares among the non-commuter sample for both bus and rail are lower than that of the commuter sample; especially the difference in rail share is much pronounced at about 19 percent. These differences are consistently found in all of the trip purposes. The mode choice distribution of the US non-commuter sample is shown in Table 3.38. The percent of trips made by SOV is c onsiderably lower in non-commuter sample (28 percent) compared to commuter sample (63 percent). At the same time, the shares of HOV2 and HOV>2 are reported at much highe r proportions of about 34 and 23 percent respectively in the non-commuter sample. Th e mode choice patterns between commuter and non-commuter market groups are also di ffered by increased share of non-motorized modes among non-commuters samples. Ag ain, these differences are followed consistently in all of th e trip purposes. School-age student samples extracted from both the data sets have shown significant differences in their mode choice patterns. Comparison between Table 3.39 and Table 3.40 shows that majority (77 percent) of the trips (school trips) reported by Indian school-age children are walk trips while in th e US sample, it is found that 75 percent of trips reported by child ren are shared by HOV2 and HOV>2 modes. These mode choice patterns are quite obvious in both contexts b ecause walking to school located within close proximity from home is very common among Indian school children while in the US context, most children are driven by thei r parents to school an d other activities.

PAGE 102

87Table 3.35 Modal Split Distribution of Commuter Sample by Trip Purpose: India All HBW HBS HBSocRecHBO NHBW NHBO Mode Percent PercentPercentPercent PercentPercent Percent Walk 27.82 27.8342.9010.3015.4032.90 0.00 Car 2.18 2.130.000.003.808.20 0.00 Scooter 7.11 7.050.000.0015.4014.10 0.00 Bicycle 3.29 3.250.000.0011.507.10 0.00 Bus 22.63 22.7128.6027.6023.1012.90 0.00 Rickshaw 5.88 5.7628.5017.2015.409.50 0.00 Rail 30.43 30.640.0037.9015.4015.30 0.00 Carpool 0.23 0.210.007.000.000.00 0.00 Other 0.39 0.400.000.000.000.00 0.00 Missing 0.02 0.020.000.000.000.00 0.00 Total 100.00 100.00100.00100.00100.00100.00 0.00 N 9621 945514292685 0.00 0%10%20%30%40%50%60%70%80%90%100%All TripsHBWHBSHBSocRecHBONHBWTrip by PurposePercent of Trips Walk Car Scooter Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.19 Modal Split Distribution of Commuter Sample by Trip Purpose: India

PAGE 103

88Table 3.36 Modal Split Distribution of Commuter Sample by Trip Purpose: USA All HBW HBS HBSocRec HBO NHBW NHBO Mode Percent Percent Percent Percent Percent Percent Percent SOV 63.35 81.6758.0539.93 38.3874.64 39.10 HOV2 20.23 9.0326.4527.18 38.8213.54 35.93 HOV>2 9.29 2.7713.269.46 18.267.11 19.93 Transit 1.35 3.710.000.00 0.490.49 0.00 Schoolbus 0.11 0.000.000.00 0.610.16 0.00 Nonmotor 4.90 2.242.1821.50 2.963.49 3.44 Other 0.58 0.240.061.93 0.350.36 1.35 Missing 0.20 0.350.000.00 0.130.21 0.25 Total 100.00 100.00100.00100.00 100.00100.00 100.00 N 23142303 748828023727202724669 19964575525208 2725725 0%10%20%30%40%50%60%70%80%90%100%All TripsHBWHBSHBSocRecHBONHBWNHBOTrip by PurposePercent of Trips SOV HOV2 HOV>2 Transit School Bus Non-motorized Other Missing Figure 3.20 Modal Split Distributions of Commuter Sample by Trip Purpose: USA

PAGE 104

89Table 3.37 Modal Split Distribution of Non-commuter Sample by Trip Purpose: India All HBS HBSocRecHBO NHBO Mode PercentPercentPercent Percent Percent Walk 45.8863.0130.4433.71 36.40 Car 0.870.241.480.90 18.20 Scooter 3.091.694.982.70 18.20 Bicycle 1.191.200.741.80 0.00 Bus 19.6616.8723.4320.22 9.10 Rickshaw 16.5212.8918.6321.35 0.00 Rail 11.383.6116.9718.88 9.10 Carpool 0.760.481.480.45 0.00 Other 0.490.001.480.00 9.00 Missing 0.160.000.370.00 0.00 Total 100.00100.00100.00100.00 100.00 N 1846830542445 11 0%10%20%30%40%50%60%70%80%90%100%All TripsHBSHBSocRecHBONHBOTrip by PurposePercent of Trips Walk Car Scooter Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.21 Modal Split Distribution of Non-Commuter Sample by Trip Purpose: India

PAGE 105

90Table 3.38 Modal Split Distribution of Non-Commuter Sample by Trip Purpose: USA All HBS HBSocRecHBO NHBO Mode Percent Percent Percent Percent Percent SOV 28.4445.4719.7123.19 26.99 HOV2 33.8232.9730.8131.61 39.54 HOV>2 23.0914.9722.8728.47 25.02 Transit 1.011.250.911.39 0.53 Schoolbus 2.410.000.007.08 2.14 Nonmotor 10.054.7722.467.80 5.30 Other 1.140.573.250.31 0.49 Missing 0.040.000.000.14 0.00 Total 100.00100.00100.00100.00 100.00 N 30506861822675874094615469218 8561774 0%10%20%30%40%50%60%70%80%90%100%All TripsHBSHBSocRecHBONHBOTrip by PurposePercent of Trips SOV HOV2 HOV>2 Transit School Bus Non-motorized Other Missing Figure 3.22 Modal Split Distributions of Non-commuter Sample by Trip Purpose: USA

PAGE 106

91Table 3.39 Modal Split Distribution of School-Age Children Sample (Age <16 years) by Trip Purpose: India All HBSCHHBS HBSocRecHBO NHBO Mode Percent Percent PercentPercent Percent Percent Walk 77.65 77.720.000.000.00 33.30 Car 0.22 0.200.0025.000.00 0.00 Scooter 0.13 0.110.0025.000.00 0.00 Bicycle 1.40 1.400.000.000.00 0.00 Bus 11.61 11.620.000.000.00 33.30 Rickshaw 6.23 6.230.000.000.00 33.30 Rail 1.09 1.100.000.000.00 0.00 Carpool 1.51 1.470.0050.000.00 0.00 Other 0.10 0.100.000.000.00 0.00 Missing 0.05 0.050.000.000.00 0.00 Total 100.00 100.000.00100.000.00 100.00 N 4572.00 4561.000.004.000.00 3.00 0%10%20%30%40%50%60%70%80%90%100%All TripsHBSCHHBSocRecNHBOTrip by PurposePercent of Trips Walk Car Scooter Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Table 3.23 Modal Split Distribution of School-Age Children Sample (Age<16 years) by Trip Purpose: India

PAGE 107

92Table 3.40 Modal Split Distribution of School-Age Children Sample (Age <16 years) by Trip Purpose: USA All HBS HBSocRecHBO* NHBO Mode PercentPercentPercentPercent Percent SOV 0.060.000.000.00 0.26 HOV2 23.7829.8714.8226.43 25.53 HOV>2 50.1167.1143.9645.26 56.81 Transit 0.630.001.340.48 0.46 Schoolbus 9.420.000.0017.23 10.45 Nonmotor 14.743.0236.149.98 6.11 Other 1.200.003.730.47 0.39 Missing 0.060.000.000.15 0.00 Total 100.00100.00100.00100.00 100.00 N 9898157118435324130942883192 2211632 Home-based school trips and Other trip purposes are coded as HBO 0%10%20%30%40%50%60%70%80%90%100%All TripsHBSHBSocRecHBONHBOTrip by PurposePercent of Trips SOV HOV2 HOV>2 Transit School Bus Non-motorized Other Missing Table 3.24 Modal Split Distribution of School-Age Children Sample (Age <16 years) by Trip Purpose: USA

PAGE 108

93 3.10.3 Modal Split Distribution by Household Car Ownership Availability of automobiles in a household is considered as an important factor that may have potential effect s on mode choice pattern of th e household members. To understand the nature of any possible underl ying relationship, modal spl it distributions of the India and US samples are further investigated based on household car ownership status. Household car ownership levels are classified into two groups in the Indian context: zero car households and non-zero car households. Due to high level of car ownership among US households, four classificati ons are used for the US samp le: zero car, one car, two cars and three or more cars households. Table 3.41 and Table 3.42 present a glimps e of modal split di stribution for trips reported by zero-car and non-zero-car household members for the India sample. Comparison between these two tables suggests that availability of cars in the Indian households potentially increases car mode share by about 20 percent on average for all the trip purposes. Particular ly, HBW and NHBW work trips s how the highest shares of private car mode at about 31 percent and 50 percent respecti vely. As expected, overall non-motorized mode share of car owned household trips is 20 percent lower than that of zero-car households. However, th e mode share of two-wheelers is consistently higher for all trip purposes reported by car owned households but to a lesser extent compared to what seen in their car shares. Overall trans it share does not seem to differ much between withand without-car household trips as it can be seen that the percent drops among bus, auto-rickshaw and rail shares are within th e range of 5 percent for with-car household trips. However, with respect to non-zero-car households, the percen t of HBW trips made by bus and rail trips are found as much as 10 pe rcent lower than zero-car households. Table 3.43 through Table 3.46 show the modal split distribution of the US sample based on household car ownership status. As expected, majority of the trips reported by the zero-car households are made by trans it and non-motorized modes. The percentages of total transit (20%) and non-motorized (40% ) trips reported by zer o-car households are about 18 percent and 30 percent higher than the corresponding shares reported by one-car households. These differences continue to incr ease consistently for all trip purposes with the level of car ownership in the US house holds. Personal vehicle trips are prevalent

PAGE 109

94 among households with cars. The highest share of SOV trips is repo rted by the three or more car household sample. However, the two-car household sample has reported the highest percent of HOV trips (HOV2 + HOV>2) at about 50 percen t of all trips as opposed to 44 percent and 43 percent repor ted by one-car and three or more car households. The shares of non-motorized mode for home-based social/recreational trips remain remarkably high in the range of 17 to 23 percent among car-owned households. However, zero-car households have reported the highest share of non-motorized mode (67%) for social/recreational trip purpose. Notably, the percent of HOV2 trips (22%) reported by zero-car households is significan tly higher compared to other personal vehicle trips (SOV: 5% and HOV>2: 6.5%) that reported by the same household sample.

PAGE 110

95Table 3.41 Modal Split Distribution of Zero-Car Household Members by Trip Purpose: India All HBW HBSCH HBS HBSocRec HBO NHBW NHBO Mode Percent Percent Percent Percent Percent Percent Percent Percent Walk 44.53 28.37 65.9163.00 30.0933.00 38.40 43.80 Car 0.26 0.42 0.040.00 0.730.00 1.40 0.00 Scooter 3.99 6.70 0.551.20 4.593.70 12.30 12.50 Bicycle 2.62 3.45 1.771.20 0.732.40 8.20 18.60 Bus 20.88 23.33 17.8517.60 24.7720.00 12.30 0.00 Rickshaw 6.68 5.95 4.8913.00 18.7220.90 11.00 6.30 Rail 20.10 31.18 7.643.50 17.4319.60 16.40 12.50 Carpool 0.69 0.20 1.270.50 2.200.40 0.00 0.00 Other 0.22 0.36 0.060.00 0.370.00 0.00 6.30 Missing 0.03 0.30 0.010.00 0.370.00 0.00 0.00 Total 100.00 100.27 100.00100.00 100.00100.00 100.00 100.00 N 17584 8945 6705808 545460 73 18 0%5%10%15%20%25%30%35%40%45%50%55%60%65%70%75%80%85%90%95%100%All TripsHBWHBSCHHBSHBSocRecHBONHBWNHBOTrip by PurposePercent of Trips Walk Car Scooter Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.25 Modal Split Distribution of Zero-Car Household Members by Trip Purpose: India

PAGE 111

96Table 3.42 Modal Split Distribution of Non-Zero-Car Household Members by Trip Purpose: India All HBW HBSCH HBS HBSocRec HBO NHBW NHBO Mode Percent Percent Percent Percent Percent Percent Percent Percent Walk 25.60 17.90 36.0155.6017.6013.30 0.00 0.00 Car 19.25 30.60 2.595.5014.7033.30 50.00 100.00 Scooter 9.53 13.40 3.6311.108.806.70 25.00 0.00 Bicycle 0.77 0.00 2.070.000.000.00 0.00 0.00 Bus 15.50 13.00 20.985.600.0026.70 16.70 0.00 Rickshaw 8.08 3.40 12.9516.6017.6020.00 0.00 0.00 Rail 18.67 20.30 18.395.6023.500.00 8.30 0.00 Carpool 1.15 0.40 2.590.000.000.00 0.00 0.00 Other 1.35 1.00 0.520.0017.800.00 0.00 0.00 Missing 1.00 0.00 0.260.000.000.00 0.00 0.00 Total 100.90 100.00 100.00100.00100.00100.00 100.00 100.00 N 1039 553 386363415 12 2 0%5%10%15%20%25%30%35%40%45%50%55%60%65%70%75%80%85%90%95%100%All TripsHBWHBSCHHBSHBSocRecHBONHBWTrip by PurposePercent of Trips Walk Car Scooter Bicycle Bus Auto Rickshaw Rail Carpool Other Missing Figure 3.26 Modal Split Distribution of Non-Zero-Car Household Members by Trip Purpose: India

PAGE 112

97Table 3.43 Modal Split Distribution of Zero-Car Household Members by Trip Purpose: USA All HBW HBS HBSocRec HBO NHBW NHBO Mode Percent Percent Percent Percent Percent Percent Percent SOV 4.56 5.604.882.873.53 7.62 4.10 HOV2 22.22 2.6629.546.8229.83 39.57 23.53 HOV>2 6.50 23.521.917.614.60 6.24 4.07 Transit 19.60 57.0522.025.4410.24 0.00 9.38 Schoolbus 2.54 0.000.000.0018.04 0.00 3.55 Nonmotor 39.90 11.1740.8667.0730.93 46.57 39.61 Other 4.68 0.000.7910.192.83 0.00 15.77 Missing 0.00 0.000.000.000.00 0.00 0.00 Total 100.00 100.00100.00100.00100.00 100.00 100.00 N 898465 1973553335946995499950 74715 122897 0%10%20%30%40%50%60%70%80%90%100%All TripsHBWHBSHBSocRecHBONHBWNHBOTrip by PurposePercent of Trips SOV HOV2 HOV>2 Transit School Bus Non-motorized Other Missing Figure 3.27 Modal Split Distributions of Zero-Car Household Members by Trip Purpose: USA

PAGE 113

98Table 3.44 Modal Split Distribution of One-Car Household Members by Trip Purpose: USA All HBW HBS HBSocRec HBO NHBW NHBO Mode Percent Percent Percent Percent Percent Percent Percent SOV 41.40 68.1254.0227.5827.32 69.61 33.81 HOV2 31.51 13.7033.5734.1130.04 13.74 42.69 HOV>2 12.67 3.326.9812.0922.46 8.30 14.79 Transit 1.71 7.950.031.222.65 1.11 0.49 Schoolbus 2.12 0.000.000.008.29 0.00 1.88 Nonmotor 9.89 6.425.2023.138.54 6.71 6.15 Other 0.55 0.490.211.870.25 0.54 0.00 Missing 0.13 0.000.000.000.46 0.00 0.18 Total 100.00 100.00100.00100.00100.00 100.00 100.00 N 23301963 1470884465526942540875198857 1689905 6032962 0%10%20%30%40%50%60%70%80%90%100%All TripsHBWHBSHBSocRecHBONHBWNHBOTrip by PurposePercent of Trips SOV HOV2 HOV>2 Transit School Bus Non-motorized Other Missing Figure 3.28 Modal Split Distributions of One-Car Household Members by Trip Purpose: USA

PAGE 114

99Table 3.45 Modal Split Distribution of Two-Car Household Members by Trip Purpose: USA All HBW HBS HBSocRec HBO NHBW NHBO Mode Percent Percent Percent Percent Percent Percent Percent SOV 38.69 85.8744.15 18.7324.1377.91 27.22 HOV2 30.00 8.7834.06 32.6534.4814.26 37.06 HOV>2 22.29 2.6519.82 23.8331.066.96 30.72 Transit 0.26 1.150.00 0.000.600.00 0.00 Schoolbus 1.24 0.000.00 0.004.130.06 1.70 Nonmotor 6.84 1.351.29 22.545.570.57 3.00 Other 0.68 0.200.68 2.240.000.25 0.30 Missing 0.01 0.000.00 0.000.030.00 0.00 Total 100.00 100.00100.00 100.00100.00100.00 100.00 N 47582242 32291568849203 9493308115280032942105 11540468 0%10%20%30%40%50%60%70%80%90%100%All TripsHBWHBSHBSocRecHBONHBWNHBOTrip by PurposePercent of Trips SOV HOV2 HOV>2 Transit School Bus Non-motorized Other Missing Figure 3.29 Modal Split Distributions of Two-Car Household Members by Trip Purpose: USA

PAGE 115

Table 3.46 Modal Split Distributio by Trip Purpose: USA 100n of Three or More Car Household Trips All HBW HBS HBSocRec HBO NHBW NHBO Mode Percent Percent Percent Percent Percent Percent Percent SOV 48.36 90.2957.5231.23 29.75 76.95 29.76 HOV2 24.52 6.8223.5721.81 33.55 11.22 38.64 HOV>2 17.17 1.0917.3222.75 23.06 6.39 23.74 Transit 0.58 0.510.001.27 0.40 0.72 0.53 Schoolbus 1.50 0.000.000.00 5.85 0.44 1.46 Nonmotor 6.14 0.091.2817.71 6.46 3.20 4.81 Other 1.50 0.140.325.23 0.94 0.39 1.07 Missing 0.23 1.050. 000.00 0.00 0.69 0.00 Total 100.00 100.00100.0 0100.00 100.00 100.00 100.00 N 23319272 23303243516007 4585646 5011223 1859240 6016833 0%10%20%30%40%50%60%70%80%90%100%All TripsHBWHBSTriPercent of Trips HBSocRecHBONHBWNHBOp by Purpose SOV HOV2 HOV>2 Transit School Bus Non-motorized Other Missing Three or More Car Household Trips Figure 3.30 Modal Split Distributions ofby Trip Purpose: USA

PAGE 116

3.11 Activity and Time Use Characteristics This section offers activity and tim is an increasing recognition in the profe patterns on individual travel behavior. The m com survey sam people trade off their tim Quantitative analyses are perform of trip frequency, travel duration and ac underlying behavioral heteroge colle exploration of activity and tim on their commuting status. m student age tim travel characteristics such as trip fre However, subsequent ch much greater details. Table 3.48 provides th for the US sam this an Another notable difference is divided into two groups base Those who reported at least one school tr commuters (age reasons behind this segm of student respondents in the sam in travel behavior between children and a commuter sam 101 e use char acteristics of the two survey samples. There ssion about the role of activity and time use ain interest of this sect ion is to explore and pare activity and time use patterns of vari ous market segments derived from different ples. It is considered that such analysis can offer valuable insights about how e into various mandatory and non-mandatory activities. ed to understand such behavioral aspects in the context tivity duration by purpose. Considering the neity in the population and various constraints that ctively dictate peoples activity-travel patterns, the emphasis has been placed into e use behavior with respect to various market groups based Table 3.47 offers activity and time use char acteristics of the India sample. Four ajor market groups were derived from the Indian survey sample: student age 15 years, 16 years, commuter and non-commuters. As mentioned earlier, activity and e use behavior of these sample groups are explored with respect to certain activityquency, travel duration and activity duration. apters have dealt with additional topics on time use behavior in e corresponding activity-t ravel characteristics ple groups. The definitions of the different marker segments included in alysis are consistent with the definitions that were used in the previous sections. that adult (age > 15 years) non-commuters are further d on the reporting of any school trip by the respondents. ip is further categorized as student non16 yr) and the rest is treated simp ly as non-commuters. The main entation of the Indian non-commuter sample are: large percent ple and sec ondly, consideration of expected variability dult student samples. However, the US nonple includes all adults who did not report any work or work related

PAGE 117

102 business trip but reported at le ast one non-work trip on the survey day. It should be noted that, this analysis focuses on only mobile indi viduals who reported at least one trip on the survey day, thus eliminating the effect of significant presence of zero-trip makers in both sample survey samples. If the zero-travel individuals are included in the sample, the averages would drop dramatically. The second column of each market segment shown in both tables provides the averages that were calculated based on non-zero observations of a particular activity. This treatment especi ally provides the averages for only those individuals who actually participated in that particular activity. 3.11.1 Trip Frequency Analysis Trip frequency distributions by trip purposes of various market segments are provided in Table 3.47 and Table 3.48 for the India and US samples respectively. The average trip frequency reported by the US sample groups are much higher than those reported by India sample groups. In the Indian context, each market segment makes, on average, two trips per day while in the US context average number of trips reporte d is about 4 per day. The average trip frequency reported by the In dian commuters is one-half as many trips reported by the US commuter sample. The av erage number of HBW trips reported by the Indian commuter sample is only about 2 per day but the number of non-work or nonhome-based trips reported is extremely lo w indicating that th e Indian commuters generally go to work (or work-related busines s) and then return home. Non-commuters in the US make nearly five trips per day; the corresponding figure for the Indian noncommuter sample is about 2 trips per day. The same trip pattern is also followed by both of the Indian school-age stude nt samples. Average school-age children in the US sample makes about one trip more than the number of daily trips reported by the Indian school age children sample. Trip rates reported by th e Indian market segments are evident of the fact that Indian travelers undertake, on average one out-o f-home activity per day (and then return home). As opposed to Indian sample, higher trip frequency reported by the US sample indicates much higher level of activity participation by US market groups. Much of these differences can be attributed to the higher dependence on transit and nonmotorized modes in the Indian context th at allow people to undertake much fewer

PAGE 118

103 multistop trip chaining than th e case in the US. However, the non-work trips (e.g. HBS, HBSocRec, HBO, and NHBO) reported by non-commuters are consistently higher than that reported by commuter sample for both coun tries. These differences are quite obvious as non-commuters are free of any work-related constraints that comm uters encounter in the course of their daily life. Therefore, they have more flexibility to pursue greater amount of discretionary or other non-manda tory activities than that available for commuters. 3.11.2 Analysis of Travel Duration Average travel time durations by purpose for both survey samples are given in Table 3.47 and Table 3.48. The average travel time dura tion for the India sample is 55 minutes in comparison to 95 minutes in the US sample. Indian commuters spend, on average 73 minutes for daily travel in which 72 minut es are reported to spend only toward HBW travel. Adult student sample in India shows the second highest daily travel time expenditure at 52 minutes followed by non-co mmuters (45 minutes) and children student sample (26 minutes). Similar to the Indian co ntext, the commuter sample in the US has the highest level of daily travel duration with 105 minutes followed by non-commuter (90 minutes) and school-age children sample (67 minutes). As evident from the comparison between the two survey samples, major diffe rences between daily average travel time durations between the two countries can be at tributed to higher rate of non-work activity participations among US respondents. For example, US commuter sample, on average spend about 35 minutes in total for all nonwork travel and about 27 minutes for NHBW travel while the corresponding figures are extremely small for the Indian commuter sample. Therefore, the need for greater amount of out-of-home activity engagement essentially entails additional travel for the US commuters. Even though overall participation in non-work activities is quite lo w in the India sample, it is quite remarkable to find that individuals who actually engaged in those trav el activities have reported about 30 minutes of daily travel expenditure for home-based shopping purposes and an hour associated with social/recreational or other travel purposes. These findings are

PAGE 119

104 consistent in all the sample groups and th e figures are highly comparable with the corresponding averages found in the US sample groups. 3.11.3 Analysis of Activity Duration Again, Table 3.47 and Table 3.48 provide the di stributions of the out-of-home activity durations by purpose for the Indian and US market groups respectively. As expected, the Indian commuters have the highest averag e activity duration (550 minutes) followed by the student (about 350 minutes) and non-commut er (about 200 minutes) samples. In the Indian context, non-mandatory activity participation is virtua lly non-existent among student and commuter samples while averag e shopping and social/r ecreational activity durations (more than one hour in both cas es) are considerably large among the noncommuter sample and quite comparable with the corresponding US averages. Similar to the Indian sample, the highest daily ac tivity duration is found for commuters (534 minutes) followed by school-age childr en (391 minutes) and non-commuter (279 minutes) samples in the US. The average work duration for the Indian commuters is about 1 hour higher than that of the US commuter sample. However, average non-work activity durations are much higher in the US commuter and children samples compared to their Indian counterparts. However, the non-commuter sample in India spends 30 min more in shopping activity than the US non-commu ters. It is also inte resting to note that even though the average durati on of social/recreational activity for the US non-commuter sample is larger than the Indian non-co mmuters, the average duration reported by individuals who actually particip ated in those activities is 40 minutes higher for Indian non-commuter sample. Although there are considerable differences found in overall tr avel and activity durations between the survey samples due to low rate of non-work activity engagement among Indian respondents, the time use patter ns of individuals who reported actual participation in a particular activity show remarkable simila rities in the amount of their non-work activity-travel time expenditure. These findings lead to a speculation that India could experience a tremendous surge in trav el demand in near future. Traditionally, opportunities for living an activ e life style like developed wo rld have been lacked in

PAGE 120

105 peoples life in India. Social norms, crippl ed transportation systems are some of the reasons that have restricted the number of op tions for them to impr ove their quality of life. But in recent time, evolving Indian society has been relaxing many constraints in peoples life and bringing simultaneous change s in their travel behavior and time use patterns. More and more people will increasi ngly take advantage of these advancements and spend more time on travel and out-ofhome activities as a m eans of improving their quality of life.

PAGE 121

Table 3.47 Activity and Time Use Characteristics of Mobile Sample: India Activity and Time Use Characteristics Student (age 15 yr) Student (age 16 yr) Commuter Non-commuter All Sample Size 2285Actually participated1266Actually participated4646 Actually participated916Actually participated9165Actually participated Average Trip Frequency 2.002.00 (2285)2.012.01(1266)2.06 2.06 (4646)2.022.02 (916)2.032.03 (9165) Home based work 00002.02 2.02 (4646)001.042.02 (4698) Home based school 2.002.00 (2285)2.002.00 (1266)0 1.50 (2)000.772.00 (3553) Home based shop 00000 2 .00 (7)0.911.99 (417)0.091.99 (424) Home based social 02.00 (2)02.00 (2)0.01 1.93 (15)0.591.99 (272)0.061.99 (291) Home based others 0002.00 (2)0.01 1.52 (17)0.502.04 (218)0.052.00 (237) Non-home based work 00000.02 1.15 (72)000.021.15 (74) Non-home based other 01.00 (3)01.0 (4)0 00.021.00 (11)01 (18) Average Travel Duration (min) 2626.1 (2285)5252 (1266)73 73 (4646)4545 (916)5555 (9133) Home based work 000072 72 (4646)0036 72 (4682) Home based school 2626 (2285)5151 (1266)0 46.5 (2)001135 (3539) Home based shop 00000 28.5 (7)1431 (417)231 (424) Home based social 048 (2)155 (2)0.25 76 (15)1757 (271)257 (290) Home based others 00050 (2)0.25 41 (17)1355 (217)254 (236) Non-home based work 00000.5 30 (72)00029 (74) Non-home based other 010 (3)015 (4)0 0122 (11)018 (18) Average Activity Duration (min) 328328 (2285)346346 (1266)547 547 (4646)193193 (916)430430 (9165) Work 0000542 542 (4646)00277542 (4677) Educational 326327.5 (2277)342343 (1261)1 334 (2)00129334 (3540) Shopping 00001 142 (7)77170 (415)8171 (423) Social/Recreational 2246 (2)2182 (2)2 372 (16)70236 (271)8243 (291) Other 002185 (2)2 278 (18)56295 (211)8292 (231) 106 106

PAGE 122

Table 3.48 Activity and Time Use Characteristics of Mobile Sample: USA Activity and Time Use Characteristics School-age Children (age 15 yr) Commuter Non-commuter All 107 Sample Size 2366415Actually participated4177119Actually participated 8417263Actually participated 12594383Actually participated Average Trip Frequency 3.623.61(2366415)4.984.98(4177119) 4.324.32(8417263)4.534.54(12594383) Trip Frequency by Purpose Home based work 0.002.00(5034)1.581.77(3726502) 0.000.000.521.77(3726502) Home based shop 0.431.60(632214)0.521.55(1395114) 0.982.01(4144964)0.831.89(5540078) Home based social 0.892.05(1030967)0.551.85(1254882) 1.062.09(4275210)0.892.04(5530093) Home based others 1.492.01(1758260)0.541.83(1229749) 1.132.09(4531308)0.932.04(5761057) Non-home based work 2.00(5740)1.232.17(2357146) 0.000.000.402.17(2357146) Non-home based other 0.791.63(1148541)0.561.97(1188704) 1.142.05(4677877)0.952.03(5866581) Average Travel Duration (min) 6767 (2366415)105105(4177119) 9090 (8417263)9595(12594383) Travel Duration by Purpose Home based work 0180 (5034)4348(3711870) 001448(3711870) Home based shop 622 (632214)823 (1395114) 1531(4144964)1329 (5540078) Home based social 1944 (1030967)1033 (1247716) 3057 (4271853)2352 (5519568) Home based others 2737 (1758260)930 (1229749) 2343 (4531308)1840 (5761057) Non-home based work 020 (5740)2749 (2357146) 00949 (2357146) Non-home based other 1430 (1145098)830 (1183348) 2240 (4674434)1738 (5857782) Average Activity Duration (min) 391391 (2366415)534541(4120813) 279281(8354901)363367 (12475714) Activity Duration by Purpose Work 2439 (10774)448461(4059960) 00149461 (4059960) Educational 199421 (1122485)10213 (187855) 76378(1707265)54362 (1895120) Shopping 1760 (664987)1640 (1704443) 3669 (4395096)2961(6099539) Social/Recreational 93204 (1083237)41104 (1657790) 107196 (4559140)84172 (6216930) Other 80193 (977215)1967 (1172539) 60147 (3445978)46127 (4618516) 107

PAGE 123

108 3.12 Trip Length Distribution This section compares trip length distributi ons by trip purpose of the Indian and US survey samples. Table 3.49 provides averag e trip lengths by purpose of the survey samples and the graphical distributions are shown in Figure 3.31 and Figure 3.32. Trip length represents the length of travel time of each trip. Trip length can also be represented by the distance between trip origin and destina tion or cost of each trip. This study offers trip length analysis based on travel time. Average trip lengt hs of the India sample are found to be consistently higher than the US sample for all trip purposes. Much of these differences can be attributed to the highe r usage of slow mode s like non-motorized and public transportation in India that tend to s ubstantially increase travel time. In both contexts, HBW trip lengths are the highest compared with all other trip purposes, which is quite expected because majority of the work trips are made during the peak travel periods under heavily congested network condition. Consiste nt with expectation, HBS trips are found to have minimum trip lengt hs in both survey samples. In a typical weekday, majority of shopping trips are mainly associated with household maintenance type activities like running household errands, wh ich involve trips of short distance and duration to a local grocery or convenience store. Table 3.49 Average Trip Length Distribution by Purpose 2001 Thane, India 2001 NHTS, FL Trip Purpose Trip Length (min) Trip Length (min) All Trip 27 21 HBW 36 28 HBS 16 15 HBSocRec 29 25 HBO 27 20 NHB 23 19

PAGE 124

02468101214161820222426283032340-511-1521-2531-3541-4551-5561-6571-7581-8591-95101-105111-115121-125131-135141-145151-155161-165171-175>180Trip Length (min)Percent Trips All Trips HBW HBS HBSocRec HBO NHB Figure 3.31 Trip Length Distributions by Purpose: India 02468101214161820222426280-511-1521-2531-3541-4551-5561-6571-7581-8591-95101-105111-115121-125131-135141-145151-155161-165171-175>180Trip Length (min)Percent Trips All Trips HBW HBS HBSocRec HBO NHB Figure 3.32 Trip Length Distributions by Purpose: USA 109

PAGE 125

1103.13 Trip Chaining Analysis Trip chaining analysis is a very effective way of understanding non-work activity engagement patterns. In most trip chaining analyses, home and work are generally accounted as the primary anchors. Although there have been substantial diversity in the definitions of the terminologies like chain, tour and trip, this analysis defines trip as a one-way segment of travel between an origin and a destination, a tour is defined by the anchors and a chain is a set of trips in a tour (McGuckin and Murakami, 1995). Figure 3.33 demonstrates a classic case of trip chaining pattern of a commuter. According to the definitions, travel from home to work is an example of one of a tour and travel from home to drug store before work is an example of a trip made by the person. Trip chaining patterns of commuters based on data sets available from three different geographical locations around the world are analyzed in this section. Table 3.50 provides non-work activity engagement patterns of commuter samples obtained from Thane (India), Kyoto (Japan), and Southeast Florida (US) survey data sets. Overall comparison shows that the US commuter sample exhibits much complex trip chaining Work Home Grocery Store Yoga Class Day Care Fast Food Restaurant Drug Store After Wor k Before Wor k During Wor k On the way to Wor k On the way to Home Figure 3.33 Demonstration of Trips, Tours and Trip Chain

PAGE 126

111 patterns compared to the Indian and Japanese commuter samples. Majority of the Indian (97 percent) and Japanese (75 percent) co mmuters do not participate in any non-work activity while only 20 percent of the US commu ter sample reported to do so. It is found that about 44 percent of the US commuters engage more than one non-work activity in a typical working day. Much of these differences can be attributed mainly to the variability in mode choice behavior across the countries. With respect to India and Japan, majority of the commuters are primarily reliant on public transportation as opposed to the US context, where commuters are heavily auto-dependent. As one can imagine that private vehicles provide greater freedom and flexibil ities to access a wide range of services as opposed to public transportation, which is c onstrained by fixed route and schedule. Nonwork activity engagement patterns of Indian and Japanese commute r samples apparently furnish more constrained travel environment compared to the US. Table 3.50 Non-Work Activity Engage ment Patterns by Commuters Around the World Before Work On way work During work On way home From Home Thane % of Commuters Kyoto % of Commuters SE FL % of Commuters 97.34 75.10 20.40 0.09 1.00 5.60 0.04 0.00 0.00 1.89 12.00 9.00 0.17 5.30 11.40 0.41 3.00 9.60 0.04 0.20 2.50 0.02 0.90 10.50 0.00 0.20 4.90 0.00 1.10 4.30 0.00 0.50 4.30 0.00 0.30 4.90 0.00 0.10 4.30 0.00 0.00 1.50 0.00 0.00 4.30 0.00 0.10 1.50 0.00 0.00 0.90 TOTAL 100.00 100.00 100.00

PAGE 127

112 3.14 Summary and Discussion This chapter offers an extensive analysis of travel characteristics and time use behavior in a developing country context. To explore the differences in activity and time use patterns between a developing and a developed count ry context, the study adopts 2001 Household Travel Survey conducted in Thane City, I ndia and 2001 NHTS, Florida Sample for the purpose of comparison. With respect to dem ographic and socio-economic profiles of the datasets, the India sample is typically ch aracterized by larger household size, more children in the household, lower household/person al income and remarkably low level of vehicle ownership relative to the US sample. Person characteristics of the survey samples are investigated based on commuting status. Ma jority of Indian commuters are male and vast majority of Indian non-commuters are fe male, which is consistent with expectation because labor force participation among females are still very low in India compared to the developed countries. Cross-classification tables of trip production rates are generated based on household and person characteristics such as household size, vehicle ownership and income, and trip purpose. Consistent with expectation, househol d trip rates tend to increase with household size and vehicle ownership. Overal l trip production rates are found much higher for the US households co mpared to the Indian households. Trip distribution analysis suggests that more than 80 percent trips reported by the Indian respondents consists of home-based work and school trips, and a very small percent of non-work trips as opposed to the US trip di stribution pattern, wher e 75 percent of total reported trips are non-work in nature. Next, time of day analysis sheds light on many distinctive characteristics of departure time choice patter ns between the two countries. Detailed modal split analysis has been conducted based on trip purpose, commuting status and household car ownership. With respect to Indian sample groups, large proportion (90 percent) of the trips is acco mplished by either public transportation or non-motorized modes, while share of auto trips are prevalent among the US market segments (above 80 percent). Analysis of activity and time use characteristics of various market segments provides valuable insights about how people trade off their time into various mandatory and non-mandatory activitie s. The average trip frequency reported by the US sample groups are much higher than those reported by the India sample groups.

PAGE 128

113 Analysis suggests that the Indian travel ers undertake, on average one out-of-home activity per day while higher tr ip frequency reported by the US sample groups apparently indicates much higher level of activity participation in the US context. Consistent with trip rate patterns, average da ily travel durations for the US market segments are much higher than the Indian market segments. Commuters are found to spend more time on travel than non-commuters in both contexts With respect to activity duration, average time spent on non-work activities by Indian comm uter sample is remarkably lower than the US commuter sample. However, durations of shopping and recreati onal activities for Indian non-commuters are signifi cantly higher than the Indian commuters and also quite comparable with the corresponding averages fo und in the US non-commuter sample. Trip length distribution by purpose are analyzed and compared between the two survey samples. Average trip lengths of the India sample are found to be consistently higher than the US sample for all trip purposes. Finally, trip chaining analysis has been performed to understand the non-work activity engageme nt patterns of commuters. Comparison between commuter samples obtained from Indi a, Japan and US suggests that the US commuters have much complex trip chaining patterns compared to the commuters in India and Japan. Much of these differences ar e attributed mainly to the variability in mode choice behavior across the countries. The Indian and Japanese commuters are likely to face greater constants in particip ating non-work activities because they are heavily reliant on public transportation while the US commuters are primarily dependent on private vehicle which provides them great er flexibility to access their activity locations.

PAGE 129

114 CHAPTER 4 EXPLORATION OF TRAVEL TIME FRONTIERS AROUND THE WORLD The previous chapter has offered a comparative analysis of travel characteristics and time use pattern based on datasets from a deve loping and a developed country. The study has explored similarities and differences between various market groups with respect to their socio-economic and demographic characteristics, trip production rate s, trip distribution and mode choice patterns, time-of-day and trip characteristics, activity, time use, and trip chaining patterns. The current chapter deals with understanding travel time expenditure by exploring the notion of travel time frontier around the world. 4.1 Travel Time Budget Vers us Travel Time Frontier Travel behavior researchers have devoted a considerable effort to conceptualize and understand the potential existence of an in trinsic fixed amount of time that people allocate to travel, called a travel time budget. Although research into travel time budgets started to appear in the literature in the 1970s, the concept has remained subject to much debate, scrutiny, and research over the years. In virtually all of the studies examining this issue, travel time expenditure s have been treated as representing travel time budgets and their variation over time and space has been studied. Zahavi (1979) was one of the first who raised the concept of a st able travel time budge t. Since then, several studies (Zahavi and Ryan 1980; Zahavi and Talvitie 1980; Schafer 2000; Schafer and Victor 2000; and Hupkes 1982) have examined fo r regularities in trav el behavior through analyses of travel time and monetary expe nditures. Their work generally supported the notion of the existence of a spatially and tem porally stable daily travel time budget. The study of Robinson et al. (1972) on daily travel time expenditure in twelve countries provided further support for this notion.

PAGE 130

115 Despite the early evidence of this concept in the literature, many researchers have not observed the existence of such a constant travel time budget. Kitamura et al. (1992) found results contradictory to th e notion of spatial stability in travel time expenditure. Purvis (1994), Levinson and Kumar (1995), and Ku mar and Levinson (1995) observed temporal instability in travel time expenditu res in the US context. Indeed, analysis of U.S. national travel survey series shows that daily travel time expenditures have been steadily rising at the rate of about two minutes per person per year between 1983 and 2001 (Toole-Holt et al., 2005), although it is unlike ly that this rate of increase can be sustained in the future and perhaps some of the increase can be attributed to improvements in survey design procedures th at resulted in better reporting of short, infrequent, and non-motorized trips over time. More recently, Mokhtarian et al. (20 01), Mokhtarian and Solomon (2001), and Mokhtarian and Chen (2004) have further expl ored issues of trav el time budgets and monetary expenditures. They present a compre hensive discussion of the subject and show that travel time expenditure can be related to personal and house hold characteristics, activity duration, and resident ial location. A behavioral c onstruct for modeling travel time expenditure is also supported by Principio and Pas (1997). Recent evidence suggests that travel tim e expenditures (or budgets) are indeed changing over time and space (Kitamur a 1992; Levinson & Kumar 1995; Kumar & Levinson 1995; Toole-Holt et al 2005). Motiv ated by these findings, this study is aimed at further exploring this s ubject by developing the notion of a travel time frontier (TTF), which is representative of the ma ximum amount of time that an individual is willing to travel in a day. It is hypothesized that travel time expenditures/budgets are showing increases over time because the TTF is considerably great er than the actual expenditure/budget. The intent of the study is to test this hypothesis and quantify the TTF with a view to understand the extent to wh ich travel time expend itures could potentially increase in different international contexts. The challenge associated with modeling or identifying the TTF is that it is an unobserved value. While travel survey data se ts provide actual travel time expenditures, they do not provide any information on the maximum amount of time that a person is

PAGE 131

116willing to allocate for travel. Thus, the TTF, as defined in this study, constitutes an unobserved frontier that influences the actual travel time expenditure. Based on observations of actual travel time expenditures, estimates of the unobserved frontier need to be obtained. This can be accomplished by employing the stochastic frontier modeling methodology in which the TTF is formulated as a production frontier model. Travel survey data sets from three countries, the United States, Switzerland, and India, are used to examine expected TTF distributions for commuter and non-commuter samples. 4.2 Modeling Methodology The stochastic frontier modeling methodology is employed to estimate the unobserved travel time frontier (TTF) that is representative of the maximum amount of travel time that an individual is willing to undertake in a day. Due to the highly skewed nature of the travel time distribution and to ensure positive predictions, a log transformation of the dependent variable is used. Let and )tln(Tii iiiuT (1) where i denotes the observation, is observed total daily travel time and is a random component that takes non-negative values. Then, it iu i which constitutes an unobserved frontier for, is always greater than or equal to. A possible model that applies to these relationships is the stochastic frontier model (Aigner et al. 1977), whose general form can be presented as: (2) Then, iT iT iiivX iiiiiiuvXXT (3) where is a vector of coefficients, is a vector of explanatory variables, is a random error term such that iX iv iv. In the context of this study, may be viewed as the location of the unobserved frontier for with the random component will not exceed ii'vX iT iv. iT iivX because is non-negative. Assuming a half normal distribution for (Aigner et al., 1997) and a normal distribution for, the distribution of iu iu iv i can be derived as:

PAGE 132

117 i22iii ; 2exp)}/(1{22)h( (4) ),,0(N~v,/ t;independenmutually are and ,)uvvar(2vivuuv2u2vii2 and iuhas the density function, .0u,2uexp22)u(gi2u2iui (5) with ui2)uE( and 2ui21)uVar( (6) The log-likelihood function for the sample of observations is then given by: (7) The maximization of the log-likelihood function allows the consistent estimation of the unknown parameters n1ii)][h(lnLL and , By replacing with in equation (3), one obtains: iT )tln(i )uexp()vX'exp(tiiii (8) As, observed travel time will not exceed 1)uexp(0i it )vX'exp(ii Therefore, may be considered as representative of the travel time frontier (TTF). The lue of the travel time frontier with random component may be denoted as: )vX'exp(iiexpected va iv )](vexpE[)X(exp])(vexp)X(expE[)E(TTFiiiii (9) Because is distributed log-normal and (Greene, 2002). Then, (10) may be estimated by maximum likelihood procedures and the resulting estim, can then be substituted into equation (10) to compute the expected TTF. The ratio between the expected travel time expenditure and expected TTF may be derived as: (11) This implies that individuals are expected to spend ),0(N~v2vi )(vexpi /2)(exp)](vexpE[2vi /2)X(exp)E(TTF2vii v and ates, and v )](/2)[1(exp2)]u(expE[))/E(TTFE(tru2uiii )TTF(Er minutes for daily travel (refer Appendix A for detailed derivation).

PAGE 133

118 A note is due here to clar ify the nature of the TTF postulated in this study. Historically, the literature has not differen tiated between the concepts of travel time expenditure and travel time budget and researchers have treated these terms synonymously (Mokhtarian et al., 2001; Mokhtarian and Solomon, 2001; Mokhtarian and Chen, 2004; Levinson and Kumar, 1995; Ku mar and Levinson, 1995). However, one could also argue that these are two distinct terms. For example, Goodwin (1981) explained that expenditure refers to th e amount of quantitative resources spent on consuming a good or service or performing an activity (travel). On the contrary, budget implies a certain amount of stab ility, referring to a maximum level of allocation of time, money, or generalized resources to a good or service (travel). If such a distinction is made, then the TTF estimated in this study may be considered to be representative of the travel time budget i.e., the amount of time that a person is willing to budget for travel, regardless of whether he or she actually spends it (expenditure). In an effort to be consistent with the larger body of literature that treats travel time expenditure and budget to be equivalent, this study util izes the travel time frontier term to refer to the estimated stochastic frontier. Model estimation in this study is accomp lished using LIMDEP (Greene 2002). In the current context, there is some ambiguity on the exact interpretation of the estimated frontier. In its original de finition, the production frontier re presents the maximum amount of goods or products that an entity (say, a ma nufacturing plant) can (o r is able to) produce given the available resources, infrastructure, and constraints (inputs). In this particular application of the stochastic fr ontier model, a slight modification of this interpretation is warranted. The maximum amount of travel that a person can produce in any day is set by the clock, i.e., 24 hours. Individuals can certainly allocate all 24 hours of a day to travel, say, when undertaking a long intern ational trip. Thus, for this particular application, the stochastic frontier is treated as being representative of the perceived or subjective maximum amount of time that a person is generally willing to undertake in a day. As this is a perceived or subjective lim it or threshold, violations can occur, i.e., the actual travel time expenditure may exceed the estimated frontier. This is similar to a situation where a person may subjectively th ink that he or she is willing to spend

PAGE 134

119 (frontier) no more than a certain amount of money for a vacation, but ends up actually spending a little more depending on the circum stances that prevail during the vacation. Although the formulation of the stochastic front ier model is not entirely consistent with this interpretation, it nevertheless offers a strong me thodological framework for attempting to estimate the unobserved upper bou nd for daily travel time expenditures. 4.3 Data Sets Three data sets from around the world are used to estimate the travel time frontiers. All three data sets are derived from household tr avel surveys in which respondent samples provided detailed trip information for a 24-hour period. The three surveys are: 2001 National Household Travel Survey of the United States 2000 Microcensus Travel Su rvey of Switzerland 2001 Household Travel Survey of the City of Thane, India While the first two surveys constitute national surveys, the survey from India is from a single metropolitan area in India as a national travel survey is not available in India. All three travel survey s are based on the trip-diary format in which respondents are asked to provide detailed information a bout trips in addition to socio-economic, demographic, and other characteristics of households and persons. However, there are differences in survey methodologies and ad ministration methods that could affect comparisons across the three data sets. Table 4.1 provides household characteristics for the three data sets. As expected, the US and Swiss survey samples show small household sizes, higher levels of licensed drivers and car ownership, and a highly urbanized population. India is characterized by large household sizes, more children, very fe w licensed drivers, and very low levels of car ownership.

PAGE 135

120 Table 4.1 Household Characteristics (US, Swiss, India) Characteristic 2001 US NHTS 2000 Swiss Travel Microcensus Survey 2001 Thane, India Household Survey Sample Size 26,038 27,918 3505 Household Size 2.56 2.43 4.12 1 person 25.82% 27.5% 2.0% 2 persons 32.63% 35.1% 12.2% 3 persons 16.53% 14.0% 19.8% 4 persons 25.02% 23.4% 66.0% Children Age: <18 yr 0.67 0.51 0.90 0 children 64.4% 71.3% 47.4% 1 child 14.6% 11.6% 26.4% 2 children 13.8% 12.4% 16.9% 3+ children 7.3% 4.7% 9.3% No. of Workers 1.31 N/A 1.34 0 workers 22.9% N/A 9.3% 1 worker 34.5% N/A 57.3% 2 workers 33.7% N/A 25.8% 3+ workers 8.9% N/A 7.6% No. of Licensed Drivers 1.75 1.51 0.10 0 licensed drivers 5.38% 12.8% 91.3% 1 licensed driver 31.85% 34.1% 7.4% 2 licensed drivers 49.25% 44.6% 1.0% 3 or more drivers 13.52% 8.5% 0.3% Annual Income Low income > $25K (29.1%)> Fr 48K (20.8%) Rs 60K (42.2%) Medium income $25-50K (33.3%)Fr48 96K (35.9%) Rs 60-180K (45%) High income > $50K (37.6%)>Fr 96K (18.4%) >Rs 180K (12.8%) Vehicle Ownership 1.90 1.17 0.06 0 auto 7.9% 19.8% 94.7% 1 auto 31.4% 50.5% 4.9% 2 autos 37.1% 24.5% 0.4% 3 autos 23.6% 5.2% 0 Residential area type Urban 79.5% 78.6% N/A Non-Urban 20.5% 21.4% N/A

PAGE 136

121 Table 4.2 offers a detailed look at the pe rson characteristics by commuting status. All individuals who made at least one work trip or work-related business trip on the travel survey day were treated as commu ters and all others were treated as noncommuters. The term worker and commuter will be used synonymously in the remainder of this study. Also, the analysis in this study is limited to mobile adu lts, i.e., individuals 18 years or above who reported at least one trip, to ensure samp le consistency across surveys and account for the fact that children are often dependent on adults for their travel needs. In the US and Swiss samples, it is found that majority of mobile non-commuters are females and retirement age oriented while the majority of mobile non-commuters in the India sample are males and young student-age oriented. A vast majority of individuals is licensed to drive in the US and Swiss samples, while only a very small minority is licensed to drive in India. With respect to trip frequencies, both mobile commuters and non-commuters in US and Swiss samples make more than four trips per day. The mobile commuters and non-commuters in India make, on average, less than one-half as many trips at just about two trips per day. Table 4.3 also provides a glimpse of av erage travel durations by purpose for mobile commuters and non-commuters in each of the three data sets. Average travel duration is about 1 hr 30 min fo r mobile adults in the US survey, about 1 hr 40 min for mobile adults in the Swiss survey, and closer to one hour for mobile adults in the India survey. To avoid definitional ambiguity of the trip purposes across surveys, this study focuses on modeling the unobserved production frontier for the to tal travel time expenditure (including work tr avel duration for commuters) as opposed to travel time expenditures by purpose.

PAGE 137

122 Table 4.2 Person Characteristics of Mo bile Adults (USA, Swiss, India) 2001 US NHTS 2000 Swiss Travel Microcensus Survey 2001 Thane, India Household Survey Characteristic Com Non-ComCom Non-Com Com Non-Com Sample Size 1762622507824714110 4623 1699 Age (in years) 42.4351.8941.551.7 36.36 30.96 18-24 years 9.35%7.5%8.6%8.3% 15.1% 53.4% 26-64 years 86.85%63.5%89.6%61.4% 84.1% 41.0% 65+ years 3.9%25.0%1.8%30.3% 0.7% 5.6% Sex Male 54.6%41.9%58.4%40.5% 85.4% 53.6% Female 45.4%58.1%41.6%59.5% 14.6% 46.4% Employment Status Unemployed 1.8%58.0%N/AN/A N/A N/A Full time 84.1%31.0%75.6%28.1% N/A N/A Part time 13.5%10.8%19.2%14.4% N/A N/A Multiple Jobs 0.5%0.2%1.3%0.8% N/A N/A Licensed Driver 97.3%92.8%89.2%75.7% 17.9% 6.6% Education College or less 57.7%65.4%81.6%87.4% 70.7% 73.0% Graduate degree 41.9%35.1%18.4%12.6% 29.3% 27.0% Occupation Sales/Service 25.3%11.7%N/AN/A 71.0% 2.8% Clerical/Admin 12.0%5.5%N/AN/A N/A N/A Laborer 20.2%7.3%N/AN/A 7.0% 1.4% Professional 40.6%17.4%N/AN/A 21.5% 6.2% Student N/AN/AN/AN/A 0.3% 47.0% Retired/unemployed N/AN/AN/AN/A 0.2% 17.4% Homemaker N/AN/AN/AN/A 0% 25.4% No. of Trips/day 4.894.924.663.54 2.06 2.01 Work trips 2.4301.600 0.84 0 Non-work trips 2.464.923.063.54 1.21 2.01 Daily Miles Traveled 52.849.331.037.8 N/A N/A

PAGE 138

123 Table 4.3 Average Daily Travel Duration by Purpose of Mobile Adults Characteristics 2001 US NHTS 2000 Swiss Travel Microcensus Survey 2001 Thane, India Household Survey Travel Duration (min) Com NonCom Com Non-Com Com Non-Com Work 26.0035.50 31.9 0 Business 9.608.44.1 4.6 0.3 School 0.61.00.41.5 0.0 13.7 Shopping 5.014.05.513.0 0.0 3.7 Social/Recreation 6.319.016.850.2 0.1 4.5 Return Home 31.032.032.243.0 36.1 25.7 Others 15.028.52.23.4 0.1 3.5 Total 93.594.5100.9115.1 72.9 51.3 4.4 Model Estimation Results This section presents a brief overview of th e model estimation results. First, Table 4.4 presents results of the model estimation effort for the US survey samples. Being male, having a drivers license, having a college education, and being employed are all positively impacting the TTF On the other hand, work activity duration is found to have a negative impact on the TTF. This is quite consiste nt with expectations as a larger amount of time spent at work w ill result in a decreased amount of time that can be allocated for travel. Commuters who live farther from their workplace appear to have higher TTFs, possibly because these workers have to allocate more time for travel to accommodate the longer commute. Weeke nds are characterized by lower TTFs, presumably because of the absence of commut e travel, while Fridays are characterized by higher TTFs, possibly due to par ticipation in discretionary ac tivities at the end of the work week. Greater household obligations such as the presence of children lead to higher TTFs for commuters, but a smaller TTF for non-commuters. Higher income levels are associated with higher TTFs, presumably b ecause higher income levels allow people to make use of other services fo r taking care of in-home activities (cleaning services, etc.) and allow people to afford out-of-home ac tivities such as eat-meal and recreation. Table 4.5 presents the stochastic production frontier models for the Swiss survey samples. Males are found to have higher TTFs, possibly because females are likely to be bearing a greater share of the household and childcare responsibilities. Indeed, larger household sizes and the presence of children are associated with smaller TTFs for non-

PAGE 139

124 commuters similar to the US context. Highe r TTFs exist for highly educated and retired non-commuters possibly due to greater awar eness of activity opportunities and absence of household obligations. Higher car ownership and income levels are associated with higher TTFs similar to the US context. As expected, a higher TTF exists for travel on Fridays. Non-commuters in the Swiss sample appear to have higher TTFs on weekends. Weekends did not have a significant imp act on commuter TTFs, possibly because the Swiss lifestyle is not as weekend-oriented as in the US. Similar to the US case, for commuters, the daily work activity duration has a negative impact on the TTF while the distance from home to work has a positive impact on the TTF. Finally, Table 4.6 presents the models for the India survey samples. Unlike the US and Swiss samples, the male commuter in India is found to exhibit a lower TTF. This is likely due to the inherent nature of male commuters in India who generally spend time at work and home during a typical co mmuting day while female commuters may undertake other chores and house hold errands in addition to wo rk travel. Having a drivers license, which is closely related to auto ow nership, is associated with higher TTFs for both commuter and non-commuter samples. Si milarly, high income individuals and those with graduate-level education exhibit higher TTF s. All of these variab les signify a greater level of affordability, awareness, and e ducation. On the other hand, low income noncommuters also appear to have higher TTFs, but presumably because they are more likely to be dependent on slower modes due to lo w vehicle ownership. For both commuters and non-commuters, a higher level of out-of-home activity participation is associated with higher TTFs, consistent with the notion that people who engage in ac tivities are likely to allocate more time for travel to undertake those activities.

PAGE 140

125Table 4.4 Stochastic Frontier Models of Travel Time Frontier: USA Mobile Commuters Mobile Non-Commuters Variable Coefficient t-stat Coefficient t-stat Constant 5.07895.4494.736 87.228 Age 0.0052.1340.012 6.293Age squared -0.00005-1.831-0.0003 -6.765Gender: Male 0.11710.5290.036 3.060Licensed driver ----0.122 5.146Education level: Bachelor or equivalent 0.0524.2510.045 3.398Employment status : full time 0.1227.541---Employed ----0.056 4.074Work activity duration (min) -0.001-45.091---Distance to work from home (miles) 0.012104.586---Occupation: professional 0.0332.776---Travel day: Sat/Sunday -0.115-7.123---Travel day: Friday 0.1006.6540.132 7.188Race: White -0.049-3.298-0.073 -4.452US immigrant 0.0291.530---Household with children 0.0312.778-0.052 -3.679HH car ownership >1 -0.050-3.455---Household vehicle ownership ----0.013 2.581High HH income ( $70K/Annum) 0.0755.9770.047 3.157Low HH income (< $25K/Annum) -0.091-5.811-0.045 -2.931Residential neighborhood: Urban -----0.056 -4.065 2.01541.5871.910 37.6220.981127.2341.222 119.149 L(C) -17625.67 -26274.80 L() -16143.57 -26082.24 2[df] 2964.200 [16] 385.120 [13] 0.7722 1.1722 0.1901 0.3214 r 0.5584 0.5013 Sample Size 15791 20740 2u 2v

PAGE 141

126Table 4.5 Stochastic Frontier Models of Travel Time Frontier: Swiss Mobile Commuters Mobile Non-Commuters Variable Coefficient t-stat Coefficient t-stat Constant 6.016264.8015.338469.619 Gender: male 0.13947.6830.12857.717Age -0.0148-3.5530.00903.397Age squared 0.00012.642-0.0002-6.107Household size -0.0228-3.896-0.0497-5.244Highly educated ----0.11404.393Occupation: retired ----0.05921.963Household car ownership = 0 0.07422.916-0.0347-1.545Household car ownership 2 ----0.07533.657High HH income (> Fr 10000/month) 0.05812.5740.07142.348Low HH income (
PAGE 142

127Table 4.6 Stochastic Frontier Models of Travel Time Frontier: India Mobile Commuters Mobile Non-Commuters Variable Coefficient t-stat Coefficient t-stat Constant 4.012794.6013.4490 55.980 Male -0.0837-4.681---Occupation: service 0.12525.028-0.1767 -2.237Occupation: student ----0.1853 3.846Occupation: retired ----0.1194 2.721Licensed driver 0.09833.1260.2147 2.771Education: graduate 0.09393.887---Household size -0.0167-2.539---No. of daily activity participated >2 0.29846.3150.3874 2.259Daily work activity duration 0.00036.264 Transit user 1.046443.9340.8743 27.078No. of children in the household 0.02192.090---Household vehicle ownership 0.03812.473---High income (> Rs. 20K monthly) 0.19141.428---Low income ( Rs. 5K monthly) ----0.0865 1.886Young (age: 18-29) ----0.0939 2.131Daily activity duration ----0.0003 5.537 2.517920.3871.3887 7.8951.086767.8770.8654 26.427 L(C) -5878.792 -2082.161 L() -4960.784 -1693.617 2[df] 1836.016 [11] 777.088[9] 1.0198 0.4928 0.1609 0.2557 r 0.5205 0.6174 Sample Size 4623 1699 2u 2v

PAGE 143

1284.5 Distributions of Travel Time Expenditures and Frontiers The stochastic frontier models can be used to estimate the expected TTF (see equation 10) or production frontier for each individual in the survey samples and generate distributions of expected TTFs vis--vis distributions of actual travel time expenditures. Plots of distributions of expected TTFs and actual travel time expenditures provide a concise picture of the relative differences between expenditures and expected frontiers in the different survey samples. In Figure 4.1, distributions of expected TTFs and actual travel time expenditures are shown for US survey samples. The expected TTF distributions are those in the figure that have sharper peaks and are to the right of the actual travel time expenditure distributions. The average values of the expected TTFs are found to be 165 minutes for commuters and 188 minutes for non-commuters. As expected, non-commuters have larger expected TTFs, possibly due to the absence of the large work activity commitment that commuters have to make in the course of a day. These values represent average expected TTFs close to three hours for mobile adults in the US samples. 05101520253035404550556000~3030~6060~9090~120120~150150~180180~210210~240240~270270~300300~330330~360360~390390~420420~450450~480480~510510~540540~570570~600600+Time (Minutes)Percentage Commuters Travel Time Expenditure (Mean = 93 min) Commuters Travel Time Frontier (Mean = 165 min) Non-commuters Travel Time Expenditure (Mean = 95min) Non-commuters Travel Time Frontier (Mean = 188 min) Figure 4.1 Distributions of Travel Time Expenditures and Estimated Frontiers: USA Mobile Commuters and Non-Commuters

PAGE 144

129In Figure 4.2, distributions are shown for the Swiss survey samples. In the case of the Swiss survey samples, the distributions of the expected TTFs are found to be more spread out and follow patterns similar to the actual travel time expenditure distributions. It is also interesting to note that the difference between commuters and non-commuters is more pronounced in the Swiss survey than in the US survey. The commuters average expected TTF is found to be about 187 minutes; the corresponding value for non-commuters is found to be about 250 minutes. This represents a one hour difference between commuters and non-commuters; the corresponding difference between the US commuter and non-commuter samples is only about 20 minutes. Consistent with expectations, non-commuters are found to have higher expected TTF than non-commuters. Thus it appears that the average commuter TTF in Switzerland is about three hours (similar to the US), but the non-commuters average expected frontier is four hours, which is about one hour more than that of the US non-commuter sample. 05101520253035404550556000~3030~6060~9090~120120~150150~180180~210210~240240~270270~300300~330330~360360~390390~420420~450450~480480~510510~540540~570570~600600+Time (Minutes)Percentage Commuters Travel Time Expenditure (Mean = 101 min) Commuters Travel Time Frontier (Mean = 187 min) Non-commuters Travel Time Expenditure (Mean = 115 min) Non-commuters Travel Time Frontier (Mean = 250 min) Figure 4.2 Distributions of Travel Time Expenditures and Estimated Frontiers: Swiss Mobile Commuters and Non-Commuters

PAGE 145

130Figure 4.3 offers a contrasting picture. In the Indian context, it is found that, on average, commuters exhibit a greater expected TTF than non-commuters, a finding that is opposite to that seen for the US and Swiss samples. The commuter sample shows an average of 141 minutes for the expected TTF. The corresponding value for the non-commuter sample is only 83 minutes. This finding is in sharp contrast to the developed countries represented by US and Swiss survey samples. In the Indian context, it is conceivable that non-commuters take on a greater share of the household obligations and childcare responsibilities. In addition, they are likely to be less mobile due to the poorer infrastructure, low auto availability, and disposable income. As a result, non-commuters allocate more time to in-home stay and consequently exhibit a shorter expected TTF. 05101520253035404550556000~3030~6060~9090~120120~150150~180180~210210~240240~270270~300300~330330~360360~390390~420420~450450~480480~510510~540540~570570~600600+Time (minutes)Percentage Commuters Travel Time Expenditure (Mean = 73 min) CommutersTravel Time Frontier (Mean = 141 min) Non-commuters Travel Time Expenditure (Mean = 51 min) Non-commuters Travel Time Frontier (Mean = 83 min) Figure 4.3 Distributions of Travel Time Expenditures and Estimated Frontiers: India Mobile Commuters and Non-Commuters Another interesting finding in the Indian context is the very bi-modal nature of the expected TTF distributions. Each distribution is characterized by two sharp peaks. In analyzing this distribution further, it was found that these two peaks (in each distribution) represent two distinct market segments, one of them consists of transit users and the other segment uses other modes of transportation such as car (very small percentage), walk,

PAGE 146

131 and bike. The first peak consists of the segm ent that is using non-tr ansit modes while the second peak in each distribution consists prim arily of individuals who are using transit almost exclusively for all of their trips. Th ese findings suggest that transit users tend to have higher TTFs (on average) compared to other mode users. I ndeed, in the Indian context, transit is a slow mode and those tr aveling longer distances (by transit) must, by necessity, allocate larger amounts of time for travel relative to others who travel shorter distances by non-tran sit modes or travel faster by car. Table 4.7 offers a summary of the interna tional comparison of average travel time expenditures, average expected TTFs, and aver age values of r, the ratio of the actual travel time expenditure to the expected TTF. While the differences in the average expected TTFs are rather modest for the commuter samples, they are quite pronounced for the non-commuter samples. The average expected TTFs range from about 2.5 hours to 3 hours for all three commuter samples. However, the corresponding range for the noncommuter samples is from about 1.5 hours to 4 hours. Table 4.7 International Comparison of Average Travel Time Expenditures and Average Estimated Travel Time Frontiers Commuters Non-commuters Study Area Statistics Travel time expenditure TTF Ratio (r) Travel time expenditure TTF Ratio (r) Mean (min) 93.5164.8 94.5 187.9 United States SD (min) 74.757.2 0.558 84.0 21.0 0.501 Mean (min) 100.9187.2 115.1 250.4 Switzerland SD (min) 93.3218.9 0.556 119.9 51.7 0.464 Mean (min) 72.9140.6 51.3 83.5 Thane, India SD (min) 67.055.0 0.521 43.0 39.1 0.617 SD: Standard Deviation Similarly, the average values of the ratio, r, are more similar across the commuter samples than for the non-commuter samples. For commuters, the average value of r is a little more than 0.5 which suggests that, on average, actual travel time expenditures are about one-half of the expected TTF. A simila r ratio is observed for non-commuters as well, except that the values ra nge from about 0.5 in the US an d Swiss contexts to a little more than 0.6 in the Indian context. In gene ral, non-commuters in India have the smallest

PAGE 147

132 expected travel time frontiers compared to all other groups. One interpretation of the findings reported here is that people, on av erage, spend about one-half of the maximum amount of time that they are willing to allocate to travel. Conversely, the maximum amount of travel that people are willing to undertake is, on average, about twice the actual travel time expenditures observed in th e data sets. In the event that household, monetary, modal, institutional, and situationa l constraints were loos ened, it is conceivable that travel could increase up to the poin t of the TTF reported in this study. 4.6 Summary and Discussion In this study, the stochastic frontier modeling methodology is employed to identify an unobserved travel time frontier (TTF) that is considered to be re presentative of the maximum amount of time that an individual is willing to allocate to travel in a day. Production frontier models are estimated for commuter and non-commuter samples drawn from three different travel survey data sets: 2001 USA National Household Travel Survey, 2000 Switzerland Microcensus Travel Survey, and 2001 Thane, India Household Travel Survey. The results presented in this study shed c onsiderable light on the variability of the TTF across international contexts. The model estimation results were used to generate distributions of expected TTFs. The averag e expected TTFs were found to be about 3 hours for US and Swiss commuters and about 2.5 hours for Indian commuters. Although the range of these average expected values is rather narrow, it is clear from the distribution that there is considerable inter-p erson variation in the expected travel time frontiers. For non-commuter samples, the distributions are even more spread-out, and the aggregate sample-wide averages of the e xpected TTFs range from about 1.5 hours to 4 hours. The findings reported in this study have im portant transport policy implications. Around the world, transport policies, infrastr ucture investments, telecommunications technology, 24-hour business establishments, moda l flexibility and av ailability, virtual workplaces, and smaller household sizes are resulting in the loosening of constraints and the increased availability of discretionary resources. The notion of the TTF provides a

PAGE 148

133 powerful framework for analyzing increases in travel time expenditures that might result from continued loosening of constraints over time; presumably, travel time expenditures can continue to rise as long as they are lower than the TTF, but would stop increasing when the TTF is reached. Thus, this con ceptualization provides a means for analyzing induced travel effects from an activity-based time use allocation perspective. Finally, a note is due here regarding the definition of a travel ti me frontier (TTF). In this study, the unobserved production frontie r is treated as a perceived subjective maximum amount of time that an individual mi ght be willing to allocate to travel (as opposed to a pure objectively defined maximu m travel time which is 24 hours). On a regular daily basis, it is c onceivable that people have some subjective judgment of the maximum amount of time that they are willing to allocate to travel. It is this subjective maximum travel time allocation that is m odeled within this study using the production frontier modeling methodology. It is felt that the actual travel time expenditure is likely to be most influenced by an individuals subjective judgment of his or her travel time frontier. Further explorations into the subjective judgments of individuals regarding their travel time allocation and expenditure are warranted in future studies In addition, future research should examine the dynamics of TTFs through an analysis of longitudinal data which would shed light on how maximum subj ective allocations of travel time may be changing or shifting over time.

PAGE 149

134 CHAPTER 5 HOW LOW CAN TRAVEL GO The previous chapter has presente d a series of models to bett er understand the travel time expenditure around the world by attempting to determine the subjective or perceived maximum amount of travel time (or travel time frontier) that people are willing to undertake in day. The findings shed considerab le light on the variab ility of TTF across international contexts. The present chapter i nvestigates travel time expenditure from a different behavioral standpoint. The study presented here attempts to quantify the theoretical minimum travel time that a pers on feels is absolutely required to accomplish the mandatory activities of the day. 5.1 Introduction to Minimum Travel Time Threshold This study is concerned with answering the question: How lo w can travel go? It is postulated that there is a minimum amount of travel that a person feels he or she must undertake to accomplish the required activities of the day. In most developed countries, the focus of transportation planning has shifte d away from capacity expansion to that of operation, management, and optimization of ex isting capacity. This shift in planning emphasis has motivated travel behavior resear chers to be concerned with relationships and trade-offs among individuals time expenditures, travel, and activities (Kitamura et al., 1997; Bhat et al., 1999 and Yamamoto et al., 1999). It is envisi oned that travel behavior models based on an understanding of peoples time use patterns offer a robust framework for analyzing the impacts of alte rnative transportation policies and control measures. If transportation control m easures are aimed at reducing (vehicular) travel, then the question arises as to the extent to which travel can be eliminat ed. In other words, what is the minimum amount (lower bound) of travel time beyond which travel can not

PAGE 150

135 be reduced further? Potentially, individuals must undertake or feel that they must undertake a certain amount of minimum travel in order to accomplish activities that are mandatory or required. For example, a person may engage in 100 minut es of travel in a day even though the mandatory or required ac tivities of the day can be accomplished in as little as 20 minutes of travel. Thus, it is clear that the subj ective minimum amount of travel that a person feels he or she must undertake is closely related to the subjective judgment (perception) of the activities that are mandatory and required. Presumably, an individual undertakes additional travel because the additional (flexible or discretionary) activities offer a positive utility that outweighs any negative u tility due to the travel that needs to be undertaken to participate in those activities. One can speculate, then, that this additional travel (beyond) the minimum is a candidate for potential elimination or reduction through the implemen tation of suitable travel demand management (TDM) strategies and transportation control measures (TCM). Modeling the minimum subjective travel time threshold would offer a basis for quantifying the potential maximum reduction in travel that can be brought about by implementing various policies. The notion of minimum required travel time expenditure may be considered analogous to that of a minimum required mone tary expenditure for subsistence. A person may spend a certain amount of money on f ood, clothing, and other goods. However, not all of this expenditure may be absolutely ne cessary for subsistence. Of all the money spent, only a small fraction may be absolutely necessary for subsistence; in the event of a crisis, the individual would not be able to spend a ny amount less than a subjective minimum threshold value. In the transportation field, modeling the subjective minimum travel time threshold is also useful from the standpoint of gaugi ng the effectiveness a nd performance of the transportation system. Presumably, if the actual travel time expenditure is considerably larger than the subjective minimum threshold value, then it means that the transportation system is performing at a level that motivates additional travel. The travel disutility is low enough that people are motivated to pursu e additional activities and travel. In a context where the transportati on system performance is very poor, one would expect the

PAGE 151

136 actual travel time expenditure to be close to the subjective minimum value. This is because the larger travel disutility hinders additional activity and travel engagement. It should be possible to identify the mi nimum required travel time threshold by simply adding all daily travel durations to ma ndatory activities. Ho wever, there are three issues associated with such a simplistic approach. First, the definition of mandatory activities (and its associated travel) is uncle ar. What may be a mandatory activity for one individual may be non-mandatory for anothe r. Thus, computing the minimum required travel time based on a defined set of mandatory activities would be problematic. Second, it is generally quite difficult to truly isolate the required travel associated with the mandatory activities from a travel diary survey data set. Trip pa tterns generally consist of a host of trip chains, journeys, and tours. Within the context of these complex patterns, one would have to make simplifyi ng assumptions to isolate the absolutely necessary trips to accomplish the predefined se t of mandatory activities. Once again, this may lead to erroneous estimates of the minimu m travel time threshold. Finally, there is a third issue in that the mandatory travel (trips ) observed in the data set may not constitute the minimum paths or the most efficient configuration for completing the mandatory activities of the day. Faced with a situation where an individual must undertake no more travel than absolutely necessary, it is possi ble that a more efficient minimum travel configuration can be found and executed wh ile accomplishing those necessary mandatory activities. Then, even if the absolutely manda tory trips could be isolated correctly in a travel diary data set, the corresponding durat ion may not constitute the minimum required travel time threshold as it may be possible to accomplish the same set of activities in an even smaller duration of time. The objective of this study is to qua ntify the theoretical minimum travel time that a person feels is absolutely required to accomplish the mandatory activities of the day. Based on the above argument, one notes th at the subjective minimum daily travel time threshold is an unobserved quantity. A tr avel survey provides actual travel time expenditure information, but no information about the persons subjective perception of the minimum travel time that must be undertak en. However, it is very likely that the subjective threshold va lue does influence the actual tr avel time expenditures observed

PAGE 152

137 and measured in travel diary surveys. A suitable methodology for modeling such an unobserved lower bound, in the presence of data on a value that is influenced by the unobserved lower bound, is the stochastic frontie r methodology. In the stochastic frontier methodology, a cost frontier modeling appro ach may be adopted to represent the unobserved lower bound. In the stochastic fr ontier modeling approach, a cost frontier represents the theoretical minimum resources (cost) that a manufacturing plant must spend to meet a production target or goal. In this particular applica tion, this is analogous to the cost frontier representing the theoretical minimum travel time (cost) that a person feels he or she must spend to accomplish a goal, i.e., complete the absolutely required mandatory activity schedule of the da y. The only difference between the two interpretations is that the threshold is consid ered to be a subjective or perceived value in the case of the travel time frontier. The argum ent in favor of this interpretation is made later in this chapter. In this study, stochastic frontier mo dels are developed to estimate the unobservable subjective minimum travel time threshold through a detailed analysis of travel survey data sets derived from thr ee countries the United States, Switzerland, and India. All of the surveys are large samp le travel diary surv eys conducted in 2000-2001 and offer a unique opportunity to examine the notion of the minimum necessary travel time in a global context. 5.2 Modeling Methodology The perceived minimum required travel time constitutes an unobserved frontier that may be estimated using the stochastic frontie r modeling methodology. This section presents the formulation of the stochastic frontier m odeling methodology in this particular context with a modification to censor the unobserved frontier (i.e., minimum required travel time) at zero. Let, ti = i + ui, (1) where i denotes the observation, ti is observed total daily trav el time expenditure and ui is a random component that takes non-negative values. i represents the perceived minimum required travel time (or unobs erved frontier) so that ti is always greater than or equal to i.

PAGE 153

138As mentioned in previous section, the stochastic frontier model (Aigner et al., 1977) is a suitable model that may be applied in this context. However, the traditional stochastic frontier model can not guarantee that the estimated i will be positive. A negative estimate on the subjective minimum travel time threshold is unreasonable and therefore, it would be appropriate to censor the unobserved frontier at zero. To solve this potential problem, one may introduce a latent variable i*. If i* > 0, then i = i* ; on the other hand, if i* 0, i = 0. In this formulation, i can never be negative. Analogous to the traditional stochastic frontier model, let, i* = 'Xi + vi (2) where, vi ~ N(0, v2) and ui is assumed to be distributed half-normal with a probability density function given by, .0,2exp22)(22uuuguu (3) Under the condition that i* > 0, substituting equation (2) into equation (1), one obtains: ti = 'Xi + vi + ui = 'Xi + i. (4) The joint distribution of i has probability density function given by, iii*ii10 |D,2exp)/(2222 (5) where Under the condition that i* 0, one obtains: i = 0 and ti = ui, ./,222vuvu 2222exp22)0|(uiu*iittD (6) where 21 u The unconditional probability density of the observations may be written as: Pr(i* > 0) D1(i | i* > 0) + Pr(i* 0) D2(ti | i* 0) (7) Now, Pr(i* > 0) = Pr('Xi + vi > 0) = ('Xi /v) Also, Pr(i* 0) = 1 ('Xi /v),

PAGE 154

139 where 2v1 Then, the unconditional probability density: 222exp)/(22)/(iiviDiX' 222exp22)]/(1[uiuvtiX' (8) The log-likelihood function for the observations is then: (9) In the log-likelihood function shown in equation (9), , and are the only unknown parameters to be estimated. To obtain the necessary travel time note that E(i*) = If then is zero. Otherwise, if, then, where is the maximum likelihood estimate of As with the standard stochastic frontier model, ni iln(DLL1) iX' 0iX' 0iX' iX' uu2)E( (10) 221)var(uu (11)The maximum likelihood estimation was done for this study using a combination of LIMDEP (Greene, 2002) and GAUSS programming language (Aptech Systems Inc., 1999). A note is due here regarding the interpretation of the stochastic frontier in the context of the perceived minimum required travel time. Under the model formulation presented here, it is theoretically not possible for the actual travel time expenditure to be lower than the estimated frontier because the term ui is greater than or equal to zero. However, unlike in a manufacturing plant operation where the frontier may be considered to be a hard, fixed, and objective threshold value, the frontier in this particular application may be representative of a more loosely defined subjective (perceived) threshold value. This is because, the hard, fixed, and objective minimum threshold value in the context of

PAGE 155

140 travel time expenditure is zero. Regardless of the activity agenda, it is theoretically possible to not engage in travel at all. However, in a beha vioral context such as the one considered in this study, there is likely to be a perceived lower bound that a person feels is representative of his or her minimum tr avel required to accomplish the mandatory activities of the day. This is a more loos e and subjective threshol d value that may be violated in a day-to-day reality. Thus, even though a person feels that he or she must dedicate, say, at least 20 minutes to travel to take care of the absolutely required activities of the day, he or she may (on occasion) end up spending less than 20 minutes for travel depending on constraints, circumstances, and unexpected situations that may arise. For example, the car may break down and the pers on may call work to cancel the important meeting, call the spouse to reallo cate child drop-off/pickup trips, and choose to skip the yoga class. When applied in a behavioral cont ext where the frontier is representative of a subjective perceived and loosely defined thresh old value, violations of the frontier are possible and consistent with expectations. 5.3 Data Sets and Sample Characteristics Three data sets from around the world are used to explore the perceived minimum required travel time expenditure. The three surveys are: 2001 National Household Travel Survey (NHTS) of the United States 2000 Microcensus Travel Su rvey of Switzerland 2001 Household Travel Survey of the City of Thane, India Detailed descriptive statistics of the three data sets are discussed in detail in the previous chapter. However, the table of average travel duration of mob ile adults is reproduced here for the convenience of the reader.

PAGE 156

141 Table 5.1 Comparison of Daily Travel Dura tion by Purpose of Mobile Adults Characteristics US Switzerland Thane, India Travel Duration (min) Com NonComCom Non-Com Com Non-Com Work 26.0035.5 0 31.9 0 Business 9.608.44.1 4.6 0.3 School 0.61.00.41.5 0.0 13.7 Shopping 5.014.05.513.0 0.0 3.7 Social/Recreation 6.319.016.850.2 0.1 4.5 Return Home 31.032.032.243.0 36.1 25.7 Others 15.028.52.23.4 0.1 3.5 Total 93.594.5100.9115.1 72.9 51.3 5.4 Model Estimation Results Censored cost frontier models of the perc eived minimum required travel time are presented in Tables 5.2 through 5.4 for commut er and non-commuter samples in all three data sets. This section presents a brief ove rview of the model estimation results as seen in these tables. Table 5.2 presents the results for the US survey sample s. It is found that the constant term is positive. While this is to be expected because the survey samples are limited to mobile adults, it nevertheless o ffers a first indication that the minimum required travel time frontier tends to be positiv e and that people perceive that they indeed have to engage in at least some amount of travel in a day. In the mobile commuter model, being male, having a college education, being employed full time, living farther away from work, and serving in a prof essional occupation positively impact the minimum required travel time frontier. All of these estimation results are consistent with expectations that highly e ducated and employed males ar e likely to be those who perceive that they have to engage in a cer tain minimum amount of travel to undertake mandatory activities. The weekend days are a ssociated with a reduc tion in the perceived minimum required travel time frontier as in dicated by the negative coefficient. This finding is consistent with expectations as activities on weekends tend to be less mandatory in nature when compared with weekdays. Variables representing the presence of children and higher levels of income are associated wi th a higher minimum required travel time frontier as indicated by the pos itive coefficients. Once again, these findings

PAGE 157

142are consistent with expectations in that households with children and higher incomes may have greater serve-child trip obligations and thus larger minimum travel time requirements. It is interesting to note that the minimum required travel time frontier is positively impacted by the variable representing Friday. It is possible that workers feel that they must travel more on Fridays, relative to other workdays, to relax and enjoy the end of a work week. In other words, some of the flexible/discretionary and joint activities (with other household members) undertaken on a Friday may be perceived as required, thus contributing to the enhanced minimum travel time frontier. Table 5.2 Stochastic Frontier Models of Minimum Required Travel Time: USA Commuters Non-Commuters Variable Coefficientt-stat Coefficient t-stat Constant 18.269 12.1794.711 5.797 Licensed driver ---1.782 2.482Gender: male 2.585 3.777---Education level: Bachelor 3.430 4.428---Employment status: full time 3.310 3.263---Daily work duration (min) -0.026 -12.460---Distance to work from home 0.749 37.533---Occupation: professional 2.313 3.032---Travel day: Saturday/Sunday -7.456 -6.427-0.836 -2.338Travel day: Friday 4.013 4.238---Race: White ----1.725 -3.496Household with children 1.779 2.670-1.071 -2.862Household car ownership >1 -1.905 -2.0560.914 2.049High HH income ($70K/Annum) 4.343 5.633---Low HH income (< $25K/Annum)-4.303 -4.054---Residential neighborhood: urban ---1.157 2.8576.155 23.48031.392 8.487100.370 151.860121.867 191.13 L(C) -86328.8 -115045.0 L() -85655.1 -115067.6 2 [df] 1347.330[12] 45.200[6] Var(v) = 259.047 985.457 9815.029 14851.57 E(u) 79.03 97.22 Var(u) 3569.102 5400.569 Sample Size 15791 20740 2v 2u

PAGE 158

143 For non-commuters, it is found that being a licensed driver has a positive impact on the perceived minimum required travel time. Having a drivers license and higher car availability provide an indivi dual the ability to u ndertake some (or all) of the household mandatory activities, thus leading to a larger minimum required travel time frontier. One interesting difference between commuters and non-commuters is that the presence of children negatively impacts the minimum required travel time frontier for noncommuters. This may be indicative of the non-commuters perception that, because they have greater household obligations and in-hom e childcare responsibilities, their minimum travel time obligation is lower than that for those who do not have the same constraints. In Table 5.3, which shows results of mode l estimation for the Swiss samples, the constant terms are found to be positive once agai n. In addition, as in the US samples, it is found that the constant term fo r mobile commuters is substantially larger than that for mobile non-commuters indicating that, ceteris paribus, the perceived minimum required travel time frontier is larger for commuters than non-commuters. This result is quite reasonable because commuters probably feel th at they have to make the obligatory trips to and from work. All of the other findings ar e generally consistent with expectations. There are a few unique findings here, relative to the US model results. Zero car ownership has a positive impact on the perc eived minimum require d travel time for mobile commuters, but a negative impact on th at for non-commuters. This is presumably because commuters without access to a car have to commute by slower modes. As the commute is generally constrained in time and space, this results in a perception that the minimum amount of travel time required is gr eater than in a situation where a car is available. On the other hand, non-commuters without access to a car may schedule less mandatory activities on their agenda (than noncommuters who have access to a car), and for the few activities that they do schedule on their agenda, they may end up choosing activity locations that are very close toge ther. People in rural areas have a lower perceived minimum required trav el time frontier as do commuters who work part time. Rainy weather is associated with a negative coef ficient; it is likely that travelers think of some travel as not necessarily absolutely requir ed when the weather is not travel-friendly. Fridays once again show a positive impact on the perceived minimum required travel

PAGE 159

144time for commuters. For non-commuters, income is found to make a difference with higher income associated with higher minimum required travel times. Table 5.3 Stochastic Frontier Models of Minimum Required Travel Time: Swiss Commuters Non-Commuters Variable Coefficientt-stat Coefficient t-stat Constant 33.996 5.9273.4564 2.330 Gender: male 4.784 4.278-0.1836 -0.531Age -0.8304 -3.2350.0057 0.089Age squared 0.0084 2.796-0.0002 -0.327Household size 2.7095 2.3910.8970 2.013Household car ownership = 0 5.3813 3.440-0.5270 -1.015Residential neighborhood: rural -2.7889 -2.632-1.1273 -3.220Nationality: Swiss 3.0127 2.294---Employment Status: part time -7.2539 -5.058---Daily work duration (min) -0.0526 -13.973---Logarithm of distance to work from home 13.1321 24.215---Travel day: Friday 4.8602 3.867---Travel day: Sat/Sunday ----0.2510 -0.725Travel day weather: rainy -4.4204 -3.593---High HH income (>Fr 10000 monthly) ---0.0904 0.170Low HH income (
PAGE 160

145 found to have a smaller perceived minimum required travel time as evidenced by the negative coefficient; this is probably due to the higher levels of household obligations that they undertake in the Indian context. In the Indian context, larg er household sizes are associated with smaller minimum travel time frontiers, possibly due to greater in-home childcare obligations and activities. Also, trav eling with the entire family in India is a more burdensome experience than in the deve loped world. This may motivate persons in larger households to lower their perceived mi nimum required travel ti me frontier relative to persons in smaller households. In the I ndian context, is found that those with a professional occupation and higher education level have a lower perceived minimum required travel time frontier. It is likely that these individuals have the resources and means to travel by faster mode s and therefore perceive that their minimum required travel time frontier is lower than others. Indeed, it is found that commuters using transit have a much larger minimum tr avel time threshold as indicated by the large positive and significant coefficient associated with transit use. As transit is a slow mode, commuters using transit in the Indian context feel that their minimum required travel time to accomplish required mandatory activities is quite large. Also as expected, commuters with low income are found to perceive lowe r minimum travel time frontier possibly due to their limited engagements in travel activi ties resulted by resource constraints. Having a drivers license is positively re lated to the subjective minimum required travel time frontier. It is c onceivable that commuters who have a drivers license have the means and resources to travel more, visit pref erred destinations that are farther away, and take on a greater amount of the household ma ndatory activities in the Indian sociocultural context. Commuters with longer work durations have a larger perceived minimum travel time frontier. This is consistent with expectations in that individuals with full-time jobs who work longer also commute longer distances to access the specialized occupations. As a result, these individuals perceive a higher minimum travel requirement. Among non-commuters, younger adults have a higher minimum required travel time frontier. These individuals are genera lly college students in the Indian survey sample with little in-home obligations. As school is a mandatory activity, they exhibit

PAGE 161

146larger minimum required travel time frontier values. It was also found that the total out-of-home activity duration and pursuing more than two activities outside home in the day positively impacted the perceived minimum travel time frontier. It is likely that these individuals are those who undertake more of the household obligations outside home and therefore have higher thresholds for the minimum required travel time. Table 5.4 Stochastic Frontier Models of Minimum Required Travel Time: India Commuters Non-Commuters Variable Coefficientt-stat Coefficient t-stat Constant 20.823212.2606.7072 5.204 Male 3.72973.592---Occupation: service -----4.5045 -2.031Occupation: professional -5.8862-5.327---Occupation: homemaker -----3.1035 -3.086Education: graduate -4.2387-2.938---Daily work activity duration 0.00613.669---Licensed driver 2.66972.387---Household size -0.3855-1.632-0.4696 -1.718Low income ( Rs. 5K monthly) -1.0333-1.218---Young (age: 18-29) ----2.5705 2.818Daily activity duration ----0.0030 2.446No. of daily activity participated > 2 ----8.3298 2.957Transit user 63.710440.680--0.210827.51019.2962 4.30258.708961.45761.7544 54.086 L(C) -24169.044 -8315.246 L() -22752.046 -8289.761 2 [df] 2833.996 [8] 50.970 [6] Var(v) = 57.387 10.2147 146.346 3803.391 E(u) 9.65 49.21 Var(u) 53.179 1382.077 Sample Size 4623 1699 2v 2u In all of the models, the statistical goodness-of-fits are quite reasonable and the coefficients are statistically significant offering plausible interpretations. The model estimation results suggest that the stochastic frontier modeling methodology offers a

PAGE 162

147 suitable framework for studying unobserved frontiers related to minimum required travel time expenditures. 5.5 Distribution of Travel Time Expenditures and Cost Frontiers The stochastic frontier models presented in Tables 5.2 through 5.4 can be used to derive distributions of expected valu es of perceived minimum requi red travel time frontiers. These distributions are plotted together with the distributions of the actual travel time expenditures for commuter and non-commuter samples in each of the three data sets. The resulting plots offer an interesting pe rspective on the distributions of expected minimum required travel time frontiers and how they are related to the observed travel time expenditures. These distributions also show the proportion of individuals for whom the expected perceived minimum required trav el time frontier is zero minutes, i.e., the proportion of individuals for whom travel coul d potentially be entirely eliminated. Figure 5.1 shows the distributions of exp ected minimum required travel time and actual travel time expenditures for the US surv ey samples. The cost frontier distributions are the sharp and highly peaked distributions that are generally shifted to the left of the actual travel time expenditure distributions. An examination of the plots shows that the stochastic frontier models offer predictions that are plausible and beha viorally intuitive. The minimum required travel time distributions are quite well defined and show sharp peaks in the 0-30 minute range. As 100 percent of the non-commuter sample fell into this category, an inset graph showing a more detailed distribution for this group is also shown in the figure. For this market segment, the distribution is generall y in the range of 0-10 minutes. Consistent with these findings, the expected value of the minimum required travel time frontier is found to be about 23 minutes for commuters and 6 minutes for noncommuters. This is consistent with the expectation that commuters have a larger subjective minimum travel time requirement due to the need to commute to and from work, which is generally considered a mandato ry activity. An intere sting finding is that virtually nobody has a zero-minute minimum ex pected travel time frontier. In other words, virtually all individuals in the US survey samples feel that they must travel for at least a certain minimum level of duration to accomplish the required activity schedule of

PAGE 163

148 the day. Conversely, almost nobody in the sample feels that he or she can eliminate all travel and stay at home all day. Figure 5.2 shows the distributions for the Swiss survey samples. Once again, an inset graph is provided to show a more detailed distribution for the non-commuter sample. Similar to the United States, the Swiss samples show an expected minimum required travel time frontier of about 21 minutes for commuters and 3 minutes for noncommuters. It is to be noted that these values are smaller than those in the United States samples even though the actual travel time expe nditures are greater in the Swiss samples than in the US samples. It is possible that the higher densities and destination choices in the Swiss context offer the ab ility to lower minimum require d travel time thresholds. Another interesting finding is that, in the Swiss commuter sample, nearly 12 percent of the commuters have a perceived minimum tr avel time frontier of zero minutes. This finding merits further investigation in that co mmuters should generally consider the work activity and its associated tr avel as mandatory and one would expect virtually nobody in the commuter market segment to have a zer o-minute minimum travel time threshold. On the other hand, in the Swiss context, it is po ssible that a small segment of the commuter sample has a work arrangement flexible enough that allow zero minute minimum travel time thresholds. Finally, Figure 5.3 presents distributions fo r the Indian survey samples. It is found that no commuter shows an expected mi nimum travel time frontier value of zero minutes. An inset graph is provided to show detailed distributions; however, in this case, detailed distributions are shown in the inset graph for both commuters and noncommuters. While the non-commuters show an expected minimum required travel time frontier of about 6 minutes similar to the US and Swiss samples, the commuters show an expected value of about 60 minutes which is much higher than that seen in the US and Swiss samples. This finding is actually qui te consistent with the nature of travel undertaken by commuters in the Indian survey sample. The commuters in the Indian survey samp le made, on average, two trips per day the trips to and from work. These are esse ntially mandatory trips. Thus, one would expect that all (or nearly all) of the observ ed travel time expenditure is absolutely

PAGE 164

0510152025303540455055606570758085909510000~3030~6060~9090~120120~150150~180180~210210~240240~270270~300300~330330~360360~390390~420420~450450~480480~510510~540540~570570~600600+Time (Minutes)Percentage Commuters Travel Time Expenditure (Average = 93.5 min) Commuters Minimum Travel Time Frontier (Average = 23.1 min) Non-commuters Travel Time Expenditure ( Average = 94.5 min) Non-commuters Minumum Travel Time Frontier (Average = 6.1 min) 05101520253035404501~22~33~44~55~66~77~88~99~10Time (Minutes)Percentage Close-up of Non-commuters Minimum Travel Time Frontier Figure 5.1 Distributions of Travel Time Expenditures and Expected Minimum Required Travel Time Frontiers: USA Mobile Commuters and Non-Commuters 149

PAGE 165

0510152025303540455055606570758085909510000~3030~6060~9090~120120~150150~180180~210210~240240~270270~300300~330330~360360~390390~420420~450450~480480~510510~540540~570570~600600+Time (Minutes)Percentages Commuters Travel Time Expenditure (Average = 100.9 min) Commuters Minimum Travel Time Frontier (Average = 20.9 min) Non-commuters Travel Time Expenditure (Average = 115.1 min) Non-commuters Minimum Travel Time Frontier (Average = 2.66 min) 0510152025303540455000~11~22~33~44~5Time (Minutes)Percentage Close-up of Non-commuters Minimum Travel TimeFrontier Figure 5.2 Distributions of Travel Time Expenditures and Expected Minimum Required Travel Time Frontiers: Swiss Mobile Commuters and Non-Commuters 150

PAGE 166

0510152025303540455055606570758085909510000-3030-6060-9090-120120-150150-180180-210210-240240-270270-300300-330330-360360-390390-420420-450450-480480-510510-540540-570570-600600+Time (Minutes)Percentage Commuters Travel Time Expenditure (Average = 72.95 min) Commuters Minimum Travel Time Frontier (Average = 59.53 min) Non-commuters Travel Time Expenditure (Average = 51.32 min) Non-commuters Minimum Travel Time Frontier (Average = 6.35 min) 05101520253035404505-67-89-1020-2530-3550-7580-8590-95100-120130-140Time (Minutes)Percentage Close-up of Commuters Minimum Travel Time Frontier Close-up of Non-commuters Minimum Travel Time Frontier Figure 5.3 Distributions of Travel Time Expenditures and Expected Minimum Required Travel Time Frontiers: India Mobile Commuters and Non-Commuters 151

PAGE 167

152 necessary and mandatory in na ture. The actual travel time expenditure averages about 75 minutes for this sample segment. Considering that most of it is mandatory, the 60 minute value for the expected minimum travel time requirement is quite consistent with expectations. It is to be noted that the hi gh level of transit usage and dependency may be leading to the high minimum tr avel time frontier for a large segment of the commuter sample in the data set from India. Table 5.5 summarizes the international comparison of average travel time expenditures and expected minimum travel time frontiers. In th is table, the value of E[u] shows the average difference between the actual travel time expenditure and the expected minimum required travel time frontier. In gene ral, it can be seen that the actual travel time expenditure far exceeds the expected mi nimum required travel time frontier in all cases, except in the case of Indian commuters. The E[u] value for Indian commuters is only about 10 minutes while the correspondi ng values for US and Swiss commuter samples are 80 minutes and 95 minutes respecti vely. This clearly reflects the effects of the maturity and performance of the tran sportation system on activity and travel engagement. In developed countries, traveling offers a disutility that is small enough to motivate substantial activity engagement (a nd therefore, travel) above and beyond the perceived minimum required travel. However, in developing countries, the disutility of traveling is still so large th at additional activity engageme nt (and travel) is undertaken more sparingly. Even for non-commuters, wh ereas E[u] is 97 minutes and 130 minutes for the US and Swiss survey samples, it is only 49 minutes for the Indian non-commuter sample. Another finding to note is that, in all three samples, the E[u] value is greater for non-commuters than for commuters. This is c onsistent with the no tion that a greater proportion of travel undertaken by non-comm uters is for discretionary purposes and therefore non-commuter samples offer the potent ial for greater reductions in travel in the event of a TDM or TCM implementation.

PAGE 168

153 Table 5.5 International Comparison of Average Travel Time Expenditures and Average Expected Minimum Required Travel Time Frontiers Commuters Non-commuters Survey Area Travel Time Expenditure (min) Min Required Travel Time (min) E(u) (min) Travel time expenditure (min) Min Required Travel Time (min) E(u) (min) US 93.5 23.179.094.5 6.197.2 Switzerland 100.9 20.994.8115.1 2.7131.1 Thane, India 72.9 59.539.6551.3 6.449.2 5.6 Summary and Discussions This study has presented a series of mode ls to better understand daily travel time expenditures by attempting to determine the subjective or perceived minimum amount of travel time that people feel they must undertake to accomplish the minimum required activities of the day. An unders tanding of the minimum travel time threshold would offer the potential to quantitatively assess the maximum amount of travel reduction that may potentially be accomplished through TDM and TCM policy implementation. The study postulates that the stocha stic frontier modeling met hodology can be employed to determine the minimum required travel time fron tier that influences the actual travel time expenditure observed in travel surveys. The mi nimum travel time threshold is represented by an unobserved cost frontier in this modeling framework. Cost frontier models of minimum required travel time frontier were estimated for three survey samples drawn from the 2001 US National Household Travel Survey (NHTS), 2000 Swiss Microcensus Travel Surv ey, and the 2001 Thane, India household travel survey. Separate models were estimated for commuters and non-commuters to recognize the inherent differences between these market segments. In addition, the analysis was limited to mobile adult samples to control for unknown factors related to the survey design and reporting. The stochastic frontier models were found to offer statistically significant coefficients for several socio-economic and demographic characteri stics indicating that the minimum required travel time frontier is likely to be influenced by a persons lifecycle stage, lifestyle, income, age, and household characte ristics. The model

PAGE 169

154 estimation results were used to plot distributions of the e xpected necessary travel time frontier values vis--vis the actual travel time expenditures. The plots were found to offer plausible interpretations and suggested that the stochastic frontier modeling methodology is suitable for modeling minimum required tr avel time frontiers. The average expected minimum required travel time frontier values were found to be about 20 minutes for US and Swiss commuters, 60 minutes for I ndian commuters, and 3-7 minutes for noncommutes. The expected minimum travel time frontier is greater than zero minutes for a vast majority of the individuals in all survey samples. This finding suggests that individuals, in general, feel that they have to undertake at least a certain amount of minimum travel to fulfill the minimum required activity agenda.

PAGE 170

155 CHAPTER 6 ANALYSIS OF COMMUTING LEN GTH CHOICE BEHAVIOR The previous chapter has presented an anal ysis of travel time expenditure by modeling theoretical minimum travel time thres hold for commuter and non-commuter samples around the world. Comparison of the model esti mation results has suggested significant effects of socio-economic, demographic and cultural attributes on the amount of travel that a person feels he/she must undertake to accomplish th e minimum required activities of the day. The current chapter offers a deta iled analysis of commute travel time choice behavior around the world. 6.1 Introduction Understanding commuter behavior has been a su bject of interest in the transportation profession for many years. At this present tim e, conducting research in this subject has never been so important in the developing country context. Subject to the increasing population pressure, economic growth, auto ownership and other undergoing societal changes over the past years, many urban area s in the developing c ountries like India are experiencing rapid suburbanization and disp ersion of job and residential locations. If commuting length is considered to be the ma nifestation of physical separation between home and work locations, then the commuting pattern of the people is certainly expected to be reconfigured in response to such changes in job or residential locations. Therefore, better understanding of commuter travel behavi or and preferences has become extremely important by taking into account the impli cations of changing commuting patterns on future transport policy and planning options in the developing nati ons. Over the past decades, many studies have addressed these issu es in the developed country context. But, virtually no knowledge is available from a developing country standpoint.

PAGE 171

156 In the travel behavior literature, commuti ng is portrayed as a behavioral process that is about the awareness, attitudes, perceptions and opti ons of the people who travel regularly on a daily basis to perform a desire d or necessary activit y. Essentially, the study of commuter behavior is the understandi ng of peoples commuting preferences under certain circumstances in order to maximize th e perceived utility of a commute alternative to perform a particular or sequence of ac tivities (Vaddepalli, 2004). Pisarski (1996) recognized many dominating factors in the Amer ican society, such as increased number of two-worker households, boom in long distan ce travel, urban form and private vehicle usage that influenced contemporary commuti ng characteristics and preferences in mid 90s. With ever increasing interest in modeling and simulating activity-travel patterns and time-use behavior, the study of commuter be havior has drawn special interests from travel behavior researchers. Particularl y, understanding the trade-off between commute and discretionary travel and its implications on quality of life has gained new momentum in the travel behavior pr ofession. For instance, a long commuter spending more time on travel is likely to experience greater stress and difficulty in finding free time to spend with friends and family and that leads to an improper work-life ba lance entailing a poor quality of life. Furthermore, a current body of literature (Ewing and Cervero, 2001; Ewing et al., 2003; Handy et al., 2002; Hoe hner, et al. 2003) have investigated the relationship between commuting pattern a nd natural and built environment. The researchers have expressed their concer ns about changing pa tterns of commuting behavior and its increasing eff ects on public health issues lik e obesity and other diseases. The studies have indicated that auto-oriente d transportation infrastructure, dispersed landuse patterns are some of the major propone nts in the tremendous growth of autodependent commute and the resulting effects of all these fact ors are significantly contributing in promoting inactive life-s tyle and prolonged exposure to vehicular emission resulting in potential he alth problems including obesity at an ep idemic scale. A growing body of literature has addresse d the connection between urban spatial structure and travel behavior Many economic theories in 1960s have emphasized upon the trade-off between commuting cost and housi ng costs and placed this trade-off at the

PAGE 172

157 core of regional science and urban ec onomics (Wingo, 1961; Ka in, 1962; Alonso, 1964; Muth, 1969). Many studies have indicated the impacts of land use on commuter preferences like mode choice, vehicle ownershi p, and trip length and whereas, some other studies have addressed the association of commute length, and residential and work locations with sustainable urban developm ent (Boarnet, 2001; Krizek, 2002; Waddell, 2002; William, 2003). All these studies have further motivated some researchers to obtain more insights into the characteristics of co mmuting behavior towards a goal to achieve more sustainable cities by targeting VMT th at involves reducing commuting trip lengths (Pez, 2001). The above discussion leaves no doubt about the compelling importance of understanding commuter behavior as the subject which is increasing ly associating with the interests of several othe r research areas including ac tivity and time use studies. The intent of this present research is to anal yze the commute time choi ce behavior. The study is based on the hypothesis that commute time reflects an individuals preference for his/her residential and workplace locati ons. These preferences in turn are the manifestation of an indivi duals decision-making processes that are governed by their socio-economic, demographic characteristics a nd personal traits and these attributes are capable of partially explaining an individuals c hoice of long, medium or short commute time. The main objective of this research is to finding out who are the long, medium and short commuters and what factors are contribut ing to the choice of commute travel time. Analyzing the restrictions and preferences that lead an individual to be a particular type of commuter would help for making better pol icies to provide mobility options and job accessibility for those in real needs. Three mu ltinomial logit models have been estimated on three market segments short, medium a nd long commuter using three data sets from US, Swiss and Thane, India. Comparison across the study results in an international set up has facilitated a better understanding of the commute time choice behavior in a diverse socio-cultural context.

PAGE 173

158 6.2 Data Sets Three data sets from around the world are us ed in this study. Th e three surveys are: 2001 National Household Travel Survey of the United States 2000 Microcensus Travel Su rvey of Switzerland 2001 Household Travel Survey of the City of Thane, India The household and person characteristic s of the three survey samples are presented in the preceding chap ter (refer Chapter Five). Howe ver, detailed description of the demographic and socio-economic characteristics of the commuter segments derived from the three data sets are di scussed in the following section. 6.3 Descriptive Statistics of the Commuter Segments Table 6.1, Table 6.2 and Table 6.3 offer a deta iled look at the person characteristics by commuting length status for US, Swiss and Indi an data sets respectively. All individuals who made at least one work tr ip or work-related business trip on the travel survey day are treated as commuters. All commuters are furthe r classified into three market segments based on their commute time: short, medium and long. Commute time of a person is defined by the one-way travel time between his home and work location or otherwise. Commuters who reported only nonhome-based work trips are excluded from the sample. Short commuters are defined as those indi viduals who commute 15 minutes or less to work. Medium commuters are defined as t hose individuals who commute more than 15 minutes but less than 60 minutes to work. Fi nally long commuters are defined as those who commute 60 minutes or more to work. Thes e definitions are applied consistently to all of the three data sets. Also noteworthy is that the analyses presented here are restricted to only adult commuters who are 18 years of ag e or above. It was felt that the analysis should be limited to the adult samples because children often do not have the freedom to choose their travel patterns, ha ve less flexibility with respec t to their travel options and decisions, and have poorer response quali ty in travel survey data sets. Table 6.1 provides the descriptive statistics of commuters for the US sample. It shows that short and medium commuters comp rise about 94 percent of all commuters.

PAGE 174

159 The gender split is consistent among all th e commuter categories dominating by males though the difference for short commuter sample is very marginal. This implies that male commuters have freedom to travel l onger distance than females to access job opportunities whereas female commuters are mos tly working close to their home to keep balance between their working life and househ old responsibilities. W ith respect to age distribution, young commuters are predominantl y short commuters, but a significant percentage of the commuters in their medium age range or around retirement age are traveling for a medium to long period of time. This indicates that when young people enter the work force prefer to work near home but as they gain experience and achieve financial freedom, their preferences for life styl e tend to change and th is influences their ability to commute long distances to satisfy their needs. On the other hand, when a person retires, essentially low wage and commuting st ress restrict him to travel shorter. The educational attainment is the next important factor for commute travel time choice as clearly observed from the US data set. Highly educated people have more freedom in job and residential location choices. Individual in comes are also proporti onal to educational attainment. Therefore, highly educated peopl e would have more freedom for residential location in a low-density and suburban lifestyle which are generally at the outskirt of the main city. This could be a valid reason be hind why the share of moderately and highly educated people is present with greater share in long commuter category in the US sample. With respect to income distribution, it is found that majority of the lower income people are making short commute. The share of moderate to higher-income people is dominant in long and medium commuter category than in short commuter category. Once again, this observation supports our expecta tion that financial freedom influences the ability of individuals to travel more to satisfy their lifestyle needs. From the driver status perspective, interestingly it is found that share of the drivers in the long commuter category are less (91.1%) when compared to short commuter category (95.2%). This may be because regular transit users who are long commuters fall into n ot a driver category. The table also shows that average one-way commute time of short, medium and long commuters in the US sample are about 10, 30 and 80 minutes respecti vely.

PAGE 175

160 Table 6.2 provides descriptive statistics for the commut ers in the Swiss Sample. Similar to the US sample, it is found that a ma jority of the commuters fall into the short category and a very small share at about 7 percent is long commute r segment. The gender spilt in the Swiss sample is also very consiste nt with the US sample as majority of all commuter segments are male. Again with respect to age distribution, middle age groups are found to be dominant in long commuter category relative to all other age groups while the trend in young age groups is not ve ry clear in Swiss sample. The share of commuters in their retirement age group is hi gher in short category though. Persons with higher education are traveling longer than other groups as expected. Commuters with full-time employment are found to undertake longer travel while part-time employed prefer to work closer to their home. Ge nder role may be a valid reason behind this commuting preference. Interestingly, the share of drivers license in the Swiss sample is exactly similar to the US sample. The share of licensed driver in long commuter category is less than that in short commuter. Again, higher percentage of transit users having no drivers license in long commut er category may be the governi ng factors in this case. The average commute time of Swiss short, medi um and long commuters are about 10, 30 and 112 min respectively. Finally, Table 6.3 presents the descriptiv e statistics of the Indian commuter samples. Interestingly, long commuters take a la rge share (about 25 percent) in the Indian commuter sample. In a transit oriented society like India, long commute time is very common due to usage of slow public transportation. Similar to the US and Swiss samples, majority of the commuters are males in the I ndian context with a share of more than 80 percent in all of the three categories. The gende r role is extremely pr ominent in this case, where traditionally males are employed and they are the only earning members in the household while females bear major household and childcare responsibilities. From the perspective of age distribution, similar to th e US and Swiss contexts young commuters in India are commuting closer to home while commu ters at their mid and retirement age are traveling for a longer period. Financial freedom and lifestyle needs are still the major contributing factors behind commuting choices in the Indian society. With respect to occupational distribution, the statistics shows that individuals who wo rk in service sectors

PAGE 176

161 like government and private agencies, financial institutions, IT firms etc. are commuting long and they share about 85% of all long commuters. This is because these people generally commute from their suburba n residence to CBD area using public transportation and that makes their commute time long. On the other hand, in the Indian context, businessman or professionals are th ose who generally own small retail shops or enterprise closer to their home. This market segment share a considerable portion of short commuter sample. As expected, Indian colleg e graduates are traveling longer while low educated Indians are short commuters. Auto ownership is very lo w in India while twowheelers are very popular in the Indian societ y as they are more affordable for middle class or lower middle class people. In gene ral, two-wheeler user s are generally short commuters as Indian roads are not very conducive for long travel by two-wheelers. Long commuters are potentially transit users. Simila r to the US and Swiss samples, majority of the Indian long commuters fall into high income category while majority of the low income individuals are making short commutes. With respect to license status, consistent with the US and Swiss contexts the share of licensed driver in the Indian long commuter category is less than that in the short co mmuter category. In the Indian sample, the average commute time of short, medium and long commuters are about 10, 33 and 78 minutes respectively. In summary, descriptive statistics are quite similar in all three data sets. Gender roles are clearly reflected in each data set in a very similar way where majority of the commuters in all the three categ ories are male; however, the percentages are much larger in the India sample at about 80 percent. Y ounger people are traveling shorter while older people are commuting longer. Highly educated and affluent individuals are found to be long commuters while low educated and low inco me people may prefer to work closer to home. A vast majority of individuals is licen sed to drive in the US and Swiss samples, while only a very small minority in licensed to drive in India. However, the share of licensed driver in the long commuter category is consistently lesser than that in short commuter segment. The share of long commuters in the India sample is found considerably higher than the ot her data sets. This essentiall y reflects a clear distinction between a transit-oriented Indian society and an auto-oriented western society. More

PAGE 177

162 Table 6.1 Person Characteristics of Commuter Groups: USA Characteristic Short Commuters Medium Commuters Long Commuters All Commuters Sample Size 11876 (47.6%) 11641 (46.7%) 1431 (5.7%) 24948 (100%) Gender Male 50.5%55.1%63.6% 53.5% Female 49.5%44.9%36.4% 46.5% Age 16 to 20 years 11.2%5.0%4.0% 7.9% 21 to 24 years 7.8%7.4%7.7% 7.6% 25 to 44 years 47.5%53.1%51.9% 50.5% 45 to 64 years 30.0%31.9%34.0% 31.1% 65+ years 3.4%2.5%2.5% 2.9% Education < high school 11.3%7.3%9.2% 9.3% High school 30.7%27.9%27.5% 29.2% Tech. training 3.5%4.0%3.8% 3.8% College 19.3%18.0%14.5% 18.4% Associate 7.1%8.9%6.7% 7.9% Bachelor 16.3%19.8%22.8% 18.4% Graduate school 1.9%2.0%2.1% 2.0% Graduate degree 9.7%12.2%13.3% 11.1% Income (USD) <15,000 41.0%28.0%12.0% 33.2% 15,000 19,999 11.0%10.0%15.9% 10.8% 20,000 24,999 12.3%10.7%14.7% 11.6% 25,000 49,999 24.3%30.7%33.6% 27.8% 50,000 74,999 5.5%12.7%7.5% 9.0% 75,000 99,999 3.7%4.6%4.1% 4.2% 1000,000 2.4%3.7%12.3% 3.5% Driver License Licensed 95.2%95.2%91.1% 95.0% Not licensed 4.8%4.8%8.9% 5.0% Trips/day 5.034.594.10 4.8 Commute Time (min) 9.829.778.3 23.5

PAGE 178

163 Table 6.2 Person Characteristics of Commuter Groups: Swiss Characteristic Short Commuters Medium Commuters Long Commuters All Commuters Sample Size 4272 (52.6%) 3269 (40.2%) 586 (7.2%) 8127 (100%) Gender Male 58.7%57.0%67.1% 58.6% Female 41.3%43.0%32.9% 41.4% Age 15 years 0.4%0.2%0.2% 0.3% 16 to 19 years 3.2%4.6%5.5% 3.9% 20 to 24 years 5.6%7.5%6.8% 6.5% 25 to 44 years 47.7%49.5%47.8% 48.4% 45 to 64 years 40.9%36.8%38.1% 39.1% 65+ years 2.1%1.4%1.7% 1.8% Education Unknown 0.4%0.4%0.5% 0.4% < High school 2.1%1.7%1.2% 1.9% High school 13.1%11.2%13.0% 12.3% Tech. training 49.1%45.9%44.4% 47.5% Other college 19.4%21.1%18.8% 20.1% Bachelor 5.2%8.4%7.7% 6.7% Masters/PhD 10.7%11.3%14.5% 11.2% Employment Unknown 0.3%0.2%0.7% 0.3% Full-time 73.4%75.5%77.1% 74.5% Part-time 21.3%18.8%15.7% 19.9% Homemaker 0.7%0.3%0.3% 0.5% Student 3.5%4.9%5.5% 4.2% Retired 0.7%0.2%0.7% 0.5% Other situation 0.1%0.1%0.0% 0.1% Driver License Unknown 2.1%2.2%2.7% 2.2% Licensed 88.8%85.5%82.9% 87.1% Not Licensed 9.1%12.3%14.3% 10.8% Trips/day 4.984.253.90 4.61 Commute Time (min) 9.230.51111.39 25.14

PAGE 179

164 Table 6.3 Person Characteristics of Commuter Groups: India Characteristic Short Commuters Medium Commuters Long Commuters All Commuters Sample Size 1567 (34.0%) 1923 (41.8%) 1116 (24.2%) 4606 (100%) Gender Male 83.2%85.9%87.8% 85.4% Female 16.8%14.1%12.2% 14.6% Age 18 to 20 years 5.0%4.4%3.1% 4.3% 21 to 24 years 11.3%10.8%10.0% 10.80% 25 to 44 years 56.5%57.4%57.1% 57.0% 45 to 64 years 26.3%26.8%29.0% 27.2% 65+ years 0.80%0.60%0.70% 0.7% Education Illiterate 7.4%4.6%2.9% 5.1% Up to SSC 57.6%54.8%44.5% 53.3% Up to HSC 12.3%11.9%13.0% 12.3% Graduate 22.7%28.8%39.6% 29.3% Occupation Service 57.6%74.7%84.1% 71.1% Farmer/Laborer 8.7%7.9%3.0% 7.0% Professional 33.1%16.9%12.6% 21.4% Student 0.30%0.3%0.2% 0.2% Homemaker 0.30%0.1%0.0% 0.1% Retired 0.20%0.2%0.1% 0.2% Income (INR) No income 2.6%2.7%1.6% 2.4% 5,000 60.7%58.9%50.0% 57.4% 5,001-15,000 34.8%36.9%45.7% 38.3% 15001+ 1.9%1.5%2.7% 1.9% With driver license 20.5%16.4%16.6% 17.8% Trips/day 2.142.022.01 2.06 Commute Time (min) 9.7133.0978.57 36.16

PAGE 180

165 interestingly, the average commute time of short, medium and long commuters are found strikingly similar in all thr ee data sets except Swiss long commuters are traveling 30-35 minutes longer than the US and Indian long commuter samples. With respect to trip frequencies, commuters in the US and Swiss samples make more than four trips per day while the commuters in India make, on average, less than one-half as many trip at just about two trips per day indicating that th ey generally go to work (or work-related business) and then return home. 6.4 Model Estimation Results Multinomial logit models (MNL) (Ben-Akiva and Lerman, 1985) were estimated for the three data sets to analyze the influenc e of individual, household and area related characteristics on the commute time choice. In all the models, the utility function of medium commuter type has been set to zero as the base alternative. The utility functions for short and long commuters were first defined by including all possible best combination of characteristics available in the corresponding datasets The variables were tested for their significance at 95 percent level of confidence by running the models in LIMDEP 8.0 (Greene, 2002). All the significant variab les were retained and the models were tested for good-of-fit using standard test-statistics. MNL estimation results are presented in Table 6.4 through 6.6 for the three survey samples. This section presents a brief overview of the model estimation results as seen in these tables. Table 6.4 presents the model estimation re sults for the US survey sample. The alternative-specific constant for short commuter is appeared significant and positive while the constant for long commuter is appe ared significantly negative. These indicate that overall propensity of the US commuters is to commute for shor t duration and not to travel so long to work. As expected, male shows a negative coefficient in the short commuter utility function and positive in long commuter utility func tion indicating that males are generally engaged in long commute or lesser tendency to commute short. Age as a continuous variable is appeared to ha ve negative coefficien t on short commuters utility function, which implies that as age increas es a commuter is more likely to travel shorter. The dummy variable for white Am erican race group shows positive coefficient

PAGE 181

166Table 6.4 Commute Time Choice Model: USA Short Commuters Long Commuters Variable Variable Type -Coefft-stat-Coeff t-stat Constant 0.37083.214-1.905 -12.277 Male Dummy -0.1501-5.3350.4375 7.084 Age Continuous 0.00191.695--Mid age Age: 25-64 years -0.4387-10.526--White Race: white 0.12443.352-0.1373 -1.918 Well Educated Bachelor degree -0.1182-3.781--Driver Dummy -0.2089-2.333-0.774 -5.749 Professional Occ: Professional -0.2398-7.9980.2891 4.374 HH size Continuous -0.1038-3.261--Low HH inc HH inc <$15,000 0.20773.241--High HH inc HH inc $75,000 -0.1491-3.7250.1293 1.781 No. of children Continuous 0.13924.343--No. of drivers Continuous 0.13994.517--Area type urban cluster 0.467213.984-0.3311 -4.607 Area pop. pop 3 million -0.5198-16.4710.9112 14.221 Log likelihood function at convergence L() =-19938.7 Log likelihood function w/o constant L(C) =-20690.0 Log likelihood function w. const. only L(0) =-26070.0 [df] = 1503.4[23] 2

PAGE 182

167 associated with short commuting and nega tively associated with long commuting. The fact is, the average auto ownership among white population is predominant than any other races in the US (78.8% of all auto owners are white: NHTS 2001), which may be the major contributing factor for whites to commute shorter. Though driver status is negatively associated with both the short and long commu ter utility function and the magnitudes of the coefficients suggest that dr ivers are more likely to be short commuters than long commuters. Professionals and pe rsons with high hous ehold income are negatively associated with short commute r and positively associated with long commuters. This is consistent with the expectation because people with high job profile and affluent household are likely to live further from city center/job location and that essentially contribute s longer commuter time. Similarl y, household size as a continuous variable is found to have negative impact on short commute. On the other hand, low income household and presence of children in the household are consistently found to have positive propensity to commute shorter. All these effects appear to suggest that household constraints like financial constraint, childcare responsibilities potentially increase the chances of being a short commu ter. Interestingly, propensity of short commuting increases with number of drivers in the household. This is possibly because as the number of drivers in the household incr ease, the chances of driving alone to work increases as opposed to carpool with ot her workers in the household, which could potentially shorten the commute length (t ime) for each individual household member. Commuters who live in a dense neighborhood like urban cluster or center are most likely to commute shorter and this finding is also consistent with ex pectation. Table 6.5 presents the MNL model estim ation results for the Swiss sample. The alternative-specific constants in both s hort and long commuter utility functions are appeared negative and statistically significan t. However, overall tendency of the Swiss commuters is appeared to undertake shorter commute similar to the US context. Once again, males are associated with long commute As females are likely to be bearing a greater share of the household and childcare re sponsibilities, this finding is consistent with expectations. Similar to the US contex t, age is found to have positive impact on short commute. Higher household income and au to ownership are associated with short

PAGE 183

168 commute similar to the US context while low household income and absence of any automobile in the household show positive propensity for long commuting. This is consistent with expectations because low in come and no auto availability could be the potential reasons for those commuters to use transit and that may cause longer commute time relative to those who drive. As expecte d, highly educated commuter is found to have a negative tendency to undertake short commut e. On the other hand, variables like parttime or self-employed and household size are positively associated with short commute. These effects are appeared to be consistent with expectations and exhibit similar effects relative to the US context except household size has opposite effect in the US model. Finally, Table 6.6 presents the model for the India sample. Once again, the alternative-specific constant associated with long commuter utility function is appeared statistically significant and e xhibit negative sign similar to the US and Swiss contexts while the constant in the short commuter util ity function did not appear significant and was decided to remove from the model. This finding implies that in general Indian commuters have a propensity to not commute for a long period of time and that is consistent with our expectation. Individuals owning small businesses are likely to take short commute while those employed in serv ice sector are appeared to have long commute time. As discussed earlier that priv ate business owners in the Indian context generally commute to their workplaces closer to their residential locations and on the other hand, individuals employe d in service sectors need to undertake long commute to the Central Business District located at th e heart of the city. Individuals with low personal income is found to be negatively a ssociated with both short and long commute and the coefficients suggest th at they are more likely to be short commuters than long commuters. Similar to the US and Swiss contexts, highly educated individuals are found to be negatively associated with short commute and positively associated with long commute. All these effects are attributable to the fact that these individuals possess a greater level of affordability to travel l onger distance and flexibility of choosing job locations further from home. As expecte d, younger individuals and availability of twowheelers in the household are found to have negative effects on long commute. Persons in their middle age group are less likely to undertake short commute. Owning a home,

PAGE 184

169Table 6.5 Commute Time Choice Model: Swiss Short Commuters Long Commuters Variable Variable Type -Coefft-stat-Coeff t-stat Constant -0.3138-3.088-2.0206 -15.076 Male Dummy ---0.4365 4.596 Age Continuous 0.00924.811 --Young age Age 18 ---0.1776 1.674 High HH inc HH inc > Fr. 10,000 -0.2877-4.357 --Low HH inc HH inc < Fr. 4000 ---0.2288 1.759 High educated Education: Bachelor -0.2312-3.749 --Part-time emp Dummy; 0.14632.508 --Independent Occu: Independent 0.80439.9140.6233 4.200 Free parking Parking free at work ----0.6505 -4.915 No reserved parking at the workplace ----0.2698 -2.277 Employed in the middle/lower cadre ---0.2050 1.906 HH size Continuous 0.07664.185 --Rural resident Dummy 0.26734.8790.4017 3.962 No. HH autos Continuous -0.3472-4.4960.3853 3.005 HH auto 2 Dummy -0.1588-3.056 --Log likelihood function at convergence L() =-7084.417 Log likelihood function w/o constant L(C) =-8928.422 Log likelihood function w. const only L(0) =-7265.430 [df] =362.026[20] 2

PAGE 185

170Table 6.6 Commute Time Choice Model: India Short Commuters Long Commuters Variable Variable Type -Coefft-stat-Coeff t-stat Constant -----1.07321 -7.567 Business Occu: Business 0.7376149.815--Service Occu: Service ----0.616989 6.411 Low income Inc: Rs. 5000 -0.25935-5.353-0.37760 -4.585 High educated Edu: Graduate -0.56779-7.6250.360855 4.139 Young age Age: 18 -30 yrs -----0.20166 -2.584 Elderly Age: > 45 yrs -0.14796-1.993--Home owner Own a home ----0.343859 3.695 Female with children in the household 0.4171693.254-0.47114 -2.811 HH 2-wheelers Continuous -----0.48797 -4.941 Smaller HH HH size 4 ----0.158957 2.173 Log likelihood function at convergence L() =-4754.227 Log likelihood function without constant L(C) =-4951.2752 Log likelihood function w. constant only L(0) =-5060.2082 [df] =611.96[14] 2

PAGE 186

171 which is closely related to moderate or higher household income shows positive association with long commute and this is certainly consistent with expectations. Consistent with expectation, female commute rs with children in the household show higher propensity to commute closer to home as they are more likely to be bearing a greater share of the household and childcare responsibilities. On the other hand, an individual who belongs to a smaller house hold is likely to have lesser household obligations and is more likely to commute longer. In summary, the models are found to o ffer very plausible interpretations and strong statistical good-of-fit measures. The test statistics associated w ith the coefficients are all statistically signifi cant and have expected signs. A cross-country comparison of the estimated models implies that the effects of socio-economic and household characteristics on commute time choices are very consistent between contexts. Interestingly, according to coefficients of the alternative specific constants appeared in all models reflect that commuters show a cons istent tendency to not commute for a longer period of time worldwide. Gender role in bearing household responsibilities and its influence on commuting preference is clearly vi sible in all contexts. For example, males and individuals in a small household are more likely to be long commuters while individuals in large household or females in a household w ith children are found to be short commuters. Age is also appeared to ha ve very similar nature of effect in all contexts. Well educated and high income house hold consistently show greater propensity of being long commuter in all contexts. In general, a person working in service sector or someone with a high job profile is more like ly to commute long while part-time workers and self-employed individuals are most likel y to be short commuters. Availability of autos and licensed drivers in the household are appeared to have positive impact on short commute in the US and Swiss contexts wh ile opposite effects ha ve been found by nonavailability of auto in the household. Howe ver, no significant effect has been found by auto ownership in the Indian c ontext, but availability of twowheelers is appeared to have negative effect on long commute. In the US and Swiss contexts, residential neighborhood has showed significant effect on commute time choice. Individuals living in rural areas are very likely to commute long while t hose living in dense urban cluster type

PAGE 187

172 neighborhood are found to be short commuters. After comparing all the three models, the only exceptions found are the effect of household size in the US model and low household income in the Swiss model. The di fference occurred in case of household size in the US context could be mainly because of the share of multi-worker household in the US sample is quite higher relative to Swi ss and India samples (2+ workers household in US: 43%; India: 33%; Swiss: Not available). Therefore, as househol d size increases the overall tendency of long commute increases in the US, but opposite effect is found in case of India and Switzerland. With regard to the second exception, it is seen that low household income is positively associated with long commuters as against to what found in the US and India models where individua ls with low household income tend to make short commute. Much of these differences can be explained by the fact that members in a low income household are most likely to be transit users in a European country like Switzerland, which simply makes them to take longer commute. 6.5 Summary and Discussion Over the past few decades, the study of commut er behavior has gained a lot of attention in the fields of travel beha vior and land-use modeling. Most of the research efforts in understanding commuter travel choices and pref erences have been made in the developed country context while virtually no knowledge is available on this subject in a developing country standpoint. The study is solely motiv ated by the importance of the topic and intended to identify the potential factors that influence commute time choice behavior in a developing country context. This research is primarily based on the hypothesis that commute time reflects an individuals prefer ence for residential and work place location. These preferences are the outcomes of indi vidual decision-making that are governed by their socio-economic, household ch aracteristics and personal traits. Three travel survey data sets: 2001 National Household Travel Survey, 2000 Switzerland Microcensus Travel Survey, and 2001 Thane, India Household Travel Survey are used in this study. Commuter sample extracted from all three data sets are further categorized into three market segments based on commute time: short ( 15 min), medium (16-59 min) and long ( 60 min). Detailed descriptiv e statistics of the commuter

PAGE 188

173 segments from all the data sets are described and multinomial logit models of commute time choice are estimated setting medium commut er type as the base alternative to better understand the behavioral aspects of the co mmuter in making short and long commuting length. Model estimated on the data sets around the world have facilitated better understanding about how contextual diffe rences influence commute time choice behavior. Model estimation results provide stro ng goodness-of-fit measures, meaningful and significant coefficients, and high-degree of sensitivity to socio-economic and demographic characteristics. With respect to international compar ison, it is found that commute time choice behavior is very simila r across the geographical locations used in this analysis. Commuters from both devel oped and developing countries consistently show negative propensity to commute longer. However, factors like working-status of the household members, socio-cultural differences (e.g. auto-oriented vs. transit oriented society) were identified as the key players in resulting some of the clear distinctions in commute length choice between the contexts.

PAGE 189

174 CHAPTER 7 ACTIVITY-TRAVEL INTERACTIONS AMONG ADULT HOUSEHOLD MEMBERS The previous has offered a detailed analysis of commute travel time choice behavior in an international context. MNL models are estimated for the commuter samples extracted from three data sets available from around th e world to better unde rstand their decision making mechanisms in choosing short, medium and long commute time. The current research presented in this chapter intends to explore interacti ons among adult household members with respect to activity and trav el engagement patterns from a developing country perspective. The current study will be compared with a similar study undertaken by Meka et al. (2002) which was based on a sample extracted from Southeast Florida Household Travel Survey. 7.1 Within-Household Interact ion in Activity Engagement Activity-based travel demand models are se eing rapid development around the world as they begin to replace traditional four-step trav el demand models in several major cities in developed countries. These models simulate daily activity-travel pa tterns of individual travelers at the disaggregate level of the deci sion-maker with a view to better capture the behavioral basis underlying human activity-trav el engagement. These models constitute major enhancements over traditional four-step m odels as they explicitly consider the role of time and space in determining activity-tra vel patterns. In addition, activity-based models are breaking new ground in the represen tation of agent-based interactions with explicit focus on intra-house hold interactions in activ ity-travel engagement. Activity-based model systems are becomi ng increasingly sophisticated in their ability to incorporate a variety of household interactions and constraints that influence individual activity-travel patterns. Over the past few years, signifi cant advances have

PAGE 190

175 been made in understanding the nature of household interactions and its role in explaining activity participation of and tr avel-activity decisions made by household members (Townsend, 1987; van Wissen, 1989; Golob, 1998; Lu and Pas, 1999). Interactions among household members may o ccur in several ways. Household members may allocate tasks among one another, make jo int decisions regarding activity scheduling and task allocation, undertake joint activities, and depend on one another for undertaking activities and travel (particularly in the cas e of children who depend on adults for their transport) (Golob and McNally, 1997; Browni ng and Chiappori, 1998; Fujii et al., 1999; Golob, 2000). As interactions among household members are undoubtedly important determinants of individual ac tivity-travel behavior, an unders tanding of such interactions and task allocation behavior is critical to the development of activity-based travel demand modeling systems (Becker, 1965; Chandras ekharan and Goulias, 1999; Simma and Axhausen, 2001). Even though most models in corporate household-le vel socio-economic variables as explanatory factor s, they may not be sufficient to explicitly account for the range of possible household interactions and task allocations that may take place. There have been several studies in the recent past aimed at exploring and modeling interactions among adult household members with respect to activity and travel engagement; however, virtually all of these studies are base d survey data sets collected in developed countries. This study intends to explore interac tions among adult household members with respect to activity and travel engagement patt erns from a developing country perspective. Structural equations models of activity e ngagement and task/time allocation among adult household members are developed and estimated in order to identify the trade-offs and complementary relationships among household members activity and travel patterns. Considering the relevance of household interactions in the development of activity-based travel demand models and the diffe rences that are likely to exist in such interactions between developi ng and developed countries, the time is ripe to explore and model household activity-travel interactions in a very different socio-cultural and demographic context. This study presents a comprehensive analysis and a series of

PAGE 191

176 structural equations models to identify the nature of interactions among adult members in the sample of households contained in the 2001 Thane survey data set. In order to focus the analysis, the modeling effort in this study is limited to the subset of households that contain two adults, one of whom is a worker. The structural equations model development effort examines the relationships between work and non-work activity engagement both withinand between-adult members in the household, while accounting for the effects of childre n, vehicle availability, and other socio-economic and demographic variables. The study presents a series of models and offers a rich comparison between the inter actions found in the India da ta set and those found in a similar study conducted a few years ago using a travel survey data set from the Southeast Florida region in the United St ates (Meka et al., 2002). 7.2 Data Sets and Sample Characteristics This study is based on 2001 Thane Household Trav el Survey Data set. To facilitate the particular interest of this research, a subset of 1275 households was extracted from the survey sample of 3505 households. Each household comprises of exactly two adults, at least one of whom is a worker. In order to make a meaningful compar ison between the adult household members, the following method was used to distinguish between the two adults. In a given household, the adult with the longer work duration is assigned person number 1, If two adults had identical work dur ations, then the older individual was assigned person number 1, If two adults had exactly the same work durations and age, then the person 1 was selected randomly between them. It is to be noted that the terms per son 1 and 2 and adult 1 and 2 have been used interchangeably in this study. Table 7.1 presents the descri ptive characteristics of th e two-adult households for the Thane and Southeast Florida sample. Th e descriptive statistics of the Southeast Florida is obtained from Meka et al. (2002) The average household size in both samples

PAGE 192

177 Table 7.1 Demographic Characterist ics of Two-Worker Households Characteristics Thane, IndiaSoutheast Florida No. of households 1275 1262 Average household size 2 adults (no child) 2 adults + 1 child 2 adults + 2 children 2 adults+ 3 or more children 3.28 32.7% 26.0% 26.0% 15.3% 3.4 32.5% 23.6% 26.8% 17.0% Average car ownership 0 car household 1 car household 2 car household 3+ car household 0.04 95.8% 3.9% 0.3% 0.0% 2.3 0.9% 10.1% 63.1% 24.9% Average two wheeler ownership 0 two wheeler household 1 two wheeler household 2 two wheeler household 0.16 85.2% 13.9% 0.9% NA NA NA NA Home ownership Own Rented Govt. Quarter Company home 75.7% 20.7% 3.3% 0.3% NA NA NA NA Home built-up area < 250 sq. ft 250-500 sq. ft 501-750 sq. ft 751-1000 sq. ft > 1000 sq. ft 34.3% 51.5% 12.1% 1.2% 0.9% NA NA NA NA NA Average No. of Workers 1 workers 2 workers 3+ workers 2.0 88.5% 11.5% 0.0% 2.0 21.0% 60.9% 18.1%

PAGE 193

178 Table 7.2 Person Characteristics of the Adult Household Members Thane, India Survey 2001 Southeast Florida Survey 1999 Characteristics Adult 1Adult 2Characteristics Adult 1 Adult 2 Sample size 12751275Sample Size 1262 1262 Gender Gender Male 92.2%9.6%Male NA NA Female 7.8%90.4%Female NA NA Age Age Average in years 3934 Average 42 43 18 to 30 years 20.2%41.8% 18 to 30 years 16.0% 17.9% 31 to 40 years 45.1%42.8% 31 to 40 years 28.4% 28.7% 41 to 50 years 26.8%9.6% 41 to 50 years 30.7% 25.0% 51 to 60 years 4.5%3.2% 51 to 60 years 15.8% 15.1% 61 years and over 3.4%2.5% 61+ years 6.6% 10.8% Missing 0%0% Missing 2.5% 2.5% Education Illiterate 6.0%12.6% Up to SSC 58.4%60.5%NA NA NA Up to HSC 11.6%9.0% College Graduate 23.9%17.8% Occupation Employment Service 64.6%9.4%Status Farmer/Laborer 7.4%1.2% Business/Professional 21.6%3.7%Full Time 90.3% 63.0% Student 0.0%1.1%Part Time 7.5% 12.8% Homemaker 0.9%80.6%Not Employed 1.3% 23.0% Retired/Unemployed 5.5%4.0% Personal Inc/Month Person Inc/Yr No income 3.5%85.1%$0-20 K 9.9% 23.8% Rs. 5,000 56.2%8.0%$21K-40K 26.1% 23.7% Rs. 5,001-15,000 38.2%6.5%$41K-60K 19.0% 11.7% Rs. 15001 and above 2.1%0.4%$61K-80K 5.8% 4.7% $81K+ 8.3% 3.4% Missing 30.9% 32.7% Vehicle ownership Work Mode No vehicle 75.1%96.4% SOV 87.6% 84.2% Car 6.1%0.7% Car/Van Pool 10.3% 12.5% Two wheeler 11.8%2.3% Transit 0.9% 1.5% Bicycle 7.0%0.6% Other 1.4% 1.8% Driver License 19.5%4.0%NA NA NA Transit pass 27.8%4.7%NA NA NA

PAGE 194

179 is quite comparable, which is about 3.5. A bout one-third of the households do not have children in both samples. With respect to household car ownership, the India two-adult households show extremely low auto availabil ity relative to the US sample (US: 2.3 vs. India: 0.04). However, about 14 percent of th e Indian households re ported at least one two-wheeler. Another very intere sting contrast betw een the Indian and US context is the distribution of the workers in the household. Consistent w ith the contemporary Indian society, vast majority of the Indian households have only one worker (88 percent) while 60 percent of the Northeast Florida sample is comprised of two workers. In the Indian context, it is very common to find househol ds with only one earning member, who is predominantly male. Traditional gender role is clearly visible here. The person characteristics of the Indi an adults presented in Table 7.2 are describes in this section. As e xpected, majority of the Indian adult 1 members are male as much as 92 percent while the other group is dominated by female (90 percent). With respect to age profile, the av erage age is marginal betw een the two groups. Younger age group share about 20 percent of the indivi duals classified as adult 1 while the corresponding percentage is 40 percent for the Adult 2 group. About two-third individuals in the adult 1 segment are in their middle age range. With respect to educational attainment, higher percent of adult 1 group member s reported themselves as college graduate compared to the other group. Vast majority of the adult 1 individuals reported their occupation as service or busin ess professional. On the other hand, adult 2 segment consists of as much as 80 percent of homemakers as opposed to the other group where the corresponding share of homemakers is negligible. Again, these findings reveal prominent gender role in the Indian househol ds. These facts are fu rther supported by the income distribution, where it is seen that 85 percent of the adult 2 individuals are nonearning members while only 3.5 percent reporte d no income in the adult 1 category. Consistent with expectation, gr eater percent of adult 1 members have drivers license and transit pass compared to the ot her adults in the household. Table 7.2 also provides corresponding person characteristics for the US adult samples (Meka et al., 2002). It is observed that average ages of the US adult samples are higher than the Indian adults. In addition, the age distributions in th e US context are very

PAGE 195

sim to em em contexts. Consistent with em reported higher incom 7.3 Activity-Travel Engagement P This section provides a descri patterns of the adult samples. Table 7.3 provi frequencies, travel and activit and non-work categories for both data sets. business activities are categorized as work considered as work-related businesses only if professions like service, bus business activity is considered as personal activity. The non-work activity duration is defi educational, personal business, 7.3 m adults who actually participated in a particul non-zero set and the corresponding sam the Thane sample, 141 persons (am activity. The average work activity duration for this is m work trip frequency reported findings are quite expected as non-w percentage com category in the US sa expectedly shows higher non-work trip freque should be noted that differences betw 180 ilar between the adult 1 and adult 2 sample s unlike the Indian samples. With respect ployment status, about 60 percent of i ndividuals in the adult 2 group are full-time ployed in the US sample, which is a remark able difference between the US and Indian ployment st atus, both adult groups in the US sample e and greater auto us age compared to the Indian adult groups. atterns of Adult Household Members ptive analysis of daily out-of-home activity engagement des the statistics of average daily trip y durations for adult 1 and 2 groups with respect to work In this study, all work and work-related activities. Notably, business activities are the persons have one of the legitimate iness/professional or farmi ng/laborer. Otherwise, any business, which is tr eated as a non-work ned in this study as the aggregation of social, recreational and any other activity durations. Table akes a distinction between the entire samples of adult 1 and 2 and the subset of the ar activity. The latter set is considered as the ple size is provided in parenthesis. For example in ong the 1275 classified as person 2) pursued work set of 141 persons is 415 minutes. With respect to the India sample, average work trip frequency of the adult 1 group uch higher than the corresponding aver age for the adult 2 gr oup while average nonby the adult 2 group is highe r than the other group. These orkers in adult 2 group are present with much higher pared to the adult group 1. Consistent with the India samples, adult 1 mple exhibits higher work trip frequency and adult 2 category ncy relative to the other adult group. But it een aver age trip frequencies between the two groups

PAGE 196

181 are much lower compared to the correspondi ng differences found in the India samples. As expected, both average work and non-work trip frequencies in the US samples (Meka et al., 2002) are quite higher as much as tw ice the averages of the India samples. With respect to activity durations in th e India sample, average work duration for those classified as adult 1 is about 8 hr 20 min. Though the average work duration for the entire adult 2 sample is c onsiderably low, the corres ponding average for the non-zero sample is found as high as 7 hours. Again as expected, group of a dult 2 on average spent more time on non-work activities compared to the other group, but the corresponding averages for the non-zero sample in each group is quite comparable. Consistent with the activity duration pa tterns in the India sample, adult 1 group is found to have much higher average work activity duration than the adu lt 2 group in the US sample (Meka et al., 2002) while the group of adult 2 is found to have higher non-work ac tivity duration than the group of adult 1. Comparison of activity durations between the two countries indicates that the adult 1 group in the India sample is spending exactly the same amount of time on work activity as their US counterpa rt while both adult groups in the US sample are spending much greater time in out-of-home non-work activities compared to the Indian adult samples. It is also found that th e average work activity duration of the adult 2 group in the US sample is quite higher than the corresponding average found for the Indian adult 2 sample. The differences in ac tivity durations between the two contexts can be attributed to the fact that the number of respondents who reporte d at least one out-ofhome activity on the survey day in the India sample is much lower than the US sample. Average travel durations show similar tre nds like activity durations. In the India sample, those classified as adult 1 spend about 35 minutes traveling to work while those classified as adult 2 spend, on average 4 mi nutes traveling to work. But the non-zero sample of the adult 1 group are spending same amount of time traveling to work as the adult 2 sample. The average non-work travel dur ation for the group of adult 2 is slightly higher than the corresponding average for the group of adult 1. This pattern can be found to reverse if one were to l ook at the non-zero samples in bo th groups. In the US sample, the group of adult 1 is traveling 40 minutes to work while the corresponding figure for the other group is 30 minutes while the non-work travel duration is higher for the

PAGE 197

182 individuals classified as adult 2 than the other adu lt group. Consistent with the differences in activity duration between th e contexts, the average work and non-work travel durations of the US adult samples are higher than their Indi an counterparts. The individuals classified as adult 1 in the India sample are spending only 4 minutes to nonwork activities while the same group in the US sample is spending about 27 minutes. Similarly, the corresponding aver ages for the adult 2 groups in the India and US samples are about 4 minutes and 50 minutes respectively. In summary, the activity-travel engagement patterns are consistent with expectations given the way in which the samp les were defined and person numbers were assigned. Comparison between a developing a nd a development country context reveal clear evidences of the influences of sociocultural differences in individuals activitytravel patterns. Nevertheless, these findings reflect a potential task allocation process where one person undertakes subsistence activ ities like work or work-related business activities and the other person participates in in-home or out-of-home household maintenance activities. These differences ar e very much reflective of activity interrelationships between household adults in the India sample. Th erefore, in the process of development of an activity based model sy stems in a developing context, one should attempt to better reflect such relationships. This study provides an exploratory analysis into the nature of these relati onships using a structur al equations modeling system. Table 7.3 Activity-Travel Patterns of the Adult Household Members Thane, India Survey 2001 1999 Southeast Florida Survey Purpose Adult 1 (All) Adult 1 (Non-zero) Adult 2 (All) Adult 2 (Non-zero) Adult 1 (All) Adult 2 (All) Daily Trip Work 1.03 1.03 (1275)0.121.05 (141)1.19 0.73 Non-work 0.07 1.03 (87)0.361.03 (227)1.15 2.06 Activity Duration (min) Work 498 498 (1275)46415 (141)498 207 Non-work 20 291 (87)41230 (227)44 105 Daily Travel Duration (min) Work 35 35 (1275)435 (141)43 30 Non-work 2 24 (87)421 (227)25 66

PAGE 198

183 7.4 Model Specification The model specification in this study is expl icitly framed to understand the interactions among two adult household heads in their activity-travel engagement patterns. The modeling of within-household in teractions in activity enga gement involves dealing with multiple endogenous variables in a simulta neous equations framework. Work and nonwork activity frequencies, activity durations, and travel durations are all activity and travel related endogenous variables that are inter-connected with one another. When modeling the interactions among several inter-dependent endogenous variables, simultaneous equations systems offer an a ppropriate framework for model development and hypothesis testing (Golob, 1998; Lu and Pa s, 1999; Fujii, et al., 1999; Golob, 2000; Golob and McNally, 1997; Simma and Axhaus en, 2001; Chandrasekharan and Goulias, 2001; Fujii and Kitamura; 2000). 7.4.1 Endogenous Variables Out-of-home non-work activity and travel dur ations of the two adult members in the household are considered as the endogenous variab les. It is considered that participation and the amount of time spent in non-work rela ted activities are potentially dependent on an individuals engagement in work-related ac tivities and travels and his/her personal and household attributes. An indivi dual may not have full contro l in scheduling and engaging in his/her work-related activity and travel, but based on his/her work-related and personal or household constraints he/she makes decision to participate in any non-work activity or travel. As mentioned earlier, non-work activity and travel durati ons used in this study are the aggregation of durations of various nonwork type activity and travel categories reported in the survey such as, school, s hopping, social/recreational and other. One could treat these activity types separately in the model specification, but in that case the number of free parameters to be estimated in the model would increase, and accurate estimation of the parameters would require larger sample size. Also, the incide nce of zero activity and travel durations will increase with the number of activity categories (Golob and

PAGE 199

184McNally, 1997). In the India data set, these problems are very pronounced and therefore, all non-work activity categories are meaningfully aggregated into one category. 7.4.2 Exogenous Variables In specifying the exogenous variables in the model structure, work activity and travel durations of the household adults and their personal and household characteristics such as income, household size, vehicle ownership, age, number of children etc are adopted as explanatory variables in the analysis. 7.4.3 Structural Equations Modeling Methodology A typical structural equations model (without latent variables) can be written as: XBYY (1) where Y is a column vector (m1) of endogenous variables, (In this study, m = 4); B is a matrix (mm) of parameters associated with right-hand-side endogenous variables; X is a column vector (n) of exogenous variables; (In this study, n = 12); is a matrix (mn)of parameters associated with exogenous variables, and is a column vector (mn) of error terms associated with the endogenous variables. Structural equations systems are estimated by covariance-based structural analysis, also called method of moments, in which the difference between the sample covariance and the model implied covariance matrices is minimized (Bollen, 1989). The fundamental hypothesis for the covariance-based estimation procedures is that the covariance matrix of the observed variables is a function of a set of parameters as shown in Equation 2: = () (2) where is the population covariance matrix of observed variables, is a vector that contains the model parameters, and () is the covariance matrix written as a function of

PAGE 200

185The relation of to () is basic to an understanding of identification, estimation, and assessments of model fit. The matrix () has three components, namely, the covariance matrix of Y, the covariance matrix of X with Y, and the covariance matrix of X. Let = covariance matrix of X and = covariance matrix of Then, it can be shown that (Bollen, 1989): (3) Before estimating model parameters, it is first necessary to ensure that the model is identified. Model identification in simultaneous structural equations systems is concerned with the ability to obtain unique estimates of the structural parameters. The identification problem is typically resolved by using theoretical knowledge of the phenomenon under investigation to place restrictions on model parameters. The restrictions usually employed are zero restrictions where selected endogenous variables and certain exogenous variables do not appear on the right hand side of certain equations and selected error correlations are specified to be zero. For identification of system (1), B must be chosen such that (I-B) non-singular, where I denotes the identity matrix of dimension m. The unknown parameters in B, , and are estimated so that the implied covariance matrix, is as close as possible to the sample covariance matrix, S. In order to achieve this, a fitting function F(S, ()), which is to be minimized, is defined. The fitting function has the properties of being a scalar, greater than or equal to zero, equal to zero if and only if () = S, and continuous in S and (). Available methods for parameter estimation include maximum likelihood (ML), unweighted least squares (ULS), generalized least squares (GLS), scale free least squares (SLS), and asymptotically distribution-free weighted least square (ADF-WLS). Each of these methods minimizes the fitting function and leads to consistent estimators of Ideally, one would use the ADF-WLS method of estimation to estimate parameters of structural equations models because of its ability to accommodate limited dependent variables with different asymptotic distributions. However, this method requires a 1111)BI()BI()BI)(()BI()(

PAGE 201

186greater sample size, due to its asymptotic properties (Golob and McNally, 1997; Meka et al., 2002). Golob and McNally (1997) suggested that in a situation where sample size one should use normal-theory based ML estimation procedure as it produces decent approximation of ADF-WLS method. Hence, in this study, the ML method of estimation may be employed without adversely influencing the estimation results. The fitting function that is minimized in the ML method of estimation of structural parameters is shown in Equation 4. FML = log | () | + tr (S -1 ()) log | S | (G + K) (4) where G = Number of excluded endogenous variables on RHS of the model, and K = Number of included exogenous variables on RHS of the model. The asymptotic covariance matrix for the ML estimator is given by, 1ML2FE1N2 (5) When is substituted for an estimated asymptotic covariance matrix that allows tests of statisal significance on parameters of is obtained. The combination PRELIS 2.0/LISREL 8.0 (Jreskog and Srbom, 1999) has been used in this study to estimate the model. It can be shown that the total effects of the endogenous variables on each other are given by: (6) The total effects of the exogenous variables on the endogenous variables in a structural equations model are given by: (7) which are the parameters of the reduced form equations (Golob and McNally, 1997). 7.4.4 Postulated Activity-Travel Causal Structure The postulated causal structure among the endogenous variables is shown in the Figure 7.1. The causal effects between the non-work activity-travel durations of the household adult members are hypothesized as a recursive system, which is explicitly based on the tic I)BI(1yyT 1x)BI(yT

PAGE 202

187 consideration that non-wor k activity engagement patte rns among adult 2 (secondary) members in the household are likely to be influenced by the non-work activity-travel patterns of the adult 1 (hous ehold head) members while the reverse effects are very unlikely to occur. Consistent with the socio-economic characteristics and activity-travel patterns previously seen of th ese groups, these expectations can be judged to be fairly realistic in the Indian context. Secondly, it is postulated that non-work travel is derived by the necessity to participat e in non-work activity, thus, a unidirectional causal effect between non-work activity and travel durations (non-work activity non-work travel) has been considered in the model structure. Under this framework, the coefficient matrix B will become a lower sub-diagonal matrix and the correlation among the error components ( ) can be set to zero which transforms to become a diagonal matrix. The postulated direct effects betw een endogenous-endogenous and exogenousendogenous in the model structure can be broadly classified into seven sections: 1) intraperson activity-travel interactions; 2) inter-person activity-travel interactions 3) intraperson activity interactions; 4) inter-person activity interactions; 5) intra person travel interactions; 6) inter person travel interac tions; 7) effects of demographic and socioeconomic characteristics on activity-trave l duration. 7.5 Model Estimation Results A comprehensive structural equation m odel was estimated on the sample of 1275 households to explore causal linkages betw een two adult persons in the household. Hypotheses regarding in ter-person interaction coupled with statisti cal measurers of fit and significance were use to guide the mode l development process. Finally, a model was accepted when it offered behaviorally sound in terpretations and satisfactory statistical indications. The model estimation pro cess was accomplished using PRELIS 2 (Jreskog and S rbom, 1999a ) and LISREL 8 (J reskog and S rbom, 1999b) software. The model structure and specification can be seen in Table 7.4 and Table 7.5. The causal relationships are presented in terms of direct, indirect a nd total effects. To understand these different effects, lets consider Table 7.4, where person 1 non-work activity duration affects person 2 non-work activ ity duration. However, person 1 non-work

PAGE 203

Figure 7.1 Postulated Activity-Travel Causal Structure 188 Person 1 Work Activity Duration Person 1 Work Travel Duration Person 2 Non-Work Travel Duration Person 2 Non-Work Activity Duration Person 1 Non-Work Activity Duration Person 1 Non-Work Travel Duration Demographic Characteristics Person 2 Work Travel Duration Person 2 Work Activity Duration 2 3 4 Exogenous Variables Endogenous Variables Error Components 1

PAGE 204

189 activity duration also affect s person 1 non-work travel dur ation. In turn, person 1 nonwork travel duration affects person 2 nonwork activity duration. Thus, person 1 nonwork activity duration also i ndirectly affects person 2 non-work activity duration through the mediating variable person 1 non-work travel duration. Th e indirect effect of person 1 non-work activity duration on person 2 non-work activity dura tion is the product of the two direct effects that cause the indirect effect. The total effect of person 1 non-work activity duration on pers on 2 non-work activity duration is the sum of the direct and indirect effects. All of the model coefficients presented in Table 7.4 and Table 7.5 are statistically significant at the 0.05 level of significance with a few excepti ons that are significant at the 0.1 level or below. The model 2 goodness-of-fit statistic indicates that the hypothesis that the matrix of sample moments is equal to the matrix of model implied moments can not be rejected at the 0.0 5 level of significance. The adjusted goodness-of-fit index (AGFI) is a measure of the relative moment of the sample variances and covariances that are predicted by the model, adjusted for the df of the model relative to the number of variables (Bollen, 1989; Golob and McNally, 1997). The AFGI for the model estimated in this study is 0.999. Thus the model fit is st atistically acceptable. The other measures of fit provided at the bottom of the Table 7.5 are al so in line with agreeable standards of fit for a structural equations model of this nature. It should be noted that some not statistically significant coefficients are retained in the models for model sensitivity and because the coefficients offered plausible be havioral interpretation. In the discussions that follow, it should be noted that person 1 represents th e adult in the household who spent the longest time working and older in age and may therefore be considered the primary worker or head of the household. The model estimation results provide very insightful and logically consistent findings. In the following section, the result s obtained from this present study on the India sample will be compared with the previous study (Meka et al., 2002) carried out on US sample whenever such comparisons apply.

PAGE 205

7.5.1 Intra-Person Activi Table 7.4 presents the interactions among th expected, fo on his/her non-work travel duration, wh engagem model provides a m am the estim work travel estim m travel for person 1 and 26 m travel patterns were found to be consistent showed that non-work activity duration of duration. The causal effects are consistent for indicate that non-work trav hour of non-work activity requires about 17 m and the corresponding figure for person 2 is 8 2002). respect to w negatively affects non-work tr work travel decreases with increase in wo suggests that every one hour in duration by 9 where work activity duration is duration. One hour increase in work activity work travel duration by 5 m household heads are less likely to engage in their long w 190 ty-Travel Interaction e endogenous activity-travel variables. As r both person 1 and person 2, nonwork activity duration has positive impact ich indicates longer non-work activity ent is likely to cause longer non-work travel duration. The structural equations echanism for quantifying the effects between the variables. The ount of travel generating from these activi ties can be determined by the magnitudes of ated coefficients. For instance, the total effects of non-work activity on nonated for person 1 and person 2 are 0.38 and 0.44 respectively, which eans that one hour of non-work activity dur ation requires about 23 minutes of non-work inutes of non-work travel for person 2. In the US context, the intr a-person interactions with respect to non-work activitywith the Indian situation. The US model person 1 directly affects non-work travel person 2 as well. The positive coefficients el duration increases with non-work activity duration. One inu tes of non-work travel for the person 1 minutes in the US context (Meka et al., The model structure in Table 7.5 shows the intra-person causal effects with ork and non-work activity-travel inte ractions. Person 1 work activity duration avel durations for the same person, which indicates nonrk activity duration. Estimated coefficient crease in work activity tends to reduce non-work travel minutes. However, reverse relationship is found with respect to person 2 positively affecting the same individuals non-work travel duration of person 2 increases his/her noninutes. These fi ndings can be explained by the fact that non-work travel after work possibly due to orking hour and at the same time if the second adult in the household is a

PAGE 206

191 worker, then the person shows greater te ndency to undertake household maintenancerelated non-work activities/travels mainly be cause of shorter working hours compared to his/her partner. This is a t ypical example signifying trade-off between the adult members in their household task allocation in the Indian household. However, in the US context (Meka et al., 2002), work activit y duration negatively affects non-work travel duration for both adul t samples. A 60 minutes increase in work activity duration for person 1 w ould bring about 4 minutes de crease in non-work travel duration for that person and the corresponding value for person 2 is 5 minutes. These findings convey a clear distinction in task allocation patterns between the adult members in a household in a developed a nd a developing country context. Consistent with the effects of work ac tivity duration, ones work-related travel duration negatively affects his/her non-work activity duration in both Indian and US contexts. However, work travel duration of person 2 in the India model does not show significant effect on non-work activ ity duration of that person. 7.5.2 Inter-Person Activi ty-Travel Interaction The India model indicates no significant tota l effect of non-work activity duration of person 1 on the non-work travel duration of pe rson 2. However, non-work travel duration of person 1 has significantly positive net eff ect on non-work activity duration of person 2. The model shows that a 30 minutes increase in non-work travel duration of person 1 contributes to a 2.4 minutes in crease in non-work activity du ration person 2. Similarly in the US context, positive interactions were found between these va riables and the model showed that an hour of non-work activity of person 1 would induce 5 minutes increase in non-work travel duration of person 2. These fi ndings indicate the complementary nature of discretionary activity participation by the adult members in the household. In addition, it can be seen from Table 7.5 th at the net total eff ect of work activity duration of person 1 on non-work activity durat ion of person 2 is negative. It appears from the model that every 60 minutes of wo rk activity duration of person 1 would bring about 1 minute of non-work travel durati on of person 2. This finding again suggests complementarities in activity -travel interaction because wh en person 1 works longer, the

PAGE 207

192 tendency of non-work activity participation jointly with other member significantly reduces. Conversely, work activity duration of person 2 has positive effect on non-work travel duration of person 1, which rather signifies task allocation b ecause when person 2 is heavily involved in subsis tence activities, the person 1 is more likely to undertake longer non-work travel to pick up some main tenance activities to run the household. A 60 minutes increase in the work-activity duration of person 2 contributes about 5 minutes increase in non-work travel duration of person 1. These findings are quite consistent with that found in the previous study on th e US sample (Meka et al., 2002). 7.5.3 Intra-Person Activity Interaction Table 7.5 presents intra-person causal eff ects between work and non-work activity durations for the adult members in the Indi an households. As expected, work activity duration of person 1 has a significant effect on non-work activity duration of that person indicating that as the person works longer, the amount of tim e he/she is spending on nonwork activity duration d ecreases. Corresponding model coeffi cient indicates that an hour of work activity reduces non-work activity duration by 11 minutes for a typical Indian household head. The US model exhibits sim ilar effects for both person 1 and person 2 adult segments where an one hour increase in work activity duration is associated with a decrease of about 9 minutes and 12 minutes in the duration of non-work activities respectively (Meka et al., 2002). 7.5.4 Inter-Person Activity Interaction The model also provides insightful causal relationships between work and non-work activity durations. Non-work activity duration of person 1 in the India model affects positively the non-work activity duration of pe rson 2 suggesting joint participation in non-work activities by the house hold adult members. But, work activity duration of person 1 is found to have negative effect on non-work activity of person 2, thus, capturing the complementary nature of non-wo rk activity engagement between the adult members while work activity duration of pe rson 2 affects positively the non-work activity duration of person 1. The model shows that a 60 minutes increase in work activity

PAGE 208

193 duration of person 2 entails 5 minutes incr ease in non-work activities for person 1. The trade-offs in task allocation between the a dult members are captured here because when the secondary adult member in the household is heavily involved in s ubsistence activities, the household head possibly tends to pi ck up some maintenance activities. The US model suggests that a 10 minut e increase in the non-work activity duration of person 1 contribute s to a 3.25 minute increase in non-work activity duration for person 2. Thus, it appears that pers on 1 and person 2 would jointly spend 3.25 minutes together while person 1 would sp end the other 6.75 minutes performing a nonwork activity outside home alone. Consistent with the Indian model, the net effect of person 1 work activity duration on person 2 non-work activity durat ion is found to be negative in the US model, but no signifi cant relationship was found between person 2 work activity duration and person 1 non-work activity duration. It appears in the US context that trade-offs in task allocati on among the adult members occur in a rare occasion because they possibly maintain an individual-specific designated set of daily activity agenda and therefore, any change in ones mandatory ac tivity duration does not seem to affect others non-work activity pattern (M eka et al., 2002). 7.5.5 Intra-Person Travel Interactions Estimated causal linkages among the intra-pe rson travel variables are found to be consistent with intra-person activity inter actions. Evident from Table 7.5, work travel duration of person 1 shows no direct effect on his/her non-work trav el duration, however, the model suggests a negative effect between these two variables suggesting longer workrelated activity-travel durati on of person 1 is likely to reduce the amount of time allocation for his/her non-work related travel The work travel duration of person 2 does not show any significant effect on the non-work travel of the same person. The US model suggested similar net effects among these travel variables with respect to the India model (Meka et al., 2002).

PAGE 209

194 7.5.6 Inter-Person Travel Interactions The model also provides very meaningfully c onsistent findings with respect to interperson travel interactions. It is found th at person 1 non-work travel duration has a positive direct effect on person 2 non-work travel duration. It appears from the model coefficient that a 30 minutes increase in th e non-work travel of person 1 leads to an increase in non-work travel duration by 4 minutes for person 2. Therefore, person 1 and person 2 would jointly spend 4 minutes of non-work travel while person 1 would spend 26 minutes of daily non-work tr avel alone. The model also suggests that work travel duration of person 1 is negatively affecting the non-work travel duration of the other. This effect holds true for person 2 as well. Ag ain, this is an example of complementarity in non-work travel interac tion suggesting that when one of the household adult is spending more time on work-related travel, th e chances of traveling jointly with other member in the household reduces. The model estimated on the US sample (Meka et al., 2002) shows remarkably similar patterns of in ter-person travel interaction as the Indian model. So far, it is seen in the India model th at as work activity duration of an adult member goes up, the propensity of the ot her adult to undertake non-work activity increases indicating the trade-off in task al location between the persons. The model also suggests that work travel duration of one of the members negatively affects the non-work travel duration of the other member. The previo usly described interaction may seem to be contradictory with the later one. But there ma y be a plausible interpretation to explain these effects such as, even though a person is not undertaking any nonwork travel jointly with the other member spending more time on work-related activity and travel, still he/she could solely undertake a short non-work travel by looking for destinations in close proximity to accomplish his/her desired non-work activity. 7.5.7 Effects of Demographic and Socio-Economic Attributes Socio-economic and demographic characteristic s of the adult members are found to have significant and meaningful e ffects on inter-person and intra-person non-work activitytravel engagement patterns in the Indian context. The estimated model coefficients

PAGE 210

195 presented in Table 7.5 suggest that adults living in a household with no children consistently show positive effect on non-work activity and travel duration for both adult categories. This finding is quite consistent with expectation because in a household with no children adults are less burdened w ith household obliga tion and childcare responsibilities. Therefore, they would have more freedom to spend greater amount of time in non-work activity and travel. Person 1 in his/her middle age group, high educated, employed with a high-profile job in the service sector is found to have negative effect on the non-work activity-travel of person 2. This is again an example of complimentary nature of non-work activitytravel engagement, where household head is a busy working person and has less time available to participate in nonwork activities with person 2 a nd then person 2 is also less likely to undertake such activities. A similar negative effect is found on nonwork activity and travel duration of person 1 when the person 2 is a worker. It is quite expected that being a worker, person 2 would have less time available to jointly part icipate in a non-work ac tivity or travel with the other adult member in the household and ther efore, person 1 would also be less likely to undertake any non-work activity or tr avel. Again, this e ffect captures the complementary nature of non-work activ ity and travel among household members. On the other hand, when person 2 is low e ducated, he or she influences positively the non-work activity duration of person 1. In the Indian context, low educated individuals who belong to the person 2 categor y are most likely to be female homemakers who dont usually participate in any out-of-home activity in a regular basis and prefer to stay at home to bear the major share of in-home household maintenance activities. Therefore, the household head who is most likely to be the male member (person 1) undertakes most of the non-work maintenance activities like shopping, child pick-up/drop off etc. in a daily basis. This is a typi cal example of in-home and out-of-home task allocation between male and female partners in an Indian two-adul t household. Again, the same variable affects the non-work travel dur ation of person 1 in a negative manner. The possible reason behind this is that if the pe rson 1 has less time available to undertake such non-work activities because of his/her work-related obligations then he or she could look

PAGE 211

196 for destinations in close proximity to get th eir non-work activities done to minimize their non-work travel duration. In the US context (Meka et al., 2002 ), socio-economic and demographic characteristics are found to have signif icant effects on non-work activity-travel engagement patterns of adult members in the household as well. The model shows availability of two vehicles in the household negatively affects non-work activity and travel duration of the adult members in th e household. This is possibly because greater availability of vehicles is lik ely to eliminate the need of joint participation in maintenance activities and travels for the hous ehold members; rather it allo ws them to distribute their responsibilities among themselves in pursu ing certain designate d activities. Such arrangements tend to reduce non-work activity an d travel durations potentially for each of the household members. Presence of children in the household positively affects the nonwork activity and travel dur ation of person 1 while it a ffects negatively the non-work activity and travel duration for person 2. Th ese effects clearly captures designated task allocation between the members suggesting th at person 1 tends to be responsible for undertaking out-of-home childcare responsibilities such as child drop-off and pick-up while person 2 is more likely to pick up major in-home child care responsibilities. In general, the model provides a compre hensive understanding of the intra-person and inter-person interactions in the context of their non-work activity and travel engagement patterns. Within-p erson interactions clearly show trade-offs between work and non-work activities. As work engage ment increases, that persons non-work engagement decreases. However, between pers ons, the potential complementary and joint nature of non-work activity engagement am ong household members is an interesting and important finding. In general, as non-work activity engagement increases for one person, it is also found to increase for the othe r person suggesting jointness in activity engagement. So, when one person works l onger and has less time for non-work activity engagement, the other person also has reduced non-work activity engagement. These relationships are important ingredients of the overall household activity-travel dynamics that need to be reflected in comprehe nsive activity-based tr avel demand modeling systems.

PAGE 212

197 Table 7.4 Structural Equation Model Esti mation Results (Causal Effects between Endogenous Variables) Endogenous variable Effect Person 1 Non-work activity duration Person 1 Non-work travel duration Person 2 Non-work activity duration Person 2 Non-work travel Duration Person 1 Non-work activity duration Direct Indirect Total ----Person 1 Non-work travel duration Direct Indirect Total 0.38 0.00 0.38 ---Person 2 Non-work activity duration Direct Indirect Total 0.03 0.02* 0.05** 0.04 0.00 0.04* --Person 2 Non-work travel Duration Direct Indirect Total -0.06 0.06 0.00 0.11 0.02* 0.13 0.44 0.00 0.44 -*Significant below 90 percent level **Significant at 90 percent level All other variable significant at 95 percent level

PAGE 213

198Table 7.5 Structural Equation Model Estimation Results (Causal Effects of Exogenous Variables on Endogenous Variables) Endogenous Variable Effect Person1 Work Activity Duration Person1 Work Travel Duration Person2 Work Activity Duration Person2 Work Travel Duration HH Size=2 (no children) Person1 Age: 31 45 Person1 Highly Educated Person1 Occu: Service Person1 Income Rs.5K+ Person2 Male Person2 Worker Person2 Low educated Person1 Non-Work Activity Duration Direct Indirect Total -0.18 0.00 -0.18 -0.07 0.00 -0.07 0.09 0.00 0.09 0.00 0.00 0.00 0.13 0.00 0.13 -0.16 0.00 -0.16 0.25 0.00 0.25 -0.08 0.00 -0.08 -0.13 0.00 -0.13 0.00 0.00 0.00 -0.07** 0.00 -0.07 0.23 0.00 0.23 Person1 Non-Work Travel Duration Direct Indirect Total -0.08 -0.07 -0.15 0.00 -0.03 -0.03 0.04* 0.03 0.07 -0.04* 0.00 -0.04 0.16 0.05 0.21 -0.18 -0.06 -0.24 -0.09** 0.10 0.01 -0.10 -0.03 -0.13 -0.10 -0.05 -0.15 0.00 0.00 0.00 0.00 -0.03* -0.03 -0.11 0.09 -0.02 Person2 Non-Work Activity Duration Direct Indirect Total 0.00 -0.01 -0.01 0.04* 0.00* 0.04 0.17 0.01** 0.18 0.00 0.00* 0.00 0.00 0.01 0.01 0.00 -0.02 -0.02 -0.24 0.01 -0.23 -0.06 -0.01 -0.07 0.00 -0.01 -0.01 0.32 0.00 0.32 -0.58 0.00* -0.58 -0.43 0.01* -0.42 Person2 Non-Work Travel Duration Direct Indirect Total 0.00 -0.01** -0.01 -0.05 0.02* -0.03 0.00 0.08 0.08 0.00* 0.00* 0.00 0.06 0.02 0.08 0.00 -0.02 -0.02 0.10 -0.12 -0.02 0.00 -0.04 -0.04 -0.05** -0.01 -0.06 0.18 0.00 0.18 -0.20 -0.26 -0.46 0.00 -0.20 -0.20 N=1275; 2 =4.95 with 15 df ; p-val = 0.99; GFI=0.99; AGFI=0.999; RMSEA=0.001 *Significant below 90 percent level **Significant at 90 percent level All other variable significant at 95 percent level 198

PAGE 214

199 7.6 Summary and Discussion This study intends to explor e interactions among adult hous ehold members with respect to their activity and travel engagement patt erns from a developing country perspective. Structural equations models of activity e ngagement and task/time allocation among adult household members are developed and estimated in order to identify the trade-offs and complementary relationships among household me mbers activity and travel patterns. A sample of 1275 households drawn from the 2001 Household Travel Survey in the Thane Metropolitan Area, India has faci litated to achieve the goal of this research. In order to focus the analysis, the modeling effort in this study is limited to the subset of households that contain two adults, one of whom is a worker. The person with the longer work activity duration in the hous ehold was designated as pers on 1 and the other person was defined as person 2. Under such framework, pe rson 1 may be considered as the head of the household. In general, the estimated model has pr ovided a comprehensive understanding of the intra-person and inter-person interactions in the context of their non-work activity and travel engagement patterns. Within-person in teractions clearly show trade-offs between work and non-work activities. As work engagement increases, that persons non-work engagement decreases. However, between persons, the model exhibits many important findings about complementary nature of non-wo rk activity engagement and trade-offs in task allocation between the adult members in the household. In general, as non-work activity engagement increases for one person, it is also found to increase for the other person suggesting jointness in activity engagement. Especially, when the household head works longer and has less time for non-work activity engagement, the other person also has reduced non-work activity engagement. But, when the secondary adult member in the household is working longer, the household head shows greater tendency to pick up some household maintenance work. These relationships are important ingred ients of the overall household activity-travel dynamics that need to be reflected in co mprehensive activitybased travel demand modeling systems. Considering the relevance of household interactions in the development of activity-based travel demand models and the diffe rences that are likely to exist in such

PAGE 215

200 interactions between developi ng and developed countries, the study presents a series of models and offers a rich comparison between the interactions found in the India data set and those found in a similar study conducted a fe w years ago using a travel survey data set from the Southeast Florida region in th e United States. Comparison between the two studies reveals remarkable consistencies in non-work activity-travel engagement patterns and task allocation behavior of the house hold members in a developing and a developed country context. The structural equati ons modeling framework has also allowed accounting for the effects of children, vehicl e availability, and other socio-economic and demographic variables. The socio-economic ch aracteristics of the household adults are found to have very insightful effects on in tra-person and inter-pe rson non-work activitytravel patterns in a developing country context. In this particular study, non-work activities had to be grouped together due to the rather low participation rates in maintena nce and leisure activities (when treated separately). This is consider ed as a limitation of this study as such arrangement may cause unwanted confusion in interpreting th e model coefficients. Therefore, it is important to preserve the disaggregate act ivity purpose classification in terms of maintenance and leisure activity categories in the model structure whenever possible. One should note that the nature of household interaction in pursuing these activities may be very different from each other because maintenance activities may be allocated between adults (suggesting a trade-off), le isure activities may be conducted jointly (suggesting a complementary effect). The primary objective of the study was to offer insights into the nature of household interactions in non-work activity engagement patterns and offer a methodology that can effectively capture and qu antify these interacti ons in a developing country context. The model results have provide d very insightful findings consistent with a developing country perspective. It is envi sioned that such models can make valuable contribution to the development of advanced activity-based mode l system in India.

PAGE 216

201 CHAPTER 8 SUMMARY AND CONCLUSIONS This chapter offers a summary and draws conclusions of the whole research effort presented in the previous chapters. Flourishing economy, rapid industrialization and increasing trend of motorization have been shaping societies in the developi ng countries like India in an unprecedented rate. Infrastructure backlog amid such rapid growth in all imaginable directions have heavily exacerbated the urban transport cr isis in the developing world by alarming increase in vehicular travel demand, road fatalities, and environmental pollution. To encounter urban transportation challe nges, the necessary development and implementation of effective transport planning and policies have been generally lagged in the developing countries that s een in the developed countries due to several constraints including resource constraints, knowledge cons traints, institutional constraints and so on. In most developed countries, the focus of transportation planning ha s shifted away from capacity expansion to that of operation, management, and optimization of existing capacity. Constraints on the availability of fi nancial resources to maintain and expand the existing infrastructure have led to the consideration of alternative mobility management options such as travel demand management and smart growth practices. This shift in planning emphasis has motivated travel beha vior researchers to be concerned with relationships and trade-offs among individuals time expenditures, travel, and activities. It is envisioned that travel behavior models based on an understanding of peoples time use patterns offer a robust framework fo r analyzing the impacts of alternative transportation policies and control measures. Ho wever, in the recent past, with the rapid development seen by several emerging ec onomies and the explosive growth in transportation infrastructure investment, ther e is a growing interest in the development and implementation of advanced travel demand modeling systems in developing

PAGE 217

202 countries. But lack of necessary research and exploration of travel behavior in a developing country context has left very limited knowledge for us to understand the extent of applicability of th ese advanced theories and met hodologies in a different sociocultural perspective. Unde rstanding the growing need in the development and implementation of advanced travel de mand modeling systems in the developing countries, this research a dopted a comprehensive approach to explore the activity engagement and time use behavior in a deve loping country context. Considering the contextual differences that are likely to exist between developing and developed countries, this research also offered rich comparisons of activity patterns and time use behavior across diverse socio-cu ltural and geographical contexts. A discussion on the state of the art pract ices and methodological developments in travel behavior and time use research ar ea was included in this study. A rich body of literature supports the ro le of activity-based approaches that have immensely contributed in the improvements over traditional trip -based methods by explicitly recognizing the role of time dimension on individual activity -travel patterns. The studies of activity engagement and time use patterns have enab led to enrich our u nderstanding of the complexity and variability of individual trav el behavior and that has led to increased capability of forecasting travel demand and evaluating planning options. However, the development and implementation of activity-bas ed methods have generally lagged in the developing countries that seen in the developed world over the past decades. Growing complexities in peoples travel behavior and activity engagement patterns in accordance with these social transformati ons have become extremely important in the context of the development and implementation of advanced modeling approaches for passenger travel demand forecasting in developi ng countries like India. This study offered an extensive empirical analysis of travel characteristics, activity engagement pattern and time use beha vior in a developing country context. To explore the differences in activity and tim e use patterns between a developing and a developed country context, the study adopted 2001 Thane Household Travel Survey and 2001 NHTS, Florida Sample for the purpose of comparison. With respect to demographic and socio-economic profiles of the datasets, th e India sample is typically characterized by

PAGE 218

203 larger household size, more children in the household, lower household/personal income and remarkably low level of vehicle owne rship relative to the US sample. Person characteristics of the survey samples were investigated based on commuting status. Majority of the Indian commuters are ma le and vast majority of the Indian noncommuters are female, which is consistent with expectation because labor force participation among females are still very low in India compared to the developed countries. Cross-classification tables of trip production rates were generated based on household and person characteri stics such as household si ze, vehicle ownership and income, and trip purpose. Consistent with expectation, household trip rates tend to increase with household size and vehicle ownership. Overal l trip production rates are found much higher for the US households co mpared to the Indian households. Trip distribution analysis suggeste d that more than 80 percen t trips reported by the Indian respondents consists of home-based work and school trips while a very small percent of trips reported are related to non-work purposes as opposed to the US trip distribution pattern, where 75 percent of tota l reported trips are non-work in nature. Next, time of day analysis shed light on many distinctive characteristics of departure time choice patterns between the two countries. Detailed modal split analysis was conducted based on trip purpose, commuting status and household car ownership. With respect to Indian sample groups, large proportion (90 percent) of the trips is accomplished by either public transportation or non-motorized modes, while sh are of auto trips ar e prevalent among the US market segments (above 80 percent). Analys is of activity and time use characteristics of various market segments provided valuab le insights about how people trade off their time into various mandatory and non-mandatory activities. The aver age trip frequency reported by the US sample groups are much higher than those reported by the India sample groups. Analysis suggested that the I ndian travelers undert ake, on average one out-of-home activity per day while much hi gher trip frequency reported by the US sample groups apparently indicates greater le vel of out-of-home activity participation in the US context. Consistent with trip rate pa tterns, average daily travel durations for the US market segments are much higher than the Indian market segments. Commuters are found to spend more time on travel than non-co mmuters in both contexts. With respect to

PAGE 219

204 activity duration, averag e time spent on non-work activitie s by Indian commuter sample is remarkably lower than the US commuter sample. However, durations of shopping and recreational activities for the Indian noncommuters are significantly higher than the Indian commuters and also quite comparab le with the corresponding averages found in the US non-commuter sample. Trip length di stribution by purpose were analyzed and compared between the two survey samples. Aver age trip lengths of the India sample were found to be consistently higher than the US sample for all trip purposes. Finally, trip chaining analysis was performed to unde rstand the non-work activity engagement patterns of commuters. Comparison between commuter samples obtained from India, Japan and US suggests that the US commuters have much complex trip chaining patterns compared to the commuters in India and Japa n. Much of these differe nces are attributed mainly to the variability in mode choice behavior across the count ries. The Indian and Japanese commuters are likely to face grea ter constants in part icipating non-work activities because they are heavily reliant on public tr ansportation while the US commuters are primarily dependent on priv ate vehicle which provides them greater flexibility to access their activity locations. This research introduced th e concept of travel time frontier to better understand the variability in travel time expenditure around the world (i.e. US, Switzerland and India). The stochastic frontier modeli ng methodology is employed to identify the unobserved travel time frontier (TTF) that is considered to be re presentative of the maximum amount of time that an individual is willing to allocate to travel in a day. The results presented in this study shed considerable light on the va riability of the TTF between commuter and non-commuter samples ac ross international c ontexts. The average expected TTFs were found to be about 3 hours for US and Swiss commuters and about 2.5 hours for Indian commuters. Although the rang e of these average expected values is rather narrow, it is clear from the distribution that there is considerable inter-person variation in the expected travel time frontiers. For non-commuter samples, the distributions are even more spread-out, and the aggregate sample-w ide averages of the expected TTFs range from about 1.5 hours to 4 hours. The findings reported in this study have important transport policy implica tions. Around the world, transport policies,

PAGE 220

205 infrastructure investments, teleco mmunications technology, 24-hour business establishments, modal flexibility and avai lability, virtual workplaces, and smaller household sizes are resulting in the loosening of constraints an d the increased availability of discretionary resources. The notion of the TTF provides a powerful framework for analyzing increases in travel time expe nditures that might result from continued loosening of constraints over time; presumabl y, travel time expenditu res can continue to rise as long as they are lower than the TTF but would stop increasing when the TTF is reached. Thus, this conceptualization provi des a means for analyzing induced travel effects from an activity-based time use allocation perspective. An another study was conducted on travel time expenditure to determine the subjective or perceived minimum amount of tr avel time that people feel they must undertake to accomplish the minimum required activities of the day. An understanding of the minimum travel time threshold would offe r the potential to quantitatively assess the maximum amount of travel reduction that may potentially be accomplished through TDM and TCM policy implementation. The stochast ic frontier modeling methodology has been employed to determine the minimum required tr avel time frontier that influences the actual travel time expenditure observed in travel surveys. The minimum travel time threshold is represented by an unobserved cost frontier in this modeling framework. The study is conducted in an inte rnational context by estimati ng cost frontier models of minimum required travel time frontier on surv ey samples drawn from US, Switzerland and India. Separate models were estimated for commuters and non-commuters to recognize the inherent differences between thes e market segments. The stochastic frontier models were found to offer statistically significant coefficients for several socioeconomic and demographic characteristics i ndicating that the mini mum required travel time frontier is likely to be influenced by a persons lifecycle stage, lifestyle, income, age, and household characteristics. The model estimation results were used to plot distributions of the expected necessary trav el time frontier values vis--vis the actual travel time expenditures. The plots were found to offer plausibl e interpretations and suggested that the stochas tic frontier modeling methodol ogy is suitable for modeling minimum required travel time frontiers. The average expected mini mum required travel

PAGE 221

206 time frontier values were found to be about 20 minutes for US and Swiss commuters, 60 minutes for Indian commuters, and 3-7 minutes for non-commutes. The expected minimum travel time frontier is greater than zero minutes for a vast majority of the individuals in all survey samples. This findi ng suggests that individua ls, in general, feel that they have to undertake at least a certain amount of minimum travel to fulfill the minimum required activity agenda. It is also found that the average difference between the actual travel time expenditure and the e xpected minimum required travel time frontier is much higher in the US and Swiss context re lative to the Indian context. This clearly reflects the effects of the maturity and performance of the transportation system on activity and travel engagement. In developed coun tries, traveling offers a disutility that is small enough to motivate substantial activity engagement (and therefore, travel) above and beyond the perceived minimum required tr avel. However, in developing countries, the disutility of traveling is still so large that additional activity engagement (and travel) is undertaken more sparingly. Time allocation behavior of commuters in their journey between home and work is considered in the study as an important ma nifestation of trade-offs between residential location choice and work-related spatio-temporal constraints. This research offered a detailed analysis of commuting length choice behavior around the world based on an individuals choice of short, medium and long commute time. Multinomial logit models of commute time choice are estimated on three data sets available from US. Switzerland and India setting medium commuter type as th e base alternative to better understand the behavioral aspects of the commuter in making short and long commuting length. Model estimation results provide strong goodness-of-f it measures, meaningful and significant coefficients, and high-degree of sensitiv ity to socio-economic and demographic characteristics. With respect to internati onal comparison, it is found that commute time choice behavior is very similar across the ge ographical locations used in this analysis. Commuters from both develope d and developing countries consistently show negative propensity to commute longer. Personal attri butes, such as high educational attainment, well-paid job and being male are more likely to influence positively on longer commute time choice. On the other hand, household cons traints like presence of children or low

PAGE 222

207 household income affect commuters positiv ely to choose shorter commute length. Interestingly, vehicle availability or presence of multiple drivers in the household attributes to increased chances of short comm uting as availability of multiple vehicles in the household reduce the chances of carpool wh ich contributes to shorter commute time for each individual member. However, factor s like working-status of the household members, socio-cultural differences (e.g. auto-o riented vs. transit oriented society) were identified as the key players in resulting some of the clear distinctions in commute length choice between the contexts. Finally, a study was conducted to explor e interactions am ong adult household members with respect to their activity and tr avel engagement patterns from a developing country perspective. Structur al equations models of activ ity engagement and task/time allocation among adult household members are developed and estimated in order to identify the trade-offs and complementar y relationships among household members activity and travel patterns. The estim ated model provided a comprehensive understanding of the intra-pers on and inter-person interactions in the context of their nonwork activity and travel enga gement patterns. Within-person interactions clearly show trade-offs between work and non-work activitie s. As work engagement increases, that persons non-work engagement decreases. Howe ver, between persons, the model exhibits many important findings about complementary nature of non-work activity engagement and trade-offs in task allocation between th e adult members in the household. In general, as non-work activity engagement increases for one person, it is also found to increase for the other person suggesting jointness in act ivity engagement. These relationships are important ingredients of the overall household activity-travel dynamics that need to be reflected in comprehensive activity-based travel demand modeling systems. The structural equations modeling framework has also allowed accounting for the effects of children, vehicle availabilit y, and other socio-economic and demographic variables. Comparison between similar studies conducted on India and US samples reveals remarkable consistencies in non-work activ ity-travel engagement patterns and task allocation behavior of the household members in a developing and a developed country context.

PAGE 223

208 This research facilitates to identify many c onstraints that prevail in peoples life in the Indian society resulting very different act ivity-travel and time use patterns in contrast with the societies in the developed world. Some of these constraints are highlighted below: Monetary constraints : Low vehicle ownership and limited participation in outof-home leisure activities can be directly linked to low income level of average Indian households. Monetary constraints explain much of the prominent contrasts in quality of life and standard of living between a developed and a developing country. Modal constraints : In India, majority of people do not have personal transport and have to rely on non-motorized or public transport modes. These slow modes lack flexibility and reliability th at generally deter trip chaining, make it difficult for people to engage in non-activities. Institutional/infrastructural constraints : Indian business establishments and service organization do not yet operate on a 24-hour basis such as many establishments in the western context. Most business establishments close at dark, leaving little opportuni ty to engage in other activities after work and leave people mostly home-oriented. The concept of overtime and performance driven work are only beginning to enter the Indian work culture. In addition, once work is completed, there ar e not many recreational and other opportunities for people. Not only are th e opportunities few and far between, but they tend to be expensive for the average Indian worker/household. Finally, and most notably, the transporta tion system in India does not offer a level of service high eno ugh to encourage engagement in evening non-work out-of-home activities. Household constraints: Household constraints ar e found to have prominent roles in defining activity-travel enga gement patterns and task allocation among household members. Household c onstraints define distinct gender roles in the Indian society where female s are most likely to stay at home to bear major share of household and ch ildcare responsibil ities and males

PAGE 224

209 commonly undertake work-re lated subsistence acti vities to support nonworking household members. Social/cultural constraints: Traditional values and social/cultural constraints are still visible in the Indian society and influence almost every aspect of peoples lives including their travel characteristics. For instance, travel behavior of traditional elderly peop le is quite different than younger generation who tend to adopt many ways of western cultu re. Traditionally, labor force participation among female is much lower than male who is commonly the primary earning member of the household. Out-of-home activity participation after dark among fe males is still very limited in India due to major security concerns. In the end, it can be concluded that this research shed considerable light on the unexplored area in the literature of activity e ngagement pattern and time use behavior in a developing country context. The model resu lts provided very insightful findings and plausible interpretations consistent with a developing country pe rspective recognizing a wide spectrum of differences and similarities in activity patterns and time use behavior between developed and developing countri es. Specified model structures were meaningfully able to incorporate social, cultu ral, institutional and transportation system constraints and reflected sensitivity to the behavioral variability between the contexts suggesting that advanced analyt ical techniques may be satis factorily applied on the data set from developing countries which may contribute important ingredients in the development of advanced activity-based model system in the countries like India. It is envisioned that some potential policy implications can be derived from this research effort in the context of trans portation planning and demand analysis, policy making and quality of life in the developing country context. Perf ormance and level of service of the transportation system is rated as an important determinant of quality of life. In this context, the time-use and activity patterns of people can be postulated as the manifestation of their desire to pursue activit ies that are distributed over time and space. Thus by analyzing the time use associated with an activity-travel patte rn, one may be able

PAGE 225

210 to measure the level of satisfaction that a person is deriving from his activity-travel pattern. In the context of the current study, it can be hypothesized that the quality of life of the Indians are much poorer compared to the Americans due to limited accessibility to opportunities resulting from constrained mobility and lack of freedom provided by poor transportation infrastructure, unavailability of private vehicle and dependence on public transportation services. By looking at the time-use patterns, it has become clear that the Indian commuters are spending greater am ount of time on work-re lated activity and travel while spending much lesser amount of time to discretionary leisure activities like social, recreational relative to the US commu ters, thus suggesting th at work-life balance in India is much worse than the people in the US. It was appeared in a study conducted by Nehra et al. (2005) that peopl e in India arrive at home fr om work much earlier than the workers in the US even though their perceived latest hom e arrival time is similar to the developed countries. The study indicated that the Indian workers tend to spend majority of their free time on in-home activit ies as there are not many recreational and other activity opportunities after work. However, situation in India has started changing with growing economy and explosive inve stment in revamping transportation infrastructure. One may conjecture that as th e transportation system improves, disposable income and vehicle ownership grows, and so cial norm loosen, peopl e will increasingly take advantage of growing opportunities in orde r to improve their quality of lives. It is quite obvious that people would prefer to spend their free time on out-of-home activities rather staying at home. This will induce new trips resulting increased travel demand in the near future. Incorporation of the effects of such dynamic changes in consumer choice and preferences are critical in the developm ent of future transp ortation planning and policy options. In the process of moving towards a next generation transportation planning process, time use studies coul d be a key element to assess how urban development and transportation systems contribut e to the quality of lives of people in the developing world.

PAGE 226

211 CHAPTER 9 FUTURE RECOMMENDATIONS AN D RESEARCH GUIDELINES This chapter addresses some future recomm endations and shed light on some potential future research problems for further progr ession of the current research effort. 9.1 Data Needs Much of the effectiveness of transportation planning and policy development is dependent on the richness of the travel survey data sets. In the de veloping countries like India, the main hindrance in the transportati on planning process is unavailability of rich survey data sets. India and other developing co untries lacks strategic efforts in collecting quality travel survey data necessary for tran sportation planning process. Emphasis should be placed on future data collection in the developing countries to identify the recent and emerging issues in travel behavior resear ch. The following data collection efforts are considered crucial to meet the future need s for the development of new generation of forecasting model systems and advancement of tr avel behavior analysis in the developing countries. Conducting national level travel surveys similar to National Household Travel Surveys in the US; Collection of detailed in-home and out -of-home activity and time use data; Incorporation of questions regarding usage of information technology in the survey instrument; Conducting periodic longer-term cross-sec tional surveys to capture dynamics in population characteristics and long-term ch anges in peoples travel behavior and life style;

PAGE 227

212 Conducting multi-day time use surveys in th e context of activity-travel patterns to capture shorter-term dynamics in travel be havior such as day-to-day (or between weekday and weekend) varia tions in mode choice, act ivity engagement, vehicle occupancy, departure time choice and so on; Incorporation of attitudinal questions the survey instrument for the purpose of calibration and validation of travel demand forecasting models. 9.2 Exploration of Temporal Dynamics in Travel Time Frontier in an International Context The current research explored variation in travel time expenditure over space and individual by modeling and es timating travel time frontier for samples drawn from three different geographical contexts. The study sugge sted significant inter-person and spatial variation in travel time frontie r along with variations in obser ved travel time expenditure. However, the study did not explore the variati on in travel time frontie rs over time. There is a growing interest in the profession in investigating temporal dynamics in travel time expenditure as well. There have several studies in the past that have found increases in average travel time expenditures over time when analyzing repeated cross-sectional data sets. The future extension of this current research effort will be to investigate temporal dynamics in travel time frontier employing l ongitudinal and panel data sets. Again, it is speculated that stochastic frontier modeling methodology will provide suitable framework to accomplish this task while controlling for unobserved individual specific effects. 9.3 Modeling Positive Utility of Travel fr om the Utility Maximization Framework Travel behavior researchers and travel demand forecasting models have generally assumed that travelers attempt to minimize trav el cost when choosing to undertake a trip. For example, travel choice models, formulated based on the utility maximization principle, are based on the premise that trav el time negatively impacts the utility of a mode or destination. However, with the in creasing interest in activity-based travel demand analysis and time use behavior, there is a new line of inquiry that is suggesting that all travel may not be viewed as a disuti lity by a traveler. Ba sed on this new line of

PAGE 228

213 inquiry, it appears that a certain amount of travel may actually be desirable and offer positive utility to the individual undertaking the travel. Even though there are several possible explanations for this evidence of th e positive utility of tr avel, the subject still remains illusive as it is not clear if the uti lity of the travel experience is truly being measured separate from the utility associated with or derived from the activity at the destination. If people are trav eling to a desirable recreational activity, it is likely that the travel experience to that destination/activity is also going to be part of the desirable activity-travel package. Thus, it is possible that the positive utility of the travel is simply reflecting the utility that the individual is go ing to derive from the destination activity. On the other hand, there are also other potentia l reasons for this phenomenon which suggests that people may desire to trav el just for the sake of trav eling. There is a rich body of literature dealing with this debate but still ther e is no success to quantity the amount of travel that actually offers positive utility to a person and beyond which the person begins to feel the burden of travel or apparently travel incurs negative utility to the person. Solving this problem from a utility maxi mization framework will be a worth while contribution to the profession.

PAGE 229

214 REFERENCES Adler, T. J., and M. Ben-Akiva (1979). A Theoretical and Empirical Model of Trip Chaining Behavior. Transportation Research Part B 13(3), 243-257. Aigner, D., C.A.K, Lovell, and P. Schmidt (1977). Formulation and Estimation of Stochastic Frontier Production Function Models. Journal of Econometrics, 6, 21 37. Alonso, W. (1964). Location and Landuse. Harvard University Press, Cambridge, Massachusetts. Aptech Systems, Inc. (1999). Gauss Reference Guide 3.2. Aptech Systems, Inc., Maple Valley, WA. Becker, G.S. (1965). A Theory of the Allocation of Time. The Economic Journal, 75(299), 493-517. Ben-Akiva, M., and S. R. Lerman (1985). Discrete Choice Analysis: Theory and Application to Travel Demand MIT Press, Cambridge, Massachusetts. Bhat, C. R. (1996). A Hazard-Based Du ration Model of Shopping Activity with Nonparametric Baseline Specification and Nonparametric Control for Unobserved Heterogeneity. Transportation Research Part B 30(3), 189-207. Bhat, C. R. (2000). Duration Modeling. In Hensher, D. A. and Button, K. J. (eds.), Handbook of Transport Modeling Elsevier Science, 91-111. Bhat, C.R. (2001). Modeling the Commute Activity-Travel Pattern of Workers: Formulation and Empirical Analysis. Transportation Science, 35(1), 61-79. Bhat, C. R. and F. Koppelman (1999). A Retrospective and Prospective Survey of TimeUse Research. Transportation, 26(2), 119-139. Bhat, C. R., and R. Misra (1998). Discretiona ry Activity Time Allo cation of Individuals between In-Home and Out-Of-Home a nd between Weekdays and Weekends. Transportation 26(2), 193-229.

PAGE 230

215 Bhat, C. R. and S. K. Singh (2000). A Comprehensive Daily Activity-Travel Generation Model System for Workers. Transportation Research Part A, 34(1), 1-22. Bhat, C. R. and S. Srinivasan (2005). A Multidimensional Mixed Ordered-Response Model for Analyzing Weekend Activity Participation. Transportation Research Part B, 39(3), 255-278. Bhat, C.R., and T. K. Lawton (2000) Passenger Travel Demand Forecasting. Transportation in the New Millennium, Transportation Research Board. Boarnet, M., and R. Crane (2001). The In fluence of Land Use on Travel Behavior: Specification and Estimation Strategies. Transportation Research A, 35(9), 823845. Bollen, K.A. (1989). Structural Equations with Latent Variables Wiley, New York. Browning, M., and P.A. Chiappori (1998). E fficient Intra-House hold Interactions: A General Characterization and Empirical Tests. Econometrica 66(6), 1241-1278. Chandrasekharan, B., and K. G. Goulias (1999). Exploratory Longitudinal Analysis of Solo and Joint Trip Making in the Puget Sound Transportation Panel. Transportation Research Record. 1676, 77-85. Chaplin, F.S. (1974). Human Activity Patterns In The City John Wiley & Sons, New York. Ettema, D. F., A. W. J. Borgers, and H. J. P. Timmermans (1993). A Simulation of Activity Scheduling He uristics (SMASH). Transportation Research Record, 1413, 1-11. Ewing, R., and R. Cervero (2001). Travel and the Built Environment. Transportation Research Record 1780, 87. Ewing, R., T. Schmidt, R. Killingsworth, A. Zlot, and S. Raudenbush (2003). Relationship between Urban Sprawl and Physical Activity, Obesity and Morbidity. American Journal of He alth Promotions Inc, 18(1), 47-57. Fujii, S., R. Kitamura, and K. Kishizawa (1999) Analysis of Individuals Joint Activity Engagement Using a Model System of Activity-travel Behavior and Time Use. In Transportation Research Record, 1676, 11-19. Fujii, S. and R. Kitamura (2000). Evaluati on of Trip-inducing Effects of New Freeways Using a Structural Equations Model System of Commuters' Time use and Travel, Transportation Research B 34(5), 339-354.

PAGE 231

216 Grling, T., K. Brannas, J. Garvill, R. G. Golledge, S. Gopal, E. Holm, and E. Lindberg (1989). Household Activity Scheduling. Proceedings of the Fifth World Conference on Transport Research Yokohama, Japan. Golob, T. F. and M. G. McNally (1995). A M odel of Household Interactions in Activity Participation and the Derived Demand in Travel. Proceedings of the Conference on Activity-Based Approaches European Institute of Retailing and Service Studies (EIRASS), Eindhoven, The Netherlands, May. Golob, T.F., and M.G. McNally (1997). A Mode l of Activity Participation and Travel Interactions between Household Heads. Transportation Research B 31(3), 177194. Golob, T.F. (1998). A Model of Household Demand for Activity Participation and Mobility. In T. Grling, T. La itilla and K. Westin (eds.) Theoretical Foundations of Travel Choice Modeling Pergamon, 365-398. Golob, T. F. (2000). A Simultaneous Model of Household Activity Participation and Trip Chain Generation. Transportation Research Part B 34(5), 355-376. Goodwin, P.B. (1981). The Usef ulness of Travel Budgets. Transportation Research A, 15, 97-106. Greene, W. H. (2002). LIMDEP Reference Guide 8.0, Econometric Software Inc., Plainview, New York. Hgerstrand, T. (1970). What a bout People in Regional Science? Papers of the Regional Science Association 24, 7-21. Handy, S. L., M. G. Boarnet, R. Ewing and R. E. Killingsworth (2002). How the Built Environment Affects Physical Activity: Views from Urban Planning. American Journal of Preventive Medicine, 23(2), Supplement 1, 64-73. Harvey, A.S. (1986). Social Contact and Use of Time in Canada and Norway: A Comparative Analysis of Time Use Patterns. In Aas, D., Havey, A.S., Wunlipinski, E., Niemi, I. (eds.), Time Use Studies: Dimensions and Applications. Central Statistical Office of Finland, Helsinki. Hamed, M. M., and F. L. Mannering (1993) Modeling Travelers Postwork Activity Involvement: Towards A New Methodology. Transportation Science 27(4), 381394. Hanson, S. and J. O. Huff (1988). Repetition and Day-To-Day Variab ility in Individual Travel Patterns: Implicat ions for Classifications. In Golledge, R. and

PAGE 232

217 Timmermans, H. (eds.), Behavioral Modelling in Geography and Planning Croom Helm, New York. Hoehner, C., L. K. Brennan, R. Browns on, S. Handy, and R. Killingsworth (2003). Opportunities for Integrating Public H ealth and Urban Planning Approaches to Promote Active Community Environments. American Journal of Health Promotion, 18(1), 14-20. Hirsh, M., J. N. Prashker, and M. Ben-Akiva (1986). Dynamic Model of Weekly Activity Pattern. Transportation Science 29(1), 24-36. Hupkes, G. (1982). The Law of Consta nt Travel Time and Trip Rates. Futures February, 38-48. Jones, P. M., M. C. Dix, M. I. Clarke, and I. G. Heggie (1983). Understanding Travel Behavior. Aldershot, England: Gower. J reskog, K., and D. S rbom (1999)a. PRELIS 2: Users Reference Guide. SSI Scientific Software International, Plainview, Chicago, IL. J reskog, K., and D. S rbom (1999)b. LISREL 8: Users Reference Guide. SSI Scientific Software International, Plainview, Chicago, IL. Juster, F.T. (1985). A Note on Recent Changes in Time Use. In Juster, F.T., Stafford, F.P. (eds.), Time, Goods and Well-Being University of Michigan, Ann Arbor, Survey Research Center, Institute for Social Research, 313-332. Kain, J.F. The Journey to Work as a Determinant of Residential Location. Papers and Proceedings, Regional Science Association, Vol. 9, 1962, pp. 137. Kim, S. G., M. Hamed, and F. Mannering ( 1993). A Note on Modeling Travelers HomeStay Duration and Efficiency of Propor tional Hazard Model. Working paper, Department of Civil Engineeri ng, University of Washington. Kitamura, R. (1984). A Model of Daily Time Allocation of Discre tionary Out-of-Home Activities and Trips. Transportation Research Part B, 18(3) 255-266. Kitamura, R., E. I. Pas, C.V. Lula, T. K. Lawton, and P. E. Benson (1996). The Sequenced Activity Mobility Simulator: An Integral Approach to Modeling Transportation, Land Use and Air Quality. Transportation, 23, 267-291. Kitamura, R., J. Robinson, T. F. Golob, M. Bradley, J. Leonard, and T. van der Hoorn (1992). A Comparative Analysis of Time Use Data in the Netherlands and California, Research Report Number: UCD-ITS-RR-92-9, Institute of Transportation Studies, University of California, Davis, June.

PAGE 233

218 Kitamura, R., and S. Fujji (1996). Two Comput ational Process Models of Activity-Travel Behavior. Proceedings of the Conference on th e Theoretical Foundation of Travel Choice Modeling Stockholm, August 7-11. Kitamura, R., S. Fujii, and E.I. Pas (1997) Time-use Data, Analysis and Modeling: Toward the Next Generation of Tr ansportation Planning Methodologies. Transport Policy, 4(4), 225-235. Kitamura, R., T. Yamamoto, K. Kishizawa, and R. M. Pendyala (2000). Stochastic Frontier Models of Prism Vertices. Transportation Research Record 1718, 18-26. Kondo, K., and R. Kitamura (1987). Time-Space Constraints and the Formation of Trip Chains. Regional Science and Urban Economics 17, 49-65. Koppelman, F. S. and T. A. Townsend (1987). Task Allocation among Household Members: Theory and Analysis. Proceedings of the Fifth International Conference on Travel Behavior Aix-en-Provence, France. Kostyniuk, L.P., and R. Kitamura (1986). Hous ehold Lifecycle: Predictor of Travel Expenditure. In Behavioural Research for Transport Policy, VUN Science Press, Utrecht, The Netherlands, 343-362. Krizek, K. J., and P. Waddell (2002). Analys is of Lifestyle Choices: Neighborhood Type, Travel Patterns, and Activity Participation. Transportation Research Record, 1807, 119-128. Muth, R. (1969). Cities and Housing: The Spatial Pa ttern of Urban Residential Land Use. University of Chicago Press, Chicago, IL. Kumar, A., and D. Levinson (1995). Tempor al Variations on Allocation of Time. Transportation Research Record, 1493, 118-127. Levinson, D., and A. Kumar (1995). Activit y, Travel, and the Allocation of Time. APA Journal, Autumn, 458-470. Levinson, D., and Y. Wu (2005). The Rationa l Locator Reexamined: Are Travel Times Still Stable? Presented at 84th Annual Meeting of the Transportation Research Board, Washington, D.C. Lingsom, S., and A. L. Ellingsaeter (1983). Wo rk, Leisure and Time Spent With Others: Changes in Time Use in the 70s. In Statistiske analyser Kongsvinger, Statistisk Sentralbyra, Oslo. Lu, X., and E. I. Pas (1999). Socio-Demogr aphics, Activity Participation, and Travel Behavior. Transportation Research Part A, 33(1), 1-18.

PAGE 234

219 Lu, X., and E. I. Pas (1997). A Structural Equation Model of the Relationships among Socio-Demographics, Activity Par ticipation and Travel Behavior. Proceedings of the 76th Annual Meeting of the Tr ansportation Research Board, Washington D.C. Ma, J., and K. G. Goulias (1997). A Dynamic Analysis of Person and Household Activity and Travel Patterns using Data from the First Two Waves in the Puget Sound Transportation Panel. Transportation 24(3), 309-331. Mannering, F., E. Murakami, and S.G. Kim (199 2). Models of Travelers Activity Choice and Home-Stay Duration: Analysis of Functional Form and Temporal Stability. Working paper, Department of Civil Engineering, University of Washington. Meka, S., R. M. Pendyala, and M. W. A. Ku mara (2002). Structural Equations Analysis of Within-Household Activity and Ti me Allocation between Two Adults. Proceedings of the 81st Annual Meeting of the Transportation Research Board Washington, D.C. Melon, I., L. Guala, and A. Loddo (2004). Ti me Allocation to Discretionary In-home, Out-of-home Activities and to Trips. Transportation, 31, 69. Miller, E. J. (1991). Modeling Accessibility Using Space-Ti me Prism Concepts within Geographical Information Systems. International Journal of Geographical Information Systems, 5, 287-301. Mokhtarian, P.L. (2005). Travel as a Desired End, Not Just a Means. Transportation Research A, 39(2-3), 93-96. Mokhtarian, P. L., and C. Chen (2004). TTB or Not TTB, That is the Question: A Review and Analysis of the Empirical Literature on Travel Time (and Money) Budgets. Transportation Research A 38(9-10), 643-675. Mokhtarian, P. L., and I. Salomon (2001). Ho w Derived Is the Demand for Travel? Some Conceptual and Measurement Considerations. Transportation Research A 35(8), 695-719. Mokhtarian, P. L., I. Salomon, and L. Redmond (2001). Understanding the Demand for Travel: Its Not Purely Derived. Innovation: The European Journal of Social Science Research 14(4), 355-380. Ministry of Road Transport and Highways (2003). Handbook on Transport Statistics in India. Transport Research Office, Ministry of Road Transport and Highways, New Delhi, India.

PAGE 235

220 Nehra, R. S., A. Banerjee, R. M. Pendyala, and S. L. Dhingra (2004). An Analysis of Personal Time Space Prism Vertex Locations in a Developing Context. Transportation Research Record, 1898, 19-27. Ng, W., and L. Schipper (2006). China Mo torization Trends, Consequences, and Alternatives. Proceedings of the 85PthP Annual Meeting of Transportation Research Board, Washington D.C. Niemeier, D. A., and J. G. Morita (1996) Duration of Trip-Making Activities by Men and Women. Transportation, 23, 353-371. Office of the Registrar General of India (2001). The Census of India 2001: Rural-urban distribution of population in India New Delhi, India. Available at: http://www.censusindia.net/results/rudist.html. Organization for Economic Co-operation and Development (2002). National Accounts Database. Available at: http:// www.oecd.org/statsportal/. Pas, E. I., and F. S. Koppelman (1987). An examination of the determinants of day-today variability in individual s’ urban travel behavior. Transportation 14, 3-20. Pas, E. I., and K. G. Sundar (1995) IntraPersonal Variability in Daily Urban Travel Behavior: Some Additional Evidence. Transportation 22, 135-150. Pendyala, R. M. (2003). Time Use and Travel Behavior in Space and Time. In Goulias, K. G. (ed.), Transportation Systems and Planni ng: Methods and Applications CRC Press, New York, 2-1 to 2-37. Pendyala, R.M., and C.R. Bhat (2004) Emergi ng Issues in Travel Behavior Analysis. Resource paper for workshop on Emerging Issues : National Household Travel Survey Conference Washington, D.C., November. Pendyala, R. M., and K. G. Goulias (2002). Time Use and Activity Perspectives in Travel Behavior Research. Transportation 29, 1-4. Pendyala, R. M., T. Yamamoto, and R. Kita mura (2002). On The Formulation of Time Space Prisms to Model Constraints on Personal Activity-Travel Engagement. Transportation 29(1), 73-94. Pez, A., P. Suthanaya, and J. Black (2001). A Spatial Analysis of Transportation Mode Specific Journey-To-Work Commuting Prefer ences: Implications for Sustainable Transport Policy. In Proceedings of the 9th World Conference on Transport Research, World Conference on Transport Research Society, Seoul, Korea.

PAGE 236

221 Pisarski, A. E. (1996). Commuting in America II. The Second Annual Report on Commuting Patterns and Trends. Eno Foundation for Transportation Incorporated, Lansdowne, VA. Principio, S.L., and E. I. Pas (1997). The Soci odemographics and Travel Behavior of Life Style Groups Identified By Time Use Patterns. Presented at 76PthP Annual Meeting of the Transportation Research Board Washington, D.C. Purvis, C. (1994). Changes in Regional Tr avel Characteristics and Travel Time Expenditures in San Francisco Bay Area: 1960-1990. Transportation Research Record, 1466, 99-109. Pucher, J., N. Korattyswaroopam, N. Mittal, and N. Ittyerah (2005). Urban Transport Crisis in India. Transport Policy 12(3), 185-198. Robinson, J, P. E. Converse, and A. Szalai (1972). Everyday Life in Twelve Countries. In Szalai, A. (ed.), The Use of Time Mouton, The Hague. Robinson, J. P., and G. Godbey (1999). Time for Life: The Surprising Ways Americans Use Their Time. Pennsylvania State University Press, University Park, PA. Schafer, A. (2000). Regularities in Travel Demand: An Interna tional Perspective. Journal of Transportation and Statistics 3(3), 1-31. Schafer, A, and V. G. Victor (2000). The Future Mobility of the World Population. Transportation Research A 34(3), 171-205. Simma, A., and K. W. Axhausen (2001). With in-Household Allocation of Travel: Case of Upper Austria. Transportation Research Record, 1752, 69-75. Sommer, B., and R. Sommer (1997). A Practical Guide to Behavioral Research: Tools and Techniques. Fourth Edition, Oxford University Press, New York. Sperling, D., and E. Claussen (2002). The De veloping World’s Motorizing Challenge. Issues in Science and Technology, Fall, 59-66. Stopher, P. R., and R.V. Vadarevu (1995). An I nvestigation of the Effect of Life Cycle on Household Travel Behavior. Presented At The 8PthP REAAA Conference Taipei, Taiwan. Szalai, A. (1972). The Use of Time Mouton Publishers, The Hauge. Timmermans, H., X. Hagen, and A. Borger s (1992). Transportation Systems, Retail Environment and Pedestrian Trip Ch aining Behavior: Modeling Issues and Applications. Transportation Research Part B 26(1), 45-59.

PAGE 237

222 Toole-Holt, L, S. E. Polzin, and R. M. Pendyala (2005). Two Minutes per Person per Day Each Year: An Exploration of the Gr owth in Travel Time Expenditures. Presented at 84PthP Annual Meeting of the Transportation Research Board Washington, D.C. Townsend, T. A. (1987). The Effects of H ousehold Characterist ics on the Multi-Day Time Allocations and Travel-Activity Patterns of Households and Their Members. Unpublished Ph.D. Dissertation, Northwestern University, Evanston, IL. Vaddepalli, S. (2004). An Analysis of Characteristics of Long and Short Commuters in the United States. M.S. Thesis, University of South Florida, Tampa, FL. van Wissen, L. (1989). A Model of Househol d Interactions in Activity Patterns. Presented at the International Confer ence on Dynamic Travel Behavior Analysis Kyoto University, Kyoto, Japan, 16-17 July. Vasconcellos, E. (2001). Urban Transport, Environment and Equity: The Case for Developing Countries Earthscan Press, London, UK. Waddell, P. (2002). UrbanSim: Modeli ng Urban Development for Land Use, Transportation and Environmental Planning. Journal of the American Planning Association, 68, 297-314. Waldman, A. (2005). Mile by Mile, India Paves a Smoother Road to Its Future. New York Times, December 4, New York, NY. William, A.V., Y. Huang, and S. Withers (2003). Does Commuting Distance Matter? Commuting Tolerance and Residential Change. Regional Science and Urban Economics 33, 199-221. Wingo, L. (1961). Transportation and Urban Land Resources for the Future, Inc., Washington, DC, pp. 132. Yamamoto, T. and Kitamura, R. (1999). An An alysis of Time Allocation to In-home and Out-of-home Discretionary Activities across Working Days and Non-working Days. Transportation, 26, 211–230. Yamamoto, T., R. Kitamura, and R. M. Pendya la (2004). Comparative Analysis of TimeSpace Prism Vertices for Out-Of-Home Activity Engagement on Working and Non-Working Days. Environment and Planni ng B: Planning and Design, 31(2), 235-250.

PAGE 238

223 Zahavi, Y. (1979). “UMOT” Project Report DOT-RSPA-DPB-20-79-3, USDOT, Washington, D.C. and Ministry of Tr ansport, Federal Republic of Germany, Bonn., August. Zahavi, Y, and A. Talvitie (1980). Regularitie s in Travel Time and Money Expenditures. Transportation Research Record, 750, 13-19. Zahavi, Y, and J. Ryan (1980). Stabi lity of Travel Components Over Time. Transportation Research Record, 750, 19-26.

PAGE 239

224 APPENDICES

PAGE 240

225 Appendix 1: Formulation of Producti on Frontier Model with Logarithmic Transformation of Dependent Variable If u is a random variable following the sta ndard half-normal distribution, then the probability density function of u f( u ) = ) 2 u exp( 2 22 u 2 u u > 0. Let m = exp(u ) and pdf( m ) and cdf( m ) represent the probability density function and cumulative distribution function with respect to m where u 0, thus 0 < m 1. cdf( m ) = P( M < m ) = P[ exp (u ) < m ] = P[ u > ln ( m )] = ) m ln( 0 ) m ln(dv ) v ( f 1 dv ) v ( f pdf( m ) = dm dv ) v ( f d dm m M d) m ln( 0 ) P(. By Leibnitz’s theorem, ), x ( )] x ( x [ f ) x ( )] x ( x [ f dy ) y x ( f dx d ) x ( ') x ( ) x ( Then, pdf( m ) = f[ ln ( m )] [ln ( m )]P’P = f[ ln ( m )]/ m = 2 u 2 u2 ) m ( ln exp m 2 2 1 0 2 u 2 u 1 0dm 2 ) m ( ln exp 2 2 dm ) m ( pdf m m E Let ln ( m ) = w then dm = exp ( w ) dw Replacing dm we obtain, 0 2 2 2 u u 2 udw ) w ( 2 1 exp 2 1 2 / exp m 2 E Let ( w -BuP B2P)/BuB = then dw = BuBd Replacing dw ) ( 1 ) 2 / exp( 2 ) ( ) 2 / exp( 2 dw 2 exp 2 1 2 / exp mu 2 u u 2 u 2 2 uu 2 E

PAGE 241

226 ABOUT THE AUTHOR Mr. Amlan Banerjee graduated with a P h.D. degree in Transportation Systems Engineering from the Department of Civ il and Environmental Engineering at the University of South Florida in 2006 unde r the guidance of Dr. Ram Pendyala. He completed his Master's degree in Trans portation Systems Engineering in 2002 from Indian Institute of Technology Bombay and he ld a Bachelor degree in Civil Engineering from Jadavpur University in India in 2000. His primary areas of doctoral research include understanding of activity engagement and ti me-use patterns in a developing country context and the advancement of state of th e art transportation modeling methods. He served as a treasurer for the USF ITE Stude nt Chapter. He was awarded the 2005 Georgia Brosch Memorial Transportation Scholarship by the Center for Urban Transportation Research at USF and in the following y ear, he was awarded International Road Federation Executive Fellowship.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001913930
003 fts
005 20071018091052.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 071018s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001729
040
FHM
c FHM
035
(OCoLC)174958785
049
FHMM
090
TA145 (ONLINE)
1 100
Banerjee, Amlan.
0 245
Understanding activity engagement and time use patterns in a developing country context
h [electronic resource] /
by Amlan Banerjee.
260
[Tampa, Fla] :
b University of South Florida,
2006.
3 520
ABSTRACT: Flourishing economy, rapid industrialization and increasing trend of motorization have been shaping societies in the developing countries like India in an unprecedented manner.Infrastructure backlog amid such rapid growth in all imaginable directions has heavily exacerbated the urban transport crisis in these countries by alarming increase in vehicular travel demand, road fatalities, and environmental pollution. To address urban transport challenges, the necessary development and implementation of effective transport planning and policies have generally lagged in the developing countries compared to that seen in the developed countries due to several constraints including resource constraints, knowledge constraints, institutional constraints and so on. However, in the recent past, with the rapid development seen by several emerging economies and the explosive growth in transportation infrastructure investment, there is a growing interest in the development and implementati on of advanced travel demand modeling systems in developing countries. But lack of necessary research and exploration of travel behavior in a developing country context has left very limited knowledge for us to understand the extent of applicability of these advanced theories and methodologies in a different socio-cultural perspective. Assessing the practical relevance of the subject, this research adopts a comprehensive approach to explore the activity engagement pattern and time use behavior from a developing country standpoint. To accomplish this goal, a series of empirical and analytical studies are performed on a household travel survey data set available from Thane Metropolitan Area in India. The study also introduces new concepts and facilitates enhancements of existing modeling methodologies in the field of travel behavior and time use research. The study results provide very insightful findings and plausible interpretations consistent with a developing country perspective reco gnizing a wide spectrum of differences and similarities in activity patterns and time use behavior between a developed and a developing country. Specified model structures are meaningfully able to incorporate various socio-cultural and institutional constraints and reflected sensitivity to the behavioral variability between the contexts suggesting that advanced analytical techniques may be satisfactorily applied on the data set from developing countries which may contribute important ingredients in the development of advanced activity-based model system in the countries like India.
502
Dissertation (Ph.D.)--University of South Florida, 2006.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 225 pages.
Includes vita.
590
Adviser: Ram M. Pendyala, Ph.D.
653
Travel time expenditure.
Commuter behavior.
Inter-agent interaction.
Stochastic frontier model.
Structural equations model.
International comparison.
690
Dissertations, Academic
z USF
x Civil Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1729