USF Libraries
USF Digital Collections

Modeling time space prism constraints in a developing country context

MISSING IMAGE

Material Information

Title:
Modeling time space prism constraints in a developing country context
Physical Description:
Book
Language:
English
Creator:
Nehra, Ram S
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
travel behavior
comparisons of behavior
stochastic frontier models
household travel survey
activity based travel demand modeling
Dissertations, Academic -- Civil Engineering -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Recent developments in microsimulation modeling of activity and travel demand have called for the explicit recognition of time-space constraints under which individuals perform their activity and travel patterns. The estimation of time-space prism vertex locations, i.e., the perceived time constraints, is an important development in this context. Stochastic frontier modeling methodology offers a suitable framework for modeling and identifying the expected vertex locations of time space prisms within which people execute activity-travel patterns. In this work, stochastic frontier models of time space prism vertex locations are estimated for samples drawn from a household travel survey conducted in 2001 in the city of Thane on the west coast of India and National Household Travel Survey 2001, United States. This offers an opportunity to study time constraints governing activity travel patterns of individuals in a developing as well as developed country context. The work also includes comparisons between males and females, workers and non-workers, and developed and developing country contexts to better understand how socio-economic and socio-cultural norms and characteristics affect time space prism constraints. It is found that time space prism constraints in developing country data set can be modeled using the stochastic frontier modeling methodology. It is also found that significant differences exist between workers and non-workers and between males and females,possibly due to the more traditional gender and working status roles in the Indian context. Finally, both differences and similarities were noticed when comparisons were made between results obtained from the data set of India and United States. Many of these differences can be explained by the presence of other constraints including institutional, household, income, and transportation accessibility constraints that are generally significantly greater in the developing country context.
Thesis:
Thesis (M.S.C.E.)--University of South Florida, 2004.
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 Ram S. Nehra.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 80 pages.

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 - 001469398
oclc - 55731053
notis - AJR1152
usfldc doi - E14-SFE0000299
usfldc handle - e14.299
System ID:
SFS0024994:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Modeling Time Space Prism Constraints in a Developing Country Context by Ram S. Nehra A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Civil Engineering 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., P.E. Jian J. Lu, Ph.D., P.E. Date of Approval: March 31, 2004 Keywords: Activity based travel demand modeling, tr avel behavior, household travel survey, stochastic frontier models, comparisons of behavior Copyright 2004, Ram S. Nehra

PAGE 2

DEDICATION This work is dedicated to my family in India.

PAGE 3

ACKNOWLEDGEMENTS I wish to express sincere appreciation to my major professor Dr. Ram Pendyala for his valuable guidance and support for the thesis work. Thanks to Dr. Steven Polzin and Dr. Jian Lu for agreeing to serve on my supervisory committee. I would like to express my gratitude for my co lleague Mr. Amlan Banerjee for helping me at every step. His hard work and company helped me in getting this work done. Thanks to Dr. S.L. Dhingra from IIT Bombay, India for making Thane data available. Special thanks are also extended to Xin Ye and Abdul Pinjari and other colleagues from Transportation Engineering group and my friends at University of South Florida who never let me feel away from home.

PAGE 4

i TABLE OF CONTENTS LIST OF TABLES ii LIST OF FIGURES iii ABSTRACT vi CHAPTER 1: INTRODUCTION 1 1.1 Activity Based Travel Demand Modeling 1 1.2 Concept of Space and Time 1 1.3 Problem Definition 2 1.4 Objective 3 1.5 Organization of the Thesis 4 CHAPTER 2: LITERATURE REVIEW 5 CHAPTER 3: MODELING METHODOLOGY 8 CHAPTER 4: DATA PREPARATION AND SAMPLE CHARACTERISTICS 11 4.1 Thane Household Travel Survey 11 4.2 Household and Person Socio-economi c Characteristics of Thane Survey 12 4.3 National Household Travel Survey 17 4.4 Household and Person Socio-econo mic Characteristics of NHTS 2001 18 CHAPTER 5: MODEL ESTIMATION RESULTS: THANE DATA 23 5.1 Origin Vertex Location 29 5.2 Terminal Vertex Location 34 CHAPTER 6: MODEL ESTIMATION RESULTS: NHTS DATA 39 6.1 Origin Vertex Location 43 6.2 Terminal Vertex Location 48 CHAPTER 7: INTERNATIONAL COMPARISONS 53 7.1 Origin Vertex Location 54 7.2 Terminal Vertex Location 60 CHAPTER 8: SUMMARY AND CONCLUSIONS 66 REFERENCES 70

PAGE 5

ii LIST OF TABLES Table 4.1 Household Characteristics of 2001 Thane, India Survey Sample 13 Table 4.2 Personal Characteristics of 2001 Thane, India Survey Sample 14 Table 4.3 Personal Transportation Characteristics of 2001 Thane, India Survey Sample 15 Table 4.4 Household Characteristics of NHTS 2001 Sample 19 Table 4.5 Personal Characteristics of NHTS 2001 Sample 20 Table 5.1 Description of Explanatory Variables Used in Thane Models 23 Table 5.2 Stochastic Frontier Models of Origin Vertex Location: Thane Sample 25 Table 5.3 Stochastic Frontier Models of Terminal Vertex Location: Thane Sample 27 Table 6.1 Description of Explanatory Variables Used in NHTS 2001 Models 39 Table 6.2 Stochastic Frontier Models of Origin Vertex Location: NHTS 2001 41 Table 6.3 Stochastic Frontier Models of Terminal Vertex Location: NHTS 2001 42 Table 7.1 Comparisons of Average Differences between Expected Vertex Location and Actual Departure/Arrival Time (E[u] values) 53

PAGE 6

iii LIST OF FIGURES Figure 1.1 Hagerstrand Time Space Prism 2 Figure 5.1 Distribution of Origin Vertex and First Departure Time (Thane): All Workers vs. All Non-workers 29 Figure 5.2 Distribution of Origin Vertex and First Departure Time (Thane): Male Workers vs. Female Workers 30 Figure 5.3 Distribution of Origin Vertex and First Departure Time (Thane): Male Non-workers vs. Female Non-workers 31 Figure 5.4 Distribution of Origin Vertex and First Departure Time (Thane): Male Workers vs. Male Non-workers 32 Figure 5.5 Distribution of Origin Vertex and First Departure Time (Thane): Female Workers vs. Female Non-workers 33 Figure 5.6 Distribution of Terminal Vert ex and Final Arrival Time (Thane): All Workers vs. All Non-workers 34 Figure 5.7 Distribution of Terminal Vert ex and Final Arrival Time (Thane): Male Workers vs. Female Workers 35 Figure 5.8 Distribution of Terminal Vert ex and Final Arrival Time (Thane): Male Non-workers vs. Female Non-workers 36 Figure 5.9 Distribution of Terminal Vert ex and Final Arrival Time (Thane): Male Workers vs. Male Non-workers 37 Figure 5.10 Distribution of Origin Vertex and First Departure Time (Thane): Female Workers vs. Female Non-workers 38 Figure 6.1 Distribution of Origin Vertex and First Departure Time (NHTS): All Workers vs. All Non-workers 43 Figure 6.2 Distribution of Origin Vertex and First Departure Time (NHTS): Male Workers vs. Female Workers 44 Figure 6.3 Distribution of Origin Vertex and First Departure Time (NHTS): Male Non-workers vs. Female Non-workers 45

PAGE 7

iv Figure 6.4 Distribution of Origin Vertex and First Departure Time (NHTS): Male Workers vs. Male Non-workers 46 Figure 6.5 Distribution of Origin Vertex and First Departure Time (NHTS): Female Workers vs. Female Non-workers 47 Figure 6.6 Distribution of Terminal Ve rtex and Final Arrival Time (NHTS): All Workers vs. All Non-workers 48 Figure 6.7 Distribution of Terminal Ve rtex and Final Arrival Time (NHTS): Male Workers vs. Female Workers 49 Figure 6.8 Distribution of Terminal Ve rtex and Final Arrival Time (NHTS): Male Non-workers vs. Female Non-workers 50 Figure 6.9 Distribution of Terminal Ve rtex and Final Arrival Time (NHTS): Male Workers vs. Male Non-workers 51 Figure 6.10 Distribution of Origin Vertex and First Departure Time (NHTS): Female Workers vs. Female Non-workers 52 Figure 7.1 Distribution of Origin Vertex and First Departure Time (NHTS): All Workers Thane Survey vs. All Workers NHTS 54 Figure 7.2 Distribution of Origin Vertex and First Departure Time: All Non-workers Thane Survey vs. All Non-workers NHTS 55 Figure 7.3 Distribution of Origin Vertex and First Departure Time: Male Workers Thane Survey vs. Male Workers NHTS 55 Figure 7.4 Distribution of Origin Vertex and First Departure Time: Female Workers Thane Survey vs. Female Workers NHTS 56 Figure 7.5 Distribution of Origin Vertex and First Departure Time: Male Non-workers Thane Survey vs. Male Non-workers NHTS 57 Figure 7.6 Distribution of Origin Vertex and First Departure Time: Female Non-workers Thane Survey vs. Female Non-Workers NHTS 57 Figure 7.7 Distribution of Terminal Vertex and Final Arrival Time: All Workers Thane Survey vs. All Workers NHTS 60 Figure 7.8 Distribution of Terminal Vertex and Final Arrival Time: All Non-workers Thane Survey vs. All Non-workers NHTS 61 Figure 7.9 Distribution of Terminal Vertex and Final Arrival Time: Male Workers Thane Survey vs. Male Workers NHTS 61 Figure 7.10 Distribution of Terminal Vertex and Final Arrival Time: Female Workers Thane Survey vs. Female Workers NHTS 62

PAGE 8

v Figure 7.11 Distribution of Terminal Vertex and Final Arrival Time: Male Non-workers Thane Survey vs. Male Non-workers NHTS 63 Figure 7.12 Distribution of Terminal Vertex and Final Arrival Time: Female Non-workers Thane Survey vs. Female Non-Workers NHTS 63

PAGE 9

vi MODELING TIME SPACE PRISM CONSTRAINTS IN A DEVELOPING COUNTRY CONTEXT Ram S. Nehra ABSTRACT Recent developments in microsimulation modeling of activity and travel demand have called for the explicit recognition of time-space constraints under which individuals perform their activity and travel patterns. The estimation of time-spa ce prism vertex locations, i.e., the perceived time constraints, is an important development in this context. Stochastic frontier modeling methodology offers a suitable framework for m odeling and identifying the expected vertex locations of time space prisms within which people execute activity-travel patterns. In this work, stochastic frontier models of time space prism vertex locations are estimated for samples drawn from a household travel survey conducted in 2001 in the city of Thane on the west coast of India and National Household Travel Survey 2001, United St ates. This offers an opportunity to study time constraints governing activity travel pattern s of individuals in a developing as well as developed country context. The work also includes comparisons between males and females, workers and non-workers, and developed and deve loping country contexts to better understand how socio-economic and socio-cultural norms and characteristics affect time space prism constraints. It is found that time space prism constraints in developing country data set can be modeled using the stochastic frontier modeling me thodology. It is also found that significant differences exist between workers and non-workers and between males and females, possibly due to the more traditional gender and working status roles in the Indian context. Finally, both differences and similarities were noticed when comparisons were made between results obtained from the data set of India and United States. Many of these differences can be explained by the presence of other constraints including institu tional, household, income, and transportation accessibility constraints that are generally significan tly greater in the developing country context.

PAGE 10

CHAPTER 1 INTRODUCTION 1.1 Activity Based Travel Demand Modeling Over the past couple of decades, after the realization that building roads is not permanent solution to congestion and increasingly complex patterns of travel demand, the emphasis of transportation planning has shifted to the effective management of travel demand. A host of strategies, so called Travel Demand Management (TDM) strategies, are aimed at effectively managing and distributing travel demand both in spatial and temporal dimensions. As the role of TDM strategies spread and transportation planners had more insight in the human travel behavior patterns, it became apparent that the traditional trip based four-step procedure for travel demand forecasting and modeling is not able to address the complex questions raised by TDM implementation. Meanwhile, activity based approach entered into travel behavior research area and offered potentially effective and practical tool for TDM. Activity-based approaches explicitly recognize that travel demand is derived from the need to pursue activities that are dispersed in time and space. Moreover, these approaches recognize the inter-dependence among decisions for a series of trips made by an individual. They also recognize the interactions among various members of the household, that arise when household members allocate resources (such as household vehicles) to themselves, assign and share tasks, and jointly engage in activities. As such, it has been argued that activity-based approaches provide a theoretically and conceptually stronger framework within which travel demand modeling may be performed. 1.2 Concept of Space and Time In a given day, a person has only 24 hours available and much of that time may be spent on basic subsistence activities including sleeping, working and person/household care. The temporal aspect of these activities tend to be rigid and impose constrain on an individuals potential travel

PAGE 11

engagement patterns. Similar arguments may be made regarding space. Depending on the time available and the speed of travel, there is only a certain amount of distance that can be covered and the number of possible locations that can be visited (within this distance) is limited as well. Thus, it can be seen that time-space prism constraints play an important in shaping peoples activity-travel pattern and an individuals travel behavior is greatly influenced by space and time constraints. Hgerstrand (1970) introduced the concept of the time-space prism to represent the finite spatio-temporal action space in which an individual can pursue activities. B Time v A H W Location Figure 1.1 Hgerstrand Time Space Prism The Hagerstrand prism (Figure 1.1) explains that one can not be outside the prism formed by locations W and H in give time A to B due to speed (v) constraints. So all the out of home activities of an individual are performed within the prism. Thus, if one desires to predict activity engagement patterns such as activity type, activity start time, activity duration, trip start time, trip duration, and so on, it is imperative that the end points of the time space prism be determined. If the temporal end points or extremities of the time space prism are determined, then the spatial constraints can simply be determined by multiplying the time span available in the prism with the speed of travel. 1.3 Problem Definition As microsimulation models of travel demand become increasingly prevalent, it is important to initiate research where models traditionally being estimated on data sets from developed contexts

PAGE 12

are now estimated on data sets from developing country contexts. The time is ripe to explore the feasibility of applying these methods in very different socio-cultural and transportation system contexts. The study of differences between males and females and between workers and non-workers is a classic one that has been of interest to travel behavior researchers for decades. Separate models of activity and travel demand are often estimated for different socio-economic groups because of potential structural differences in their behavior and decision making. Workers face constraints that are not typically encountered by non-workers due to the rigidity of the work location and schedule. Similarly, gender roles in households (particularly in a developing country context such as India) may suggest fundamental differences in time space prism constraints that warrant separate models for these groups. This hypothesis is tested within this work. 1.4 Objective This work builds upon the previous work done in this area to estimate and identify time space prism vertex location distributions for a survey sample from a developing country. Two large survey samples one from the city of Thane on the west coast of India (near Bombay/Mumbai) and another from National Household Travel Survey (NHTS) offer data sets for accomplishing this goal. In addition to facilitating a comparison of time space prism vertices between developed and developing country contexts, the data set also allows comparisons between workers and non-workers and between males and females within the data sets used in this study. India has experienced significant economic and technological growth over the past few decades. The high-tech industry in India has flourished during the past decade and contributed to considerable growth in Indias economy and participation in global trade. With more than one billion people in population, the country is the second most populous nation after China. While there has been a trend towards adopting some of the ways of the west, particularly among the younger age groups, the country has retained much of the social norms and culture that have historically defined it. In general, labor participation rates of women are significantly lower than those of men, gender roles are rather traditional with women undertaking the major share of household obligations, family ties are strong (strong intra-household interactions), and vehicle ownership and affordability are low although vehicle ownership is growing rapidly. So a large number of socio-economic factors make different the developing country India to developed country United States and consequently the travel behavior and transportation characteristics of people in these countries are different too.

PAGE 13

In this work, models are developed to locate prism vertices along the time axis. The models are formulated as stochastic frontier models, which are used to estimate the location of unobservable frontiers (prism vertices) based on the measurement of an observable variable that is governed by the frontier. The observable variable is the trip departure or arrival time that must occur either after an origin (beginning) prism vertex or before a terminal (end) vertex respectively. In particular, the work focuses on following items of interest: Formulation and estimation of time vertices in a developing country (India) context using stochastic frontier models Formulation and estimation of time vertices in a developed country (United States) context using stochastic frontier models Comparison of space-time vertices between workers and non-workers Comparison of space-time vertices between males and females Comparison of space-time vertices between developed and developing countries 1.5 Organization of the Thesis The remainder of this work is organized as follows. The literature on role of space and time in activity based travel demand modeling and stochastic frontier modeling has been reviewed in Chapter 2. The stochastic frontier modeling methodology is described in Chapter 3. Chapter 4 describes the Thane and National Household Travel Survey samples and data preparation process and provides descriptive statistics of the sample characteristics. Chapter 5 provides results of the stochastic frontier model estimation effort for the Thane Survey sample and provides discussion on various comparisons accomplished; males vs. females and workers vs. non-workers. Origin and Terminal Vertex has been estimated for all market segments. Chapter 6 provides results for National Household Survey Sample and provides all the comparisons performed on NHTS sample. Chapter 7 provides a discussion on among various market segments in developing country and developed country context. Chapter 8 offers a concluding discussion and directions for future research in this area.

PAGE 14

CHAPTER 2 LITERATURE REVIEW The notion of time-space prim was introduced by Hgerstrand [1] in early 1970 to describe the spatio-temporal constrains in which people make activity and travel decisions. The concept of time-space prim or time geography focus on relationship interrelationship among activities in time and space and constrained imposed by those interrelationships. Since 1970s many researchers in the travel behavior arena have addressed or utilized the concept of time-space prisms for modeling activity and travel engagement patterns of individuals. Fuji et al. (1997) [2] presented a micro-simulation model system of an individuals daily activity behavior. The model incorporated spatial, temporal and coupling constraints that restrict individuals daily activity behavior. The model system, entitled PCATS (Prism-Constrained Activity-Travel Simulator), is composed of an activity choice model, a mode and destination choice model and activity duration models. The estimation results point to the importance of treating daily activity and travel as a whole, and incorporating time-space constraints into analysis. PCATS is applied to estimate the effect of hypothetical transportation policies on an imaginary individual. The results indicate that PCATS can estimate the effect of change in work-hours or commuting time on individuals' activity and travel behavior. Pendyala et al. (1997) [3] developed and applied an activity-based microsimulation model system capable of simulating changes in individual travel patterns in response to a transportation control measure. Using simulator AMOS in Washington DC area, to analyze the response from activity and travel pattern survey and response to various transportation control measures, it was found that it can be used as an effective transportation policy analysis tool. Kitamura et al. (2000) [4] developed a simulator which assumes a sequential history and time-of-day dependent structure to generate daily activity travel patterns using Monte Carlo simulation. Study results show that individuals' daily travel patterns can be synthesized in a practical manner by micro-simulation.

PAGE 15

As the microsimulation approaches attempts to simulate activity and travel patterns at the level of individual travelers, when dealing with individual travelers and their behavioral responses to various transportation policies its is imperative that a mechanism be developed that facilitates modeling of an individuals time space prisms. The concept of stochastic frontier modeling has been widely used in industry where the goal is to minimize the cost and maximize the production. Models for cost are estimated as cost frontier models while for production models are formed at production frontier models. The goal is to perceive the minimum cost possible by observing the actual cost and set of explanatory variables on which the observed cost might be dependent upon. In the same manner, maximum production is perceived by observing actual production and attributes affecting the production function. Kitamura et al. (2000) [5] presented a methodology to estimate the location and size of space-time prisms that govern individuals' activity and travel. Stochastic frontier models were formulated to locate prism vertices along the time axis because the vertices of a prism are unobservable. The observable trip starting or ending times is used as the dependent variable, and commute characteristics, personal and household attributes and area characteristics are used as explanatory variables. Aigner et al. (1977) [6] provides a discussion on formulation and estimation of stochastic frontier production function models. The disturbance term is defined as the sum of symmetric normal and negative half normal. Various aspects of maximum-likelihood estimation for the coefficients of production function with respect to above disturbance are also considered. Pendyala et al. (2002) [7] estimated the temporal vertices of time-space prisms using stochastic frontier modeling technique for two activity data sets collected in San Francisco and Miami areas. Differences and similarities in temporal vertex location of various markets segments are discussed in the paper. Yamamoto et al. (2003) [8] explains that out-of-home activities are engaged within the time-space prisms but the prisms themselves are unobservable. The paper estimated stochastic frontier models for Southeast Florida data evaluates two possible distributions for frontier models for one-sided disturbance term, the half normal and exponential distributions. The results suggest that the exponential distribution has a higher goodness-offit, but the coefficient estimates of explanatory

PAGE 16

variables are similar between models with the two distributions. The work also estimated and compared origin and terminal prism vertices for working and non-working days. Recently there have been studies that recognizes role of time space in trip chaining. Some of them are discussed below. Kondo and Kitamura (2003) [9] discuss formation of trip chaining in time-space constraints. The study examines factors related to the likelihood of combining activities into a multi-sojourn trip chain. The analysis indicates that the likelihood depends on whether marginal benefit of the time spent for in-home activities diminishes or not. If marginal benefit diminishes, longer activity duration, larger distance between the home and non-work activity locations, larger distance between the home and work location, and slower speed of the travel mode used, tend to favor a multi-sojourn trip chain. When the marginal benefit does not diminish, the relation is in general reversed. It is also found that trip chaining behavior is less sensitive to the sojourn duration and travel mode in the evening period than in the morning. A study by Nishii and Kondo (2002) [10] analyzed temporal and spatial constraints underlying rail commuters' trip linkages, and examined the role of the terminal station where a commuter transfers lines or leaves. The results provide strong evidence that non-work stops in the after-work paths tend to cluster around the commuting terminal as well as the work place zone.

PAGE 17

CHAPTER 3 MODELING METHODOLOGY This section provides a brief overview of the stochastic frontier modeling methodology. By definition, a trip in a prism always starts at or after the origin vertex of the prism, and ends at or before its terminal vertex. While the beginning and ending times of a trip are almost always available from travel survey data, the origin and terminal vertices of a prism are normally unobserved. Although sometimes it is possible to infer the location of a prism vertex along a time axis, this is rather an exception than a norm. A modeling approach, therefore, is adopted in this study to estimate the location of prism vertices using unobserved variables. Adopted in the modeling approach are the inequalities, At origin vertex: oot (1) At terminal vertex: ttt (2) where o is the location along a time axis of the origin vertex of a prism, t is the location of the terminal vertex, is the beginning time of the trip, and is the end time of the trip. It is assumed that ot tt o and t are unobserved. From the inequalities, ooout (3) tttut (4) where and are the nonnegative random variables. ou tu The general form of the stochastic frontier model [6], which applies to relationships such as those presented above in Equations (3) and (4), may be written as:

PAGE 18

iiiiiiuvXXY (5) where i denotes the observation, is the observed dependent variable (in this case a trip beginning or ending time), iY a vector of coefficients, a vector of explanatory variables, and the random error terms, iX iv iu iv and .0iu iivX can be viewed as the location of origin vertex with random element Similarly, a model for a terminal vertex can be formulated as: iv iiiiiiuvXXY* (6) In econometric literature on stochastic frontier models, is typically assumed to be normal and a truncated (half) normal distribution is often used for. In this case, Aigner et al. (1977) give the distribution of iv iu i in the cost frontier model as (subscript i is suppressed below) [6, 11] ,2exp)}/(1{22)(22h (7) and the distribution of in the production frontier model as *i *22***,2exp)}/({22)(h (8) where ),,0(~,1,1,/,22222222222vvuvuvuNv and u has the density function, .0,2exp22)(22uuuguu (9) The likelihood function is said to be not entirely well behaved for models with this error density function. Waldman (1982) (11) provides the result that if the third moment of the model residuals is positive, then the least squares slope estimates and represent a local maximum of the likelihood.[6] 0

PAGE 19

This formulation is adopted in this study with an observed trip starting or ending time as ; and selected attributes of the individual and household, and workers (or non-workers) characteristics as. Because of the way the model is constructed, the inequalities of equation (1) and (2) are always satisfied. Yet, there remains the question of whether iY iX iivX in fact represents the prism constraint in the strict sense of Hgerstrand. One could argue that iivX may represent a threshold which an individual subjectively holds as the earliest possible starting time or the latest possible ending time of a trip, but may not coincide with actual constraints that are governing. For example, a worker may believe that he or she cannot possibly leave home before 7 AM in the morning; thus the origin vertex of his or her prism before the work starting time is located, at least in his or her mind, at 7 AM. But it is not likely that this is an objectively defined constraint. In fact, the same worker may leave home before 6:30 AM for a business trip. Models of prism vertices are estimated in this study with empirical data without any information on the individuals beliefs or perceptions of prism constraints. Yet observed travel behavior is governed by subjective beliefs and perceptions, e.g. I must return home by midnight or I cannot possibly leave home before 7:00 A.M. Thus some ambiguity is unavoidable about the nature of iivX ; it is unlikely that it represents a prism vertex in the strict sense of Hgerstrand. It is yet reasonable to assume that iivX is nonetheless a useful measure for the practical purpose of determining the earliest possible departure time or latest possible arrival time of a trip.

PAGE 20

CHAPTER 4 DATA PREPARATION AND SAMPLE CHARACTERISTICS This chapter provides a brief overview of data preparation process and discusses socio-economic and demographic characteristics of the Thane, India and National Household Travel Survey (NHTS), United States data. 4.1 Thane Household Travel Survey A comprehensive household travel survey was conducted in 2001 in the metropolitan area of Thane located near Bombay (Mumbai) in the state of Maharashtra, India. This city is located on the west coast of India and is a major metropolitan center of economic and business activities. The survey included a travel diary instrument that was filled out by field workers who visited households and conducted face-to-face interviews. In a developing country context such as India, it is quite common to conduct face-to-face interviews involving a large number of field workers because of the poor telecommunications system, low literacy rates, and desire to obtain high response rates. Despite face-to-face interview method, there may be number of irregularities in the Thane data. Data from developing countries are often of suspicious nature, probably due to complexities involved in conducting such surveys in developing country contexts. For example, the interviewer may prefer literate individuals over illiterate respondents making sample biased. This irregularity is also observed in large number of invalid responses in data set. But for the analysis purpose, the sample is considered free from any biasness and assumed that respondents were selected randomly and their correct responses were reported. The Thane survey data was available in Microsoft Excel format (filename.xls) and needed extensive cleaning and processing. The data was converted into a DBF file and imported into SPSS software. Using MATLAB and C++ programs missing repetitive data (example: for a particular household only first row contained information about household characteristics while it

PAGE 21

was applicable to all the trip rows relevant to that particular household) was filled. Also missing and non applicable values were replaced by -1 and -9 respectively. The original data file contained information at trip level, but the analysis required person level file (one record/row per person) and hence the trip file was aggregated at person level, requiring most of the trip variables to be removed. Some of the trip information like first trip of the day starting time and final trip of the day ending time, number of trips per day, and trip distance were retained due to their importance in analysis. The person level file was stratified by worker and gender status resulting in six segments: workers, non-workers, male workers, female workers, male non-workers, and female non-workers. Each file was saved in TAB DELIMITED format for input in LIMDEP for stochastic frontier modeling. 4.2 Household and Person Socio-economic Characteristics of Thane Survey Household and person socio-economic characteristics are described in Tables 4.1 and 4.2. The survey sample includes 14,428 individuals residing in 3,505 households. The average household size is a little over four persons per household. This is substantially higher than average household sizes found in the developed world where average household sizes are typically in the 2.5 persons per household range. About two-thirds of the households in the Thane Survey sample have four or more persons in the household. As expected, vehicle ownership level is very low with 80 percent of the households having no two-wheelers and 95 percent of the households not having any car. A high percentage of the households owned their own home, but it should be noted that a large majority of the homes are very small dwelling units. About 90 percent of the households are less than 750 sq ft in built-up area. More than 50 percent of the households have at least one child and only about 10 percent of the households have no worker. Among the 14,428 persons in the survey sample, 11,256 are adults (18 years or older). This work focuses on analyzing time space prism constraints for adults; therefore, the descriptive statistics are provided in Table 2 for the adult person sample stratified by gender and worker status. The Workers were defined by occupational category. Only those persons were included in workers category who made at least one trip on the given survey day (mobile sample), has a permanent work place and falls into service or business/professional occupation categories. Consistent with contemporary Indian society, a large proportion of the adult workers are male (85 percent). A majority of the non-workers are females. Although the average age of the different groups is

PAGE 22

Table 4.1 Household Characteristics of 2001 Thane, India Survey Sample Characteristic Statistic Sample Size 3505 Household Size 4.12 1 person 2.0% 2 persons 12.2% 3 persons 19.8% 4 persons 66.0% Two-wheel vehicles (scooter + motorcycle) 0.22 0 two-wheeler 80.2% 1 two-wheeler 17.9% 2 or more two-wheeler 2.9% Auto Ownership 0.06 0 auto 94.7% 1 auto 4.9% 2 or more autos 0.4% Dwelling Unit Own 79.3% Rent 17.6% Govt. Quarters 3.0% Company provided houses 2.0% Built-up Home Area < 250 sq ft 27.9% 250-500 sq ft 51.0% 501-750 sq ft 15.9% 751-1000 sq ft 2.4% >1000 sq ft 9.0% No. of Children (under 18) 0.90 0 children 47.4% 1 child 26.4% 2 children 16.9% 3+ children 9.3% No. of Workers 1.34 0 workers 9.3% 1 worker 57.3% 2 workers 25.8% 3+ workers 7.6%

PAGE 23

Table 4.2 Person Characteristics of 2001 Thane, India Survey Sample Worker (Adult) Non-worker (Adult) Characteristic Male Female Male Female Sample Size 3949 674 2215 4418 Average Age (in years) 36.5 35.2 36.9 37.1 18-21 years 6.0% 6.4% 31.6% 14.6% 22-49 years 77.4% 84.0% 36.1% 65.3% 50-64 years 15.8% 9.3% 19.8% 14.3% 65 years or above 0.8% 0.3% 12.5% 5.8% Occupation Service 69.6% 78.9% 4.8% 0.5% Farmer/Laborer 7.0% 6.5% 1.7% 0.3% Business/Professional 22.9% 13.1% 10.8% 0.9% Student 0.2% 0.5% 25.0% 7.4% Homemaker 0.0% 0.0% 0.0% 81.4% Retired/Unemployed 0.3% 1.0% 57.7% 9.5% Highest Education Level No School 4.5% 8.6% 5.5% 16.3% Upto 10 th grade 57.1% 31.3% 53.9% 59.2% 10 th to 12 th grade 12.7% 9.5% 18.4% 10.6% Any college 25.7% 50.6% 22.2% 13.9 Monthly Income (Rupees) No income 0.0% 0.0% 70.0% 97.4% Upto 2,000 13.9% 18.0% 6.8% 0.9% 2,001-5,000 43.4% 39.6% 12.8% 0.9% 5,001-10,000 27.7% 31.2% 3.7% 0.3% 10,001-15,000 10.3% 9.9% 6.0% 0.5% 15,001-20,000 1.3% 1.0% 0.5% 0.0% 20,001 and above 3.4% 0.3% 0.2% 0.0% License No license 80.4% 91.7% 92.2% 99.0% Two-wheeler 12.8% 5.8% 4.8% 0.7% Auto 6.8% 2.5% 3.0% 0.3% Vehicle Availability No vehicle 75.9% 93.6% 89.1% 98.8% Car/Jeep/Van 3.5% 0.9% 1.4% 0.2% Two-wheeler 11.5% 4.6% 3.9% 0.6%

PAGE 24

Table 4.3 Personal Transportation Characteristics of 2001 Thane, India Survey Sample Worker (Adult) Non-worker (Adult) Characteristic Male Female Male Female Sample Size 3949 674 2215 4418 Transit Pass 29.0% 33.4% 8.0% 2.5% Monthly Expense on Travel Rs. 358 Rs. 360 Rs. 202 Rs. 116 Mobile Sample 3949 674 911 788 Trips per day (Mobile Sample) 2.06 2.05 2.01 2.02 Work 0.83 0.91 0.00 0.00 Education 0.00 0.00 0.56 0.38 Business 0.20 0.11 0.01 0.01 Shopping 0.00 0.00 0.11 0.39 Social visits 0.00 0.00 0.13 0.09 Recreational 0.00 0.00 0.07 0.02 Medical 0.00 0.00 0.03 0.03 Return home 1.02 1.01 1.00 1.00 Other 0.00 0.00 0.11 0.09 Modal Split (all trips) Non-motorized 48% 51% 53% 60% Automobile 6% 3% 4% 1% Public Transit 40% 38% 35% 28% IPT 6% 8% 8% 11% Daily Activity Duration (min) Work 457 426 0 0 Education 0 0 197 121 Business 96 42 5 5 Shopping 0 0 23 59 Social visits 0 2 34 19 Recreational 0 0 12 5 Medical 0 0 5 8 Temporary Home Sojourn 10 11 8 7 Other 0 0 34 22 Daily Travel Duration (min) 72 69 58 45 Work 32 31 0 0 Education 0 0 17 11 Business 4 3 0 0 Shopping 0 0 2 6 Social visits 0 0 4 2 Recreational 0 0 2 0 Medical 0 0 2 1 Return home 36 35 28 23 Other 0 0 3 2

PAGE 25

quite similar, the distributions are quite different. As expected, a large proportion of workers (about 80 percent) fall in the 22-49 year range. On the other hand, male non-workers are in the 18-21 year age group (31 percent) or retired (13 percent). The former group is likely to consist mostly of college students. Only 55 percent of male non-workers are in the 22-64 age group. On the other hand, 80 percent of female non-workers are in the 22-64 age range. Occupational distribution shows that a vast majority of workers are service workers, possibly working in a number of service establishments. The next highest category of workers is business/professional followed by farmer/laborer. Non-workers show considerable differences between males and females. More than 50 percent of male non-workers are retired or unemployed. More than 80 percent of female non-workers are homemakers. No male non-worker indicates his occupation as homemaker indicating the presence of a rather strong gender role in the household. About 25 percent of male non-workers are students; only 7 percent of female non-workers are students. With respect to education, it is found that a majority of respondents have not been to college at all, except for female workers. This is consistent with expectations; if females are working in the Indian context, it is generally because they are well-educated and are putting the education to good use. The monthly income in rupees (where 1 US$ = 48 Rupees) is reported at the person level. Most non-workers report no income. Some male non-workers report income, possibly due to retirement income. Most workers report incomes in the Rs. 2001 Rs. 10000 range on a monthly basis. Thus, compared to the developed country context, the standard of living is quite low. With respect to license holding and vehicle availability at the person level, it is generally found that vehicle availability is very low. A very high percentage of respondents in all groups report that they have no drivers license and no vehicle. This leads to a discussion on the personal transportation characteristics of the survey respondents. While all workers reported at least one trip (mobile people), only 911 of the 2215 male non-workers (41 percent) and 788 of the 4418 (18 percent) female non-workers reported at least one trip on the travel survey day. As expected, workers spend more money every month on transportation expenses. About 30 percent of workers have a transit pass; less than 10 percent of non-workers do so. Travel characteristics are summarized in Table 4.3 for the mobile sample, i.e., individuals who reported at least trip on the survey day. The modal split across all trips shows that about 90 percent of trips are accomplished either by non-motorized modes or by public transit. Only about 1-5 percent of the trips are made by

PAGE 26

automobile. About 6-10 percent of the trips are made by Intermediate Public Transportation (IPT). Average daily activity durations generally show that workers mostly spend their time outside home at the work location. About 7-8 hours of the day is devoted to the work activity with some additional time devoted to business activities, most likely related to work. The temporary home sojourn refers to an individual returning home in the middle of the day for a brief period and then departing again to pursue activities outside home. Some individuals return home in the middle of the day, possibly to eat lunch or run errands. Non-workers, on the other hand, pursue a range of business, shopping, social, recreational, and other activities in addition to school/education. With respect to daily travel durations, workers spend virtually all of their time traveling to and from work (return home). There is a small amount of travel for business purposes. Non-workers, on the other hand, spend time traveling to a variety of activities consistent with their activity participation and duration patterns. In general, it appears that workers spend a total of about 70 minutes traveling per day. The corresponding figure for non-workers is 58 minutes for males and 45 minutes for females. It should be noted that these figures are averages taken over the mobile sample only. If the zero-travel individuals are included in the sample, the averages would drop dramatically. 4.3 National Household Travel Survey The National Household Travel Survey (NHTS) is a U.S. Department of Transportation (DOT) effort sponsored by the Bureau of Transportation Statistics (BTS) and the Federal Highway Administration (FHWA) to collect data on both long-distance and local travel by the American public. The joint survey gathers trip-related data such as mode of transportation, duration, distance and purpose of trip. It also gathers demographic, geographic, and economic data for analysis purposes. The NHTS 2001 was conducted using Computer-Assisted Telephone Interviewing (CATI) technology. Each household in the sample was assigned a specific 24-hour Travel Day and kept diaries to record all travel by all household members for the assigned day. The NHTS interviews were conducted from April 2001 through May 2002. There are approximately a total of 66,000 households in the final 2001 NHTS dataset. About 26,000 households are in the national sample, while the remaining 40,000 households are from nine add-on (area-specific) areas.

PAGE 27

The NHTS 2001 data is available for download at http://nhts.ornl.gov/2001/index.shtml The NHTS data is updated frequently and currently Jan. 2004 version of NHTS 2001 is available for download. This work is based on 2003 version of NHTS 2001 data. The data is available in DBF format in four files (i) Household File (ii) Person File (iii) Trip File and (iv) Vehicle File. This work required first three files only. These files were imported in SPSS and Household and Person file were merged in Trip file using Add variables option in SPSS. This created a file containing the entire trip, and personal and household information pertaining to trip in one row only. This file was aggregated at personal level. Since each individual was considered equally important in the analysis, unweighted analysis was performed for the NHTS sample. A host of dummy variables based on socio-economic and demographic characteristics were added in person level file. The person level file was stratified by worker and gender status resulting in six segments: workers, non-workers, male workers, female workers, male non-workers, and female non-workers. Each file was saved in TAB DELIMITED format for input in LIMDEP for stochastic frontier modeling. 4.4 Household and Person Socio-economic Characteristics of NHTS 2001 Household and person socio-economic characteristics are described in Tables 4.4 and 4.5. The survey sample includes 60,282 individuals residing in 26,308 households. The average household size is close to two and half persons per household. This is substantially lower than average household size found in the developing country like India where average household size is in more then four persons per household range. Significantly, about one-fifth households of NHTS sample are one-person households, deeply reflecting social structure and living style of the country. Perhaps most of them are retirees or students. As expected, almost everyone has a vehicle and one fourth of the households have three or more vehicles. A high percentage of the households (79 percent) own their own home, and almost three-forth of the households are living in detached single houses. In accordance to the small household sizes, a whopping 65 percent of the households do not have any children. About 23 percent households do not have even a single worker. This refers to a large number of retired people living with small household sizes. Among the 60,282 persons in the survey sample, 45,704 are adults (18 years or older). To facilitate comparison between Thane Survey and NHTS sample only urban households from NHTS sample were taken into consideration for analysis because Thane is also a suburb area. Only weekdays trips from NHTS sample were used in analysis to be consistent with Thane

PAGE 28

Table 4.4 Household Characteristics of NHTS 2001 Sample Characteristic Statistic Sample Size 26038 Household Size 2.56 1 person 21.6% 2 persons 38.6% 3 persons 16.0% 4 persons 23.8% Vehicle Ownership 2.05 0 auto 5.5% 1 auto 27.1% 2 autos 40.7% 3 autos 16.8% 4 and more autos 9.9% Home Ownership Type Own 78.8% Rent 20.8% Provided by Job/military 0.4% Others 0.1% Type of Housing Unit Detached Single House 72.3% Duplex 3.6% Townhouse 3.1% Apartment, Condominium 14.7% Mobile home trailer 6.1% Others 0.2% No. of Children (under 18) 0.65 0 children 65.5% 1 child 14.1% 2 children 13.5% 3+ children 6.9% No. of workers in HH 1.32 0 workers 23% 1 worker 32.6% 2 workers 36.3% 3+ workers 8.1%

PAGE 29

Table 4.5 Personal Characteristics of NHTS 2001 Sample Worker (Adult) Non-worker (Adult) Characteristic Male Female Male Female Sample Size 8215 7677 2364 4221 Mobile Sample 8215 7677 2364 4221 Average Age (in years) 42.34 42.2 63.43 57.24 18-21 years 5.1% 5.2% 5.1% 3.2% 22-49 years 64.71% 64.2% 12% 31.2% 50-64 years 25.36% 25.9% 22.1% 22.1% 65 years or above 4.83% 4.7% 60.8% 43.5% Occupation Sale or Services 24.2% 26.7% 0% 0% Clerical or Administrative 3.5% 21.3% 0% 0% Manufacture/Farming 25.6% 5.4% 0% 0% Professional/Technical 42.1% 40.4% 0% 0% Others 4.6% 6.2% 100% 100% Highest Education Level Less then High School 7.2% 4.5% 15.3% 12.8% High School Graduate 25.8% 26.3% 31.4% 35.3% College/Associate Degree 49% 53.14% 39.25% 42.5% Professional Degree 17.8% 15.6% 13.3% 8.5% Driver Status Driver 97.4% 95.9% 91.9% 85% Not a Driver 2.6% 4.1% 8.0% 15% Total HH Annual Income <= $20000 9.02% 9.02% 9.02% 9.02% $20001 $50000 31.66% 31.66% 31.66% 31.66% $50001 $80000 26.64% 26.64% 26.64% 26.64% > = $80000 27.56% 27.56% 27.56% 27.56% Missing/Others 5.12% 5.12% 5.12% 5.12% Trips per day (Mobile Sample) 4.84 5.15 4.99 5.08 Total trip distance in miles 60.43 43.72 38.03 34.06 Transit Users 4.8% 5.5% 4.1% 4.5%

PAGE 30

sample. Though this work compares whole United States personal travel with Thane, India, it will be better if the comparisons are made at city level or over geographic area where socio-economic factors and travel characteristics are rather trended or uniform. The NHTS data included only those respondents who reported at least one trip (because the data preparation started with merging trip file with household and personal files, non-trip makers were simply omitted) and hence the sample size in various segments is also mobile sample size. It can be observed that workers are almost equally distributed by gender status reflecting equal roles for males and females. This is very different from Indian scenario where only 15 percent of the workers were females. In non-workers females outnumber males with ratio of 2:1. The average age is also very different among various segments. The average age for non-workers is about 60 years in comparison to 42 years for workers. Most of the workers are in 22-49 years age range while non-workers are in 65+ years age range. The difference between average age of workers in United States (41 years) and India (35 years) is also substantial, probably attributable to life expectancy and work culture. It is also evident from age distribution and average age of non-workers that most of them are retired people. Less then 5 percent workers are in 65+ range and, in all segments, about 5 percent people are in 18-21 years range. Contrary to NHTS sample, the Thane Survey sample had a large number of female households as non-workers and a large proportion of young non-workers were students. Occupational distribution of workers shows that about two-fifth of both males and females work in Professional or Technical areas while one-forth of them are in service sectors. But there is clear gender distinction in clerical/administrative and manufacturing/farming occupations. About 21 percent of females are in clerical/administrative jobs (compare it with 3 percent males in clerical/administrative jobs) while about 26 percent of males are in manufacturing or farming related occupation (compare it with female share of 5 percent). Details about non-workers occupational distributions are not available but majority of them are retired and housemakers and fall in others categories. Combining the age distribution of non-workers with occupational distribution suggest greater role of females in housemaking as only 43 percent of females are above 65+ (possible retirees) compared to 61 percent male non-workers. With respect to education, it is found that a majority of respondents are either high school graduate or have been to college or vocational school (associate or diploma degrees). In a developed country like United States this is consistent with literacy rate and societys emphasis on education and is in sharp contrast with Indian/developing country context where illiteracy prevails and higher education, to a large population, is either unavailable or not affordable. The annual income in

PAGE 31

NHTS data is reported at household level excep t for few cases where it is reported at personal level. Only 10 percent of the households have income less then $20,000. Taking into account average household size of 2.5, this is close to poverty line and hence roughly 10 percent people are below poverty line (2001 census-12.7 percent) a nd hence the standard of living is quite good and again this is very different from a developing country like India where more then 40 percent people live under poverty line. With respect to driver status almost all the wo rkers are drivers and while 92 percent of male nonworkers and 85 percent of female non-workers are dr ivers. Since the vehicle availability is very high (95 percent) high percentage of drivers is easily understood. High vehicle availability and drivers also explains personal transportation characteristics of respondents. In United States, be it a worker or non-worker, everyone makes on average 5 trips a day. This is very high in comparison to number of trips people make in de veloping countries. In India a large proportion of non-workers did not report even a single trip on give day while workers reported on average 2.06 trips per person per day. Only 5 percent of the respondents used transit on a given day for travel, consistent with high auto use in United States and substantially different fro m developing country where more then half of the people use transit (more than 30% have trans it passes) for travel purpose. Also total distance traveled distribution shows that workers travel 44-60 miles a day and non-workers travel little less (34-38 miles) that is also very high compared to Thane where due to poor infrastructure, low vehicle availability and other constraints traveled distance is less. Though information about the total traveled distance in a day is not available in the Thane data, the total travel time in a day (60-70 min) gives indication of lesser total travel distance as the average speed in Indian context is only about 40-50km/hr. Detailed information about pers onal and socio-economic characteristics of NHTS 2001 data is available at http://nhts.ornl.gov and Bureau of Transportation Statistics (BTS) website http://www.bts.gov/programs/nationalhouseholdtravelsurvey

PAGE 32

CHAPTER 5 MODEL ESTIMATION RESULTS: THANE DATA The variables used in the model estimation effort for Thane data are shown in Table 4. Many of the variables represent dummy variables that take a value of one if the condition is satisfied and zero otherwise. Some variables capture multiple attributes, for example, VEH_HHSIZE =4 captures the effect of both household size and vehicle ownership as the variable takes a value of one if the household has a vehicle and is of size four. One can expect that vehicle ownership will have differential constraining effects depending on the number of household members. Model estimation was accomplished using the LIMDEP software package [12]. Table 5.1 Description of Explanatory Variables Used in Thane Models Variable Definition SERVICE Persons occupation is service PROFESSIONAL Persons occupation is business/professional STUDENT Person is a student UNEMPLOYED Person with no income LOW_INCOME Person with income 0 to Rs. 5000 MID_INC_HHSZ3 Person with household size 3 and income range Rs. 5000-15000 MID_INC_HHSZ4 Person with household size 4 and income range Rs. 5000-15000 MULTIACT Person performs more than two out of home activities HHSIZE=1 One person household HHSIZE=4 Household size = 4 HHSIZE>4 Household size is greater than 4 HH_CHILD At least one member of the household is below 18 years CHL_HHSZ>4 Household of size greater than 4 has at least one child less than 18 years VEH_OWNER Person owns a vehicle LIC_DRIVER Person is a licensed driver MUL_VEH_HH Person in a household with more than one vehicle VEH_HHSIZE=4 Person in a household of size 4 owning at least one vehicle LONGTRIP Total travel duration is more than one hour Tables 5.2 and 5.3 summarize the results of the model estimation effort. Origin and terminal vertex models are estimated for the following groups:

PAGE 33

iviui The dependent variables of the models presented in this work are defined with the time of day expressed in minutes, with 12:00 AM (midnight) being 0; so 6:00 AM is expressed in the model as 360, and 6:00 PM as 1080. All models assume that has a normal distribution and has a half-normal distribution. The expected value of uis evaluated as Female workers and female non-workers The models for the origin vertex of the morning prism are presented in Table 5.2. The models are formulated as cost frontier models, i.e., Y i = All workers and all non-workers Male workers and male non-workers uiuE2][2/1u (10) where u is an estimate of iiiuvX In general, all of the models offer plausible interpretations with respect to the various explanatory variables. For example, in all models, it is found that participating in multiple activities shifts the prism vertex location earlier along the time axis as evidenced by the negative coefficient on the MULIACT variable. The constant term generally indicates that non-workers have a prism vertex location shifted later in the day than workers. This is presumably due to work schedule requirements that make workers more willing to depart earlier from home than non-workers who do not have such constraints. Owning multiple vehicles appears to contribute to the vertex location shifting later in the morning as evidenced by the positive coefficient associated with the MUL_VEH_HH variable. This is also consistent with expectations as persons with greater vehicle availability may have greater flexibility to leave later from home as opposed to being influenced by the constraints and uncertain travel times of public transit or other slower modes. It is found that low income workers (all worker models) have an earlier origin vertex location. Lower income persons are likely to have less access to automobiles and may be in service oriented occupations that require earlier arrival times. On the other hand, workers in the business/professional occupation are found to have a later origin vertex. Also, non-workers whose daily travel duration exceeds one hour have earlier origin vertex location; if people have long travel times they are likely to stretch their prism to accommodate their travel patterns. On the other hand, it is also possible that those with less constraining vertex locations are able to undertake longer travel. The casual relationship between these two aspects of behavior merits further study. The presence of children in a

PAGE 34

Table 5.2 Stochastic Frontier Models of Origin Vertex Location: Thane Sample Variable Workers Non-Workers Male Workers Male Non-workers Female Workers Female Non-workers Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Constant 416.91 42.73 449.59 28.51 411.69 36.73 494.89 22.28 429.78 28.64 650.85 3.04 SERVICE -13.56 -1.78 -16.42 -1.88 -17.38 -1.70 PROFESSIONAL 37.12 4.54 37.15 4.02 UNEMPLOYED -121.06 -7.40 LOW_INCOME -15.31 -3.60 38.71 2.72 -10.66 -2.24 MID_INC_HHSZ3 51.72 2.04 MID_INC_HHSZ4 -69.15 -5.63 68.52 3.01 MULTIACT -32.75 -2.94 -71.85 -1.97 -39.23 -3.28 -106.17 -1.53 HHSIZE=1 112.00 1.91 HHSIZE=4 27.25 2.76 HHSIZE>4 12.08 2.27 -104.98 -4.38 HH_CHILD -11.96 -3.03 -15.05 -1.32 -12.87 -3.01 -40.83 -2.37 CHL_HHSZ>4 -17.94 -1.84 LIC_DRIV ER .31.36 -10 6 MUL_VEH_HH 8.15 1.60 12.08 2.27 LONGTRIP -85.94 -5.96 L(C) -29103.01 -11768.49 -24986.17 -6254.94 -4190.86 -5511.50 L() -29029.69 -11721.26 -24913.88 -6257.72 -4184.83 -5507.33 2 (df) 146.4 (7) 94.46 (5) 144.24 (7) 5.56 (4) 12.06 (3) 8.34 (3) Var(v) 12216.20 17416.64 12556.21 30022.77 8914.97 62050.89 E(u) 96.18 88.30 95.48 207.87 100.16 109.56 Var(u) 5280.49 12246.23 14321.17 24664.48 5728.24 6851.29 N 4623 1699 3949 911 674 788

PAGE 35

iiiuvX The models for the terminal vertex of the evening prism are presented in Table 5.3. In this case, the models are production frontier models and formulated as household makes the location of prism vertex earlier for all segments irrespective of gender ad work status. As the models were estimated on rather controlled homogeneous socio-demographic groups (classified by worker and gender status), it is found that several variables in the various models are not statistically significant. However, some of these variables were retained in the interest of model sensitivity and because the variables offered plausible interpretations consistent with expectations. The vertex location in this model represents the latest possible time that people may perceive for their final arrival at home. The constant term in the models is around 1260, which corresponds to 9 PM. Consistent with expectations, it is found that people who pursue multiple activities have a terminal vertex location shifted later in the day. This is presumably due to their higher activity engagement level and the need to spend more time outside home completing their activity agenda. In households with children, it is found that females (both workers and non-workers) have a tendency to have an earlier terminal vertex (as evidenced by the negative coefficients). This is indicative of a gender role where the females are caregivers for the children. In addition, it is found that male non-workers also show a negative coefficient for this variable. In households with children, a male non-worker may also be playing a child caregiver role contributing to an earlier terminal vertex location. Low income workers show a tendency to have an earlier terminal vertex location while low income non-workers show a tendency to have a later terminal vertex location. This finding is not easily explained, but is likely to be indicative of low income non-workers having to run household errands later in the evening after the worker has arrived home. The presence of multiple vehicles in the household contributes to earlier terminal vertex locations presumably due to the faster travel time in which people with access to vehicles can reach their home. Service and professional occupations are associated with later terminal vertices. Also, those with long trips have later terminal vertex locations; if people have long travel times, they are likely to stretch their prism to accommodate their travel patterns. Students are generally found to have earlier terminal vertices as evidenced by the negative coefficient associated with the STUDENT variable. Once again, it should be noted that some coefficients that are not statistically significant have been retained in the models for model sensitivity and because the coefficients offered plausible behavioral interpretation.

PAGE 36

Table 5.3 Stochastic Frontier Models of Terminal Vertex Location: Thane Sample Variable Workers Non-Workers Male Workers Male Non-workers Female Workers Female NonCoef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Constant 1256.1 125.12 1265.3 130.50 1384.02 154.11 1286.34 88.54 1198.47 76.66 1249.28 91.76 SERVICE 11.90 5.91 -35.85 -4.42 31.47 2.43 STUDENT -99.56 -10.20 -99.59 -6.94 -94.10 -7.45 LOW_INCOME -13.78 -2.40 68.71 1.56 MULTIACT 75.34 4.42 69.14 2.00 77.70 3.33 103.17 2.06 140.78 3.87 HHSIZE=1 85.83 1.88 HHSIZE=4 44.05 2.76 HH_CHILD -28.05 -2.99 -29.29 -4.21 -35.90 -2.54 -18.07 -1.46 CHL_HHSZ>4 -36.52 -1.76 VEHOWNER -40.90 -5.45 LIC_DRIVER -24.62 -2.94 VEH_HHSIZE=4 30.68 1.74 MUL_VEH_HH -29.57 -2.11 MID_INC_HHSZ 24.77 1.51 L(C) -31302.47 -11582.58 -26791.32 -6261.95 -4450.77 -5307.13 L( ) -31243.44 -11515.15 -26835.35 -6224.11 -4438.64 -5277.98 2 (df) 118.06 (4) 134.86 (5) 86.06 (4) 75.68 (4) 24.26 (4) 58.3 (4) Var(v) 7494.22 10173.28 9529.85 12797.40 6175.28 6308.38 E(u) 272.38 262.04 286.59 270.31 221.91 254.84 Var(u) 42281.21 39194.73 46854. 02 41706.60 28092.67 37069.64 N 4623 1699 3949 911 674 788

PAGE 37

All of the models estimated offer a powerful means of comparing expected prism vertex locations across socio-economic groups and between developed and developing contexts. The comparisons performed include: Workers vs. non-workers Males vs. females (workers and non-workers) The comparisons are best seen by plotting the distributions of actual first departure or final arrival times against the distributions of the corresponding expected vertex locations. The expected vertex location for each individual is calculated as iX These plots are shown in Figures 5.1 through 5.10. In addition to a comparison through a plot of the distributions, the value of E[u] offers another statistic for comparisons. The E[u] represents the average difference between the expected vertex location and the actual arrival/departure time. This value may be interpreted as a measure of proximity between the vertex location and the actual departure/arrival time. If the value of E[u] is small, then one might say that the travelers are generally arriving and departing very close to their prism constraints. Conversely, large values may suggest that people are departing or arriving with much time to spare. These values will be discussed in conjunction with the distributions of vertex locations and arrival/departure times.

PAGE 38

5.1 Origin Vertex Location Figure 5.1 shows comparison of first departure time distribution and origin vertex distribution for workers vs. non-workers. The comparisons are stratified by gender status and discussed in pair wise manner. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Workers Obs Workers Vertex Non Workers Obs Non Workers Vertex Figure 5.1 Distribution of Origin Vertex and First Departure Time (Thane): All Workers vs. All Non-workers

PAGE 39

Male Workers vs. Female Workers: Figure 5.2 compares males against females while controlling for working status. Both worker samples show clear departure time distributions and clear corresponding vertex location distributions. However, the female departure time distribution is shifted a little bit later than that of males and the vertex distribution shows a similar shift. The first departure time distributions for male and female workers peak at about 9:00 AM and 9:30 AM and vertex distributions peak at about 7:30 AM and 8:00 AM respectively. Thus, it appears that, between male workers and female workers, the female workers may be more involved in household and child obligations that contribute to constraints that shift the vertex location to the right. Both groups show an average difference between the expected vertex location and the observed departure time of a little over 1 hr 30 min. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Workers Obs Male Workers Vertex Female Workers Obs Female Workers Vertex Figure 5.2 Distribution of Origin Vertex and First Departure Time (Thane): Male Workers vs. Female Workers

PAGE 40

Male Non-workers vs. Female Non-workers: This comparison is shown in Figure 5.3. There are clear differences that suggest that male and female non-workers play differing roles in the household. The substantial presence of students in the male non-worker sample contributes to one of the two peaks of the origin prism vertex distribution to be well defined and occurring at about 6:00 AM. On the other hand, the substantial presence of homemakers and caregivers in the female non-worker sample contributes to a major peak in their vertex distribution being located at about 10:00 AM. The actual departure time distributions show rather similar wavy patterns; however, a noticeable difference is that a larger proportion of female non-workers leave home for the first time only at about 5:00 PM. This may suggest that female non-workers take care of the household obligations virtually all day long and then get out of the home for the first time at about 5:00 PM to either participate in recreational activities or run errands. It is found that male non-workers show the value of E[u[ as high as 3 hr 30 min but female non-workers show only 1 hr 50 min. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Non-Workers Obs Male Non-Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 5.3 Distribution of Origin Vertex and First Departure Time (Thane): Male Non-workers vs. Female Non-workers

PAGE 41

Male Workers vs. Male Non-workers: This comparison is shown in Figure 5.4. As expected, male workers show a clear departure time distribution with the peak around the 8:30 AM mark. On the other hand, male non-workers exhibit a more wavy departure time distribution with several small peaks. The earliest and most prominent peak occurs early in the morning at about 7:00 AM and this is most likely to constitute the student subsample among the male non-workers. Other male non-workers leave home for the first time later in the morning, as expected. The origin vertex distributions are very consistent with the departure time distributions. The male worker vertex distribution is a clear unimodal distribution with a clear peak in the 7:00 AM range. The male non-worker distribution is a bimodal distribution with one peak in the 6:00 AM range and another one in the 9:00 AM range. The former corresponds to the student subsample while the latter corresponds to the others. The clear differences between workers and non-workers are demonstrated and more interestingly, it appears that it might be important to distinguish between students and non-students within the non-worker category (at least in a developing country context). An examination of Table 7.1 or Table 5.2 shows that the E[u] is about 1 hr 35 min for male workers and about 3 hr 30 min for male non-workers. As expected, workers exhibit a smaller average time gap between the vertex location and the actual departure time, presumably due to work schedule constraints. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Non-Workers Obs Male Non-Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 5.4 Distribution of Origin Vertex and First Departure Time (Thane): Male Workers vs. Male Non-workers

PAGE 42

Female Workers vs. Female Non-workers: A rather similar, but more pronounced, difference between workers and non-workers can be seen for females as well (Figure 5.5). The female worker sample shows a clear distribution of departure time and a correspondingly clear vertex distribution. The female non-worker sample shows a wavy departure time distribution with peaks at about 7:00 AM (possibly the student subsample), 10:00 AM, and 5:00 PM. Correspondingly, the vertex distribution shows a large peak at about the 10:00 AM mark with a smaller peak preceding it at about 8:00 AM. Thus, the origin vertex location distribution for non-workers is clearly shifted later in the morning than that for the female worker sample. These findings are consistent with expectations and reflect the possibility that the female non-workers have household obligations and childcare responsibilities in the morning that shift their prism to later in the morning. An examination of the values of E[u] shows that both groups have an average difference between the expected vertex location and the observed departure time around 1 hr 40min. These findings are very similar to those found previously for male workers vs. male non-workers. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Female Workers Obs Female Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 5.5 Distribution of Origin Vertex and First Departure Time (Thane): Female Workers vs. Female Non-workers

PAGE 43

5.2 Terminal Vertex Location Figure 5.6 shows comparison of final arrival time distribution and terminal vertex distribution for workers vs. non-workers. The comparisons are stratified by gender status and discussed in pair wise manner. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Workers Obs Workers Vertex Non Workers Obs Non Workers Vertex Figure 5.6 Distribution of Terminal Vertex and Final Arrival Time (Thane): All Workers vs. All Non-workers

PAGE 44

Male Workers vs. Female Workers: This interesting comparison (Figure 5.7) shows that males and females show considerable differences in their prism constraints. The vertex location distribution for female workers is substantially earlier than that of male workers. Although both groups are workers, the female group shows an earlier vertex location distribution, most probably because of the household obligations and responsibilities carried by females in the Indian context. The arrival time distributions are more similar to one another than the vertex location distributions. Both arrival time distributions show peaks at about 7:00 PM. However, the peak for the male vertex location distribution is at about 10:30 PM and that for the female worker sample is at about 9:00 PM. Table 5.3 shows that the E[u] are 4 hr 45 min and 3 hr 40 min for male and female workers respectively. Thus, although male workers perceive to be less constrained than female workers, they dont appear to take advantage of the larger time-space prism. Their final home arrival time distribution is similar to that of female workers. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Workers Obs Male Workers Vertex Female Workers Obs Female Workers Vertex Figure 5.7 Distribution of Terminal Vertex and Final Arrival Time (Thane): Male Workers vs. Female Workers

PAGE 45

Male Non-workers vs. Female Non-workers: This comparison is shown in Figure 5.8 It is found that both groups have rather similar arrival time distributions with female non-workers showing a greater tendency to participate in evening activities at about 7:00 PM relative to males. Corresponding to this difference, it is found that the male non-workers have a vertex location distribution that is shifted to the left (earlier in the evening) when compared with female non-workers. The peak for the male non-worker vertex location distribution is at about 7:30 PM. The female non-worker sample shows a distribution with two peaks; a small peak at about 7:00 PM and a large peak at about 9:00 PM. The distributional differences may be due to the presence of students and retirees in the male non-worker sample in larger proportions than in the female non-worker sample. The students and retirees are both likely to return home earlier; on the other hand, the female non-worker sample includes many homemakers who possibly perform household errands in the evening (after the worker has arrived home) and thus have a later prism vertex location distribution. Table 5.3 shows that the E[u] for male non-workers is 4 hr 30 min and 4 hr 15 min for female non-workers. Thus, both groups exhibit a similar average time gap between the actual arrival time and vertex location. The differences are with respect to the locations of the distributions along the time axis. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Non-Workers Obs Male Non-Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 5.8 Distribution of Terminal Vertex and Final Arrival Time for (Thane): Male Non-workers vs. Female Non-workers

PAGE 46

Male Workers vs. Male Non-workers: Figure 5.9 shows a comparison between male workers and male non-workers. The actual arrival time distributions are flatter than the actual departure time distributions seen in Figure 5.4. The non-workers show a distribution with two small peaks, one at about 1:00 PM and another at about 7:00 PM. The workers show a rather flat distribution with a small peak at about 7:00 PM. The corresponding vertex location distributions reflect these patterns of arrival. The non-workers vertex distribution is substantially earlier than that of workers and peaks about 8:00 PM. The workers vertex distribution peaks at about 10:30 PM indicating that male workers perceive that they have considerable flexibility on when they can arrive home in the evening. This may be due to the socio-cultural norm where non-workers take care of much of the household obligations while workers are less responsible for in-home obligations. The values in Table 5.3 suggest that both male workers and non-workers arrive about 4 hr 45 min prior to their vertex location (on average). This is a rather large value (larger than that seen in the morning departure), but may reflect the greater flexibility perceived at the end of the day once the work of the day has been accomplished. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Workers Obs Male Workers Vertex Male Non-Workers Obs Male Non-Workers Vertex Figure 5.9 Distribution of Terminal Vertex and Final Arrival Time (Thane): Male Workers vs. Male Non-workers

PAGE 47

Female Workers vs. Female Non-workers: With respect to the terminal arrival of females, however, it is found that female workers and non-workers share terminal vertex location distributions that are quite similar to one another (Figure 5.10). Once again, non-workers show a bimodal arrival time distribution with peaks at about 1:00 PM and 7:00 PM while workers show a more unimodal but spread distribution with a small peak at about 7:00 PM. The terminal vertex location distribution for female non-workers is clearly bimodal with a small peak at about 7:00 PM corresponding to those who come home early. There is a major peak at the 9:00 PM mark for those who come home in the second peak of the bimodal distribution. Interestingly, the workers show a very similar vertex location distribution. The first peak about 7:00 PM is very small and the major peak for female workers occurs at about 9:00 PM as well. Thus, it appears that females (both workers and non-workers) have a certain level of household responsibility and obligations that contributes to them having similar vertex location distributions. In Table 5.3 it is found that the E[u] is about 3 hr 40 min for female workers and about 4 hr 15 min for female non-workers. This difference is consistent with expectations where one would expect workers to have to arrive closer to their vertex location (on average) due to their work schedule constraints. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Female Workers Obs Female Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 5.10 Distribution of Origin Vertex and First Departure Time (Thane): Female Workers vs. Female Non-workers

PAGE 48

CHAPTER 6 MODEL ESTIMATION REAULTS: NHTS 2001 DATA The modeling methodology adopted to estimate vertices for Thane sample data also used for NHTS 2001 data. Origin and Terminal vertex location models were estimated for various socio-economic segments and comparisons were made in segments to better understand time space vertex location in a developed country context. Table 6.1 provides a list of variables used into model estimation effort. Model estimation was accomplished using the LIMDEP software package [12]. Table 6.1 Description of Explanatory Variables Used in NHTS 2001 Models Variable Definition LDRIVER Person is a licensed driver CNT_DRV Count of drivers in household MULTIPUR Person makes 4 or more trips in a day LONGTRIP Sum of tip distances traveled in a day is more than 50 miles. DFULLTIM Person works full time DTRANUSE Person makes at least one transit trip on survey day LOWINC Total household income is less than $ 30000 MALE Persons gender is male HH_CHILD Household has at least one child (age less than 18) MVEH Household has more than one vehicles HHSIZE Household size DCARPL Person carpools for traveling HH_HISP Person is of Hispanic origin Tables 6.2 and 6.3 summarize the results of the model estimation effort for NHTS 2001 sample. Origin and terminal vertex models are estimated for the following groups: All workers and all non-workers Male workers and male non-workers Female workers and female non-workers

PAGE 49

The models for the origin vertex of the morning prism are presented in Table 6.2. Similar to Thane data set models are formulated as cost frontier models and a ll of the models offer plausible interpretations with respect to the various explanatory variables. For example, engagement in multiple activities resulted in shift in prism vertex location earlier along the time axis as evident by negative coefficient for MULTI ACT variable. Also, the prism vertex location for non-workers is shifted later in day than workers, shown by constant term, due to work schedule requirements for workers to depart earlier than non-workers from home. The prism vertex location for long distance travelers (travel distance greater than 50 miles) is also shifted earlier irrespective of workers and non-workers. Full time workers and transit users also depart from home earlier than part time workers. Perh aps transit does not offer flexibility and comforts associated with auto and uncerta inty in schedule and longer travel time force users to depart earlier from home. Since almost everyone has a ve hicle in United States, number of vehicles in household does not have significant impact on vertex location. Also, origin vertex location for poor is shifted earlier probably because most of th e transit users are poor. The vertex location for carpoolers is also shifted earlier perhaps due to co mplexities and need for more time involved in carpool managements. Again, it is found that several variables in the various models are not statistically significant. However, some of these variables were retained in the interest of model sensitivity and because the variables offered pl ausible interpretations consistent with expectations. The models for the terminal vertex of the eveni ng prism are presented in Table 6.3. In this case, the models are production frontier models. Consistent with expect ations, it is found that people who pursue multiple activities have a terminal vertex location shifted later in the day; easily explainable due to their higher activity engagement level and the need to spend more time outside home completing their activity agenda. Low income workers and non-workers show a tendency to have an earlier terminal vertex location proba bly due to their low level of engagement in multiple activities and more dependence on transit. Al so, those with long trips have later terminal vertex locations because they are likely to stre tch their prism to accommodate their travel patterns. Similar to the comparison performed for Thane Survey sample, the comparisons performed for NHTS sample include: Workers vs. non-workers Males vs. females (workers and non-workers)

PAGE 50

Table 6.2 Stochastic Frontier Models of Origin Vertex Location: NHTS 2001 Variable Workers Non-Workers Male Workers Male Non-workers Female Workers Female Non-workers Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Constant 401.03 72.24 508.85 77.96 332.42 73.92 495.68 44.25 410.30 107.38 511.51 64.53 LDRIVER 5.03 1.03 -8.21 -1.62 -14.57 -1.45 -10.70 -1.74 CNT_DRV -11.60 -4.44 MULPUR -2.97 -1.52 -67.46 -18.08 5.22 1.96 -69.21 -10.58 -10.66 -3.90 -66.82 -14.70 LONGTRIP -28.70 -13.17 -48.90 -10.09 -31.42 -10.68 -49.38 -6.12 -18.31 -5.68 -47.84 -7.91 DFULLTIM -63.96 -29.81 -56.47 -20.52 DTRANUSE -10.75 -2.29 -33.49 3.89 -17.77 -1.07 -18.46 -2.84 -40.45 -3.92 LOWINC -6.81 -3.03 -5.45 -1.53 -7.95 -2.73 -6.26 -1.41 MALE -22.30 -5.96 HH_CHILD -10.54 -2.39 -5.51 -1.70 -8.27 -1.66 MVEH -4.47 -0.71 HHSIZE -2.67 -2.83 DCARPL -15.20 -2.73 -17.78 -3.23 HH_HISP -37.94 -3.01 -22.05 -2.55 L(C) -101298.0 -42970.55 -52537.55 -15427.86 -48664.79 --27524.35 L() -100821.8 --42911.99 -52383.94 -15336.84 -48443.37 -27365.53 Var(v) 2366.03 5242.92 2032.82 5509.00 2267.63 5037.98 E(u) 187.1 199.17 197.49 197.85 180.40 199.70 Var(u) 20017.43 22673.63 22292.14 22372.41 18600.25 22792.95 N 15892 6585 8215 2364 7677 4221

PAGE 51

Table 6.3 Stochastic Frontier Models of Terminal Vertex Location (NHTS 2001) Variable Workers Non-Workers Male Workers Male Non-workers Female Workers Female Non-workers Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Constant 1344.6 162.54 1136.26 124.19 1350.21 242.72 1132.05 100.65 1333.98 129.43 1150.67 121.23 LDRIVER -19.86 -2.50 12.81 1.68 -25.20 -2.51 MULPUR 50.88 16.26 123.18 22.04 48.14 10.76 143.76 15.28 51.37 11.90 115.18 16.94 LONGTRIP 21.40 6.42 70.26 9.45 14.42 3.29 78.61 6.22 66.64 7.24 DTRANUSE 14.31 2.09 45.71 3.62 14.49 1.61 51.97 2.30 16.30 1.72 38.72 2.61 LOWINC -4.97 -1.56 -10.63 -2.05 -7.55 -1.78 -13.96 -2.00 MALE 7.06 2.59 HH_CHILD 26.22 4.35 MVEH -4.17 -1.18 -9.78 -1.99 -10.82 -1.17 20.84 2.95 HH_HISP 35.19 1.79 22.26 2.04 L(C) -108650.7 -45145.38 -56541.42 -16268.22 -52045.89 -28873.79 L() -108482.3 44805.36 -56475.25 -16125.67 -51950.92 -28675.97 Var(v) 3234.10 13899.39 2598.73 15448.60 3792.08 12998.13 E(u) 320.66 255.77 342.77 255.17 296.15 256.04 Var(u) 58766.71 37389.66 67150.84 37214.86 50127.71 37468.79 N 15892 6585 8215 2364 7677 4221

PAGE 52

Again, the comparisons are best seen by plotting the distributions of actual first departure or final arrival times against the distributions of the corresponding expected vertex locations. These plots are shown in Figures 6.1 through 6.10. In addition to a comparison through a plot of the distributions, E[u] values, the average difference between the expected vertex location and the actual arrival/departure time are also discussed in conjunction with the distributions of vertex locations and arrival/departure times. 6.1 Origin Vertex Location Figure 6.1 shows comparison of first departure time distribution and origin vertex distribution for workers vs. non-workers. The comparisons are stratified by gender status and discussed in pair wise manner. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Workers Obs Workers Vertex Non Workers Obs Non Workers Vertex Figure 6.1 Distribution of Origin Vertex and First Departure Time (NHTS): All Workers vs. All Non-workers

PAGE 53

Male Workers vs. Female Workers: Figure 6.2 compares male workers against female workers. It is evident from the figure that both male and female worker samples show a clear unimodal departure time distribution with equal degree of variance and peak around 8:00 AM and vertex locations are also consistent with departure time distribution and peak around 6:00 AM for both male and female workers. This shows that working females are not much different from working males in their perception for origin prism. The E[u] values for male workers and female workers are 3 hr 17 min and 3 hr respectively. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Male Workers Obs Male Workers Vertex Female Workers Obs Female Workers Vertex Figure 6.2 Distribution of Origin Vertex and First Departure Time (NHTS): Male Workers vs. Female Workers

PAGE 54

Male Non-workers vs. Female Non-workers: This comparison is shown in Figure 6.3. The actual departure time distributions for both male non-worker and female non-worker samples are similar peaking around 9:30 AM but the vertex distribution for male non-workers peaks around 7:00 AM while for female non-workers it peaks around 8:00 AM. There are clear differences that suggest that male and female non-workers play differing roles in the household. The substantial presence of homemakers and caregivers in the female non-workers shifts their vertex later than male non-workers. The E[u] values for male non-workers and female non-workers are 3 hr 18min and 3 hr 20 min respectively. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Male Non-Workers Obs Male Non-Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 6.3 Distribution of Origin Vertex and First Departure Time (NHTS): Male Non-workers vs. Female Non-workers

PAGE 55

Male Workers vs. Male Non-workers: This comparison is shown in Figure 6.4. As expected, male workers show a clear departure time distribution with the peak around the 8:00 AM mark. Similar to workers, non-workers also show a clear departure time distribution but it is flatter than workers distribution and peaks later in the day around 9:30 AM. The origin vertex distributions are unimodal and very consistent with the departure time distributions. The male worker sample vertex distribution peaks around 6:00 AM while for male non-worker sample it peaks around 7:00 AM. An examination of Table 6.2 shows that the E[u] is about 3 hr 17 min for male workers and about 3 hr 18 min for male non-workers. The E[u] value, the average difference between the expected vertex location and the actual departure time, explains that both workers and non workers perceive their origin prism vertex quite early. While male non-workers might not have work schedule constrains, their origin vertex is equally stretched to male workers who might need to drop kids to daycare or school and manage one or more activities before work or on the way to work. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Male Workers Obs Male Workers Vertex Male Non-Workers Obs Male Non-Workers Vertex Figure 6.4 Distribution of Origin Vertex and First Departure Time (NHTS): Male Workers vs. Male Non-workers

PAGE 56

Female Workers vs. Female Non-workers: This comparison is shown in Figure 6.5. Similar to male workers and non-workers, female workers and non-workers also show clear departure time distributions peaking around 8:00 AM and 9:30 AM respectively and their origin vertex distribution is around 6:00 AM and 8:00 AM respectively. Like male non-workers, female non-workers distribution is also more spread than female workers distributions consistent with the expectation that non-workers have less constrains. Also, female non-workers have household obligations and childcare responsibilities in the morning that shift their prism later in the morning. An examination of the values of E[u] shows that female workers have an average difference between the expected vertex location and the observed departure time around 3 hr and female non-workers have a difference of 3 hr and 20 min. These findings are very similar to those found previously for male workers vs. male non-workers. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Female Workers Obs Female Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 6.5 Distribution of Origin Vertex and First Departure Time (NHTS): Female Workers vs. Female Non-workers

PAGE 57

6.2 Terminal Vertex Location Figure 6.6 shows comparison of final arrival time distribution and terminal vertex distribution for workers vs. non-workers. The comparisons are stratified by gender status and discussed in pair wise manner. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Workers Obs Workers Vertex Non Workers Obs Non Workers Vertex Figure 6.6 Distribution of Terminal Vertex and Final Arrival Time (NHTS): All Workers vs. All Non-workers

PAGE 58

Male Workers vs. Female Workers: This comparison is shown in Figure 6.7. The actual arrival time distribution for male workers is very similar to that of female workers and both of them, though rather flat, peaks around 5:30 6:00 PM but the vertex distribution for male workers peaks around 12:00 while it peaks around 11:00 PM for female workers, most probably because of the household obligations and responsibilities carried by females. Table 6.3 shows that the E[u] values are 5 hr 43 min and 4 hr 56 min for male and female workers respectively. Thus, although male workers perceive to be less constrained than female workers, they dont appear to take advantage of the larger time-space prism. Their final home arrival time distribution is similar to that of female workers. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Male Workers Obs Male Workers Vertex Female Workers Obs Female Workers Vertex Figure 6.7 Distribution of Terminal Vertex and Final Arrival Time (NHTS): Male Workers vs. Female Workers

PAGE 59

Male Non-workers vs. Female Non-workers: This comparison is shown in Figure 6.8 The actual arrival time distributions for both male non-workers and female non-workers are similar and flat with two small peaks around 4:30 PM and 8:00 PM. The vertex distributions for both segments are also bimodal. While for male non-workers it peaks around 7:00 PM and 10:00 PM, for female non-workers it peaks around 8:00 PM and 10:00 PM. This is not easily explainable but probably some of trips by retired non-workers, and housemakers might result in multimodal distribution. The female non-worker sample might include some homemakers who possibly perform household errands in the evening after their working spouse or head of the family returns and thus have a later prism vertex location distribution than male non-workers for the first peak. Both groups show E[u] value around 4 hr 15 min. 01020304050607080901003456789101112131415161718192021222324252627Time of Day(hr)Frequency Distribution(%) Male Non-Workers Obs Male Non-Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 6.8 Distribution of Terminal Vertex and Final Arrival Time (NHTS): Male Non-workers vs. Female Non-workers

PAGE 60

Male Workers vs. Male Non-workers: Figure 6.9 shows a comparison between male workers and male non-workers. The final arrival time distributions for both the groups are flatter than the actual departure time distributions seen in Figure 6.4, the non-worker sample being more spread, consistent with the expectations that workers are constrained by work schedule. The male worker sample shows an arrival time distribution peaking around 5:30-7:00 PM range. The non-workers show a distribution with two small peaks, one at about 5:00 PM and another at about 8:00 PM. The vertex distribution for male workers is unimodal and peaks around 12:00 midnight while for male non-workers it is bimodal and peaks around 7:00 PM and 9:30 PM. This indicates that male workers perceive that they have considerable flexibility on when they can arrive home in the evening as their vertex (compared to male non-workers) is shifted too much later in the night. This can be explained by the fact that workers, unlike non-workers, run their errands (shopping, dinner etc.) only in evening after they finish their work and hence their perception is already made of to spare more time out of home in evening. The E[u] values in Table 12 suggest that male workers arrive about 5 hr 43 min prior to their vertex location (on average) and male non-workers are relatively closer to their actual arrival time with E[u] value 4 hr 15 min. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Male Workers Obs Male Workers Vertex Male Non-Workers Obs Male Non-Workers Vertex Figure 6.9 Distribution of Terminal Vertex and Final Arrival Time (NHTS): Male Workers vs. Male Non-workers

PAGE 61

Female Workers vs. Female Non-workers: This comparison is shown in Figure 6.10. Again, the actual arrival time distributions for both female workers and female non-workers are flatter than the departure time. The female worker sample distribution shows a rather spread peak in 5:30 to 9:30 PM range while the female non-worker sample distribution does not have a clearly defined peak and majority of the sample is distributed from 3:30 PM to 9:00 PM. The vertex distribution for female workers is unimodal and peaks around 11:30 PM while it is bimodal for female non-workers peaking at 8:00 PM and 10:00 PM. The bimodal vertex distribution for female non-workers can be attributed to difference in characteristics of students, retired non-workers and homemaker non-workers. The E[u] values are 4 hr 56 min and 4 hr 16 min for female workers and female non-workers respectively. 01020304050607080901003456789101112131415161718192021222324252627Time of Day (hr)Frequency Distribution(%) Female Workers Obs Female Workers Vertex Female Non-Workers Obs Female Non-Workers Vertex Figure 6.10 Distribution of Origin Vertex and First Departure Time (NHTS): Female Workers vs. Female Non-workers Most of the terminal arrival distributions have a very small proportion of trips in early morning hours (around 2:00-3:00 AM). This is probably attributable to people coming very late in night from clubs, movies and other night-life activities.

PAGE 62

CHAPTER 7 INTERNATIONAL COMPARISONS After estimating models for origin and terminal vertices for various market segments for Thane and NHTS samples, corresponding vertices are compared for both data sets. Thane, though a suburb on outskirts of Bombay, India, can be considered an ideal depiction of a developing country and United States is perfect example of a developed nation. Only urban households were selected from the NHTS 2001 data to facilitate comparison with a data from developing country suburb. The actual first departure time and terminal arrival time distributions have been plotted with origin and terminal vertex distribution for both workers and non-workers and also stratified by gender status for Thane and NHTS samples. Comparison of E[u] values (Table 7.1) also provides more insight into differences between developing and developed countries. Table 7.1: Comparison of Average Differences between Expected Vertex Location and Actual Departure / Arrival Time (E[u] values) Market Segment NHTS 2001 Thane 2001 First Departure Time Workers 3hr 07min 1hr 36min Non-workers 3hr 19min 1hr 28min Male-workers 3hr 17min 1hr 35min Female-workers 3hr 00min 1hr 40min Male Non-workers 3hr 18min 3hr 28min Female Non-workers 3hr 20min 1hr 50min Terminal Arrival Time Workers 5hr 21min 4hr 32min Non-workers 4hr 16min 4hr 22min Male Workers 5hr 43min 4hr 47min Female Workers 4hr 56min 3hr 42min Male Non-workers 4hr 15min 4hr 30min Female Non-workers 4hr 16min 4hr 15min

PAGE 63

7.1 Origin Vertex Location All workers Thane Survey vs. NHTS: This comparison is shown in Figure 7.1. It is clear from figure that the actual first departure time distribution for Thane worker sample is later in the day (9:00 AM) compared to NHTS worker sample (8:00 AM). There is also clear difference between vertex distributions too and for Indian workers peak is around 7:00 AM while for U.S. workers it is around 6:00 AM. The corresponding E[u] values for India and U.S. are 1 hr 36 min and 3 hr 07 min. These differences are quite clearly explained in the context of international differences. Work schedules in India are generally 9:00 AM to 5:00 PM and people tend to perceive work start times as less constraining when compared to developed countries such as the United States. There are several factors affecting the perception and they are discussed later in this section. 0.0010.0020.0030.0040.0050.0060.0070.0080.0090.00100.003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Workers Thane Obs Workers Thane Vertex Workers NHTS Obs Workers NHTS Vertex Figure 7.1 Distribution of Origin Vertex and First Departure Time: All Workers Thane Survey vs. All Workers NHTS Figure 7.2 and 7.3 shows comparison between Thane and U.S. male and female workers and in all cases the actual first departure distribution as well as vertex distribution is shifted later for Thane workers. It is important to note that while there is considerable difference in male and female worker sample departure time and vertex distribution in Indian context, in U.S. it is negligible, indicating relatively more involvement of Indian working females in household obligations and childcare.

PAGE 64

01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Workers Thane Obs Male Workers Thane Vertex Male Workers NHTS Obs Male Workers NHTS Vertex Figure 7.2 Distribution of Origin Vertex and First Departure Time: Male Workers Thane Survey vs. Male Workers NHTS 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Female Workers Thane Obs Female Workers Thane Vertex Female Workers NHTS Obs Female Workers NHTS Vertex Figure 7.3 Distribution of Origin Vertex and First Departure Time: Female Workers Thane Survey vs. Female Workers NHTS

PAGE 65

All Non-workers Thane Survey vs. NHTS: This comparison is shown in Figure 7.4. The actual departure time distribution for non-workers is wavy in Thane sample and has two small peaks compared to rather clear but flat distribution for U.S. non-workers. The vertex distribution for Thane Survey sample is bimodal peaking around as early as 5:00 AM and later around 8:00 AM while for NHTS it peaks around 7:00 AM. The Indian sample consists of a large number of students in non-workers segment and most of them are dependent on transit, the origin vertex is shifted earlier in the day. Since the analysis is for adults only (18 and more years old) most of the students must be college going students. In United States, different to India, most of the college going students work part time or full time and use auto for travel. Hence they are not clearly differentiable by workers and that is why the non-workers sample of NHTS non-workers is clearly unimodal for actual departure time as well as vertex distribution. The E[u] values for Thane non-worker sample is 1 hr 28 min and for NHTS non-worker sample it is 3 hr 19 min. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Non-Workers Thane Obs Non-Workers Thane Vertex Non-Workers NHTS Obs Non-Workers NHTS Vertex Figure 7.4 Distribution of Origin Vertex and First Departure Time: All Non-workers Thane Survey vs. All Non-workers NHTS Comparison of non-workers for developed and developing countries by gender status is shown in Figure 7.5 and 7.6 There is clear difference in Thane and NHTS sample actual departure time for both males and females. Female non-workers leave their home later in the day compared to U.S. females showing greater participation in household obligations and housemaking.

PAGE 66

01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Non-Workers Thane Obs Male Non-Workers Thane Vertex Male Non-Workers NHTS Obs Male Non-Workers NHTS Vertex Figure 7.5 Distribution of Origin Vertex and First Departure Time: Male Non-workers Thane Survey vs. Male Non-workers NHTS 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Female Non-Workers Thane Obs Female Non-Workers Thane Vertex Female Non-Workers NHTS Obs Female Non-Workers NHTS Vertex Figure 7.6 Distribution of Origin Vertex and First Departure Time: Female Non-workers Thane Survey vs. Female Non-Workers NHTS

PAGE 67

The origin prism vertex location represents the earliest time when a person can leave home if need arises. Though the actual first departure time distribution for NHTS sample is earlier than the Thane sample because the work starts at 8:00 AM in United States compared to 9:00 AM in India, it is important to note that the average difference between the actual departure distribution and vertex distribution E[u] is also substantially different in two countries. While the E[u] value for all the segments in NHTS sample is between 3 hr to 3 hr 20 min, it should be noted that the E[u] value is only in 1hr 30 min to 1 hr 50 min range in Indian context except for the male non-worker sample. This means that the United States workers perceive them more capable of leaving home early if the need arises. It is quite common in United States to make early morning trip to airport to catch a flight and travelers do not consider the system a hindrance for such early morning activities. A large number of people make early morning trips to drop their kids to daycare center or school and thus shift the vertex as well as distribution earlier. In a developing country like India, where the vehicle ownership level is very poor, most of the workers make their trips by transit and generally there is certain level of uncertainty of schedule and complexities and discomfort of transfer associated with transit and this should result in stretched prism (larger E[u]) but this is certainly not the case in India as evident by small E[u] values. Some of the possible explanations for peoples perception of constrained vertices are as follow: In the metropolitan city and suburbs of Bombay (Thane) rail is the major mode of public transportation and in peak hour its frequency is as high as 3 minutes and that makes commuters more comfortable about transit schedule because even if they miss one train they can catch next one within 3 minutes. As it was shown in Chapter 4 that the average number of trips per person per day in United States is around five compared to little over two trips per person per day in India. Two trips per day (from home to work and work to return home) leave little scope for engagement in other activities and obligations. The trip chaining analysis of Thane and NHTS data (working paper at University of South Florida) suggest that only 0.19 percent commuters in India engage in non-work related activities before work and on the way to work whereas in United States 30.2 percent of commuters shows trip chaining pattern where the commuter involve in at least one non-work related activities before work or on the way to work.

PAGE 68

It should also be noted that as trip chaining increases and becomes more complex, uncertainty in transit schedule increases but without trip chaining transit is reliable enough to constrain the prism vertices. Several socio-economic, institutional and infrastructural factors play important role in deciding the origin prism vertex location (example: In India there is no concept of dropping kids to day care center because large number of homemakers take cares of children) and hence, the need for Indians to leave their home does not arise and the origin prism vertices remains constrained.

PAGE 69

7.2 Terminal Vertex Location All workers Thane Survey vs. NHTS: This comparison is shown in Figure 7.7. The actual terminal arrival time distributions for both Thane and U.S. workers are similar and peaks in 6:00 to 7:00 PM range consistent with workers schedule constrains but the vertex distribution is substantially different. While the vertex distribution for Thane worker sample peaks around 9:00 PM, it peaks around 11:00 PM for NHTS worker sample. This can be easily explained by the fact that India is still a developing country and there is not much to do after sun set. A detailed explanation is given later. The corresponding E[u] values are 4 hr 32 min for Thane sample and 5 hr 21 min for NHTS sample. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Workers Thane Obs Workers Thane Vertex Workers NHTS Obs Workers NHTS Vertex Figure 7.7 Distribution of Terminal Vertex and Final Arrival Time: All Workers Thane Survey vs. All Workers NHTS Figure 7.8 and 7.9 shows comparison between Thane and NHTS male and female worker sample and as it is evident from the figures that despite similar arrival time distribution for male workers vertex location is shifted later for U.S. sample and the shift is more prominent in female workers probably due to their more concern for household obligations than their American counterpart and perhaps also due to feelings of insecurity later in night in traveling in transit in a developing country.

PAGE 70

01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Workers Thane Obs Male Workers Thane Vertex Male Workers NHTS Obs Male Workers NHTS Vertex Figure 7.8 Distribution of Terminal Vertex and Final Arrival Time: Male Workers Thane Survey vs. Male Workers NHTS 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Female Workers Thane Obs Female Workers Thane Vertex Female Workers NHTS Obs Female Workers NHTS Vertex Figure 7.9 Distribution of Terminal Vertex and Final Arrival Time: Female Workers Thane Survey vs. Female Workers NHTS

PAGE 71

All Non-workers Thane Survey vs. NHTS: This comparison is shown in Figure 7.10. The actual terminal arrival time distributions for both Thane and NHTS sample are wavy and flat with no clear peaks. The vertex distribution for NHTS sample is shifted later in evening compared to Thane sample (peak around 9:00 PM) and is bimodal with one small peak around 8:00 PM and other around 10:00 PM. The E[u] values for Thane non-worker sample is 4 hr 22 min and for NHTS it is 3 hr 15 min. 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Non-Workers Thane Obs Non-Workers Thane Vertex Non Workers NHTS Obs Non Workers NHTS Vertex Figure 7.10 Distribution of Terminal Vertex and Final Arrival Time: All Non-workers Thane Survey vs. All Non-workers NHTS Comparison of non-workers for developed and developing countries by gender status is shown in Figure 7.11 and 7.12. Again both male and female in Thane sample have vertex distribution peaks earlier than NHTS sample and the vertex distribution is unimodal for males and bimodal for females. The vertex distribution is bimodal for both males and females in NHTS sample.

PAGE 72

01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Male Non-Workers Thane Obs Male Non-Workers Thane Vertex Male Non-Workers NHTS Obs Male Non-Workers NHTS Vertex Figure 7.11 Distribution of Terminal Vertex and Final Arrival Time: Male Non-workers Thane Survey vs. Male Non-workers NHTS 01020304050607080901003456789101112131415161718192021222324252627Time of day (hr)Percent of Sample Female Non-Workers Thane Obs Female Non-Workers Thane Vertex Female Non-Workers NHTS Obs Female Non-Workers NHTS Vertex Figure 7.12 Distribution of Terminal Vertex and Final Arrival Time: Female Non-workers Thane Survey vs. Female Non-Workers NHTS

PAGE 73

The prism vertex locations are also dependent on several factors including occupational distribution (high income technical and business/professional workers may have a less constraining work start time, but are willing to leave early when they have to do so) number of children in household, vehicle ownership level and income level. So, for example commuters with more household obligations and children might return later because of the need to perform household obligations, running errands, and serve child trips but at the same time they perceive their vertex constraints to be earlier for exactly the same reasons as the need to get home early to take care of household members and obligations. Though both workers and non-workers in India have actual terminal arrival time distributions not much different from their U.S. counterpart they perceive their prism more constrained. This may be attributed to institutional and household factors. From an institutional standpoint, workers generally leave work quite punctually in India. The concept of overtime and performance driven work are only beginning to enter the Indian work culture. In addition, once work is completed, there are not many recreational and other opportunities for people. Not only are the opportunities few and far between, but they tend to be expensive for the average Indian worker/household. Another explanation is that Indian business establishments and service organizations 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 opportunity to engage in other activities after work. These explanations are consistent with the daily trip rates seen in Table 3 where workers made, on average, two trips per day one trip to work or business and one trip back home. Finally, and most notably, the transportation system in India does not offer a level of service high enough to encourage engagement in evening non-work activities. People do not have personal transport and have to rely on non-motorized or public transport modes. These slow modes that generally deter trip chaining make it difficult for people to engage in other activities on the way home from work. Thus, the maturity of the transportation system and personal vehicle ownership play an important role in determining peoples activity engagement pattern relative to their prism constraints. Thus, although Indian workers perceive their prisms not only constrained (though they are not much different as it is evident from E[u] values) but at the same time they are not able to take

PAGE 74

advantage of the time available within the prism due to institutional, transportation system, and social constraints. One can also view these differences in light of the social norms and culture in India. The structure of the Indian family and strong intra-household interactions suggest that people are very home-oriented in the Indian social context. While some of these norms are loosening with time, they continue to be strong determinants of peoples activity and travel patterns.

PAGE 75

CHAPTER 8 SUMMARY AND CONCLUSIONS The work has offered a detailed analysis of time space prism constrains for worker and non-worker samples by estimating time space prism vertices in the context of developing and developed countries. Using household travel survey samples from the City of Thane, India and United States (NHTS), this work estimates stochastic frontier models of origin and terminal prism vertex locations for males and females further stratified by their work status. This work builds upon previous work that recognizes the importance of modeling activity-travel patterns within the context of time space prism constraints. The origin prism vertex governs the first departure of the individual from home while the terminal prism vertex governs the final arrival of the individual at home for the day. The estimation results show that the stochastic frontier model estimation methodology can be suitably applied in a developing country context for estimating time space prism vertices. The work has facilitated unique comparisons. Comparisons between males and females, workers and non-workers, and between western and Indian samples were performed. The comparisons show that both gender and working status play a strong role in shaping the prism vertex location distributions. The analysis also shows that both gender and working status play different role in developing and developed country contexts. In the Thane non-worker sample, the presence of students and retirees was an important factor in determining prism vertex location distributions. In general, the comparisons of prism location distribution supported the notion that traditional household roles play an important part in shaping prism vertex location distributions and observed departure/arrival times. Non-workers and females, who bear most of the household obligations and child care giving duties, were found to have tighter prism constraints than counterparts. The international comparison between a developed country and a developing country revealed interesting findings. With respect to the origin vertex location distribution, it was found that the

PAGE 76

distribution of the prism vertex and the actual first departure time in Indian context were shifted about 1 hour later in the day compared to the western context. For the terminal vertex, the distribution of the actual arrival time is more or less same for both the countries but the prism vertex is shifted about 1 later in the evening. Some of the differences are: The E[u] value, the average difference between first departure time and origin vertex location distribution, for all market segments except male non-workers is about 1 hr and 30 min in Indian Context while it is about 3 hr 15 min for all market segments in United States context. The average difference between final arrival time and terminal vertex location distributions for all market segments is about 4 hr in Indian Context and 5 hr in United States Context but the actual arrival time distribution for both the countries is same. In Indian context, there are clear differences in first departure time and vertex distributions of male and female workers with female workers leaving home half an hour later due to household obligations and childcare but in United States both male and female workers show similar departure time and vertex distribution. In Indian non-worker sample presence of students was a major factor in deciding vertex locations, while in United States non-worker sample students were not differentiable perhaps because a large number of students work part time or full time and fall under workers category. These differences could be easily explained by social, institutional, cultural, and transportation system constraints and differences. Some of the factors are: Vehicle Availability and Modal Split: Vehicle ownership levels in India are very poor, only 5 percent of households have auto compared to 95 percent in United States, and hence most of the workers and non-workers use transit. Use of transit discourages trip chaining and hence lesser engagement in multiple activities resulting in earlier terminal vertex. Institutional Constrains: India is still a developing country and most of the business establishments are not open for 24 hours. The concept of overtime and night-shifts are just entering into Indian work culture and there are not many social and recreational opportunities and hence people prefer to come home earlier in the evening. Monetary Constraints: The vehicle ownership level and engagement in social and recreational activities can be directly correlated to quality of life and income level.

PAGE 77

Average Indian is not rich and hence does not have money to spend for activities/trips that keep him/her out of home for longer time. Household Constraints: Unlike United States, children in Indian context are rarely sent to daycare centers and female member of household take care of household obligations. That result in more burdens over female workers and non-workers to remain in-home for longer time. Infrastructure Constrains: The transportation system level of service is very poor and generally trip chaining and engagement in multiple activities is not feasible due to the discomfort and longer travel time associated with trips. Social/Cultural Constrains: The structure of Indian family and strong intra-household interactions suggest that people are very home oriented. Though there is a trend in young generation towards adopting ways of west, the country retains its norms and nuclear family structure resulting in one household member taking care of need of many members and hence important impact on travel characteristics. Lack of security in transit also discourages females to make late night trips. In model formulation only selected constraints were used. Variables like INCOME represented influence of monetary constrains on the location of prism vertices. But in the survey data, variables pertaining to institutional and social/cultural constrains were not included and hence their impact on vertex location is not quantified. Interaction among various household members is also not explicitly captured in data and exact nature of household obligations and child care is, to a great extent, unknown. One may conjecture that, as the transportation system improves in the developing country, vehicle ownership grows, and social norms loosen, people will increasingly take advantage of the time available within the time space prism. This has important implications for transportation planning, policy making, and quality of life. Weakening the constraints result in greater travel and activity engagements and that can be related to quality of life. The work has also shown that models of time space prism constraints can be estimated for developing country contexts. The availability of travel survey data provides the opportunity to start developing more sophisticated activity-based travel demand models in those contexts as well. Models of time space prism constraints that recognize the unique characteristics, constraints, and socio-cultural aspects of the developing country context are important ingredients in such a developmental effort. Some of the areas of further research are listed below.

PAGE 78

So far, the activity based models have been developed only for developed countries. Future research should focus on using data sets from developing countries to develop activity based travel demand models in those contexts and estimating time space prisms is important contribution to the development of such modeling efforts. Though the models have been estimated for two-day and three-day samples and it is found that prism locations are very consistent from one day to next day, the scheduling and undertaking of activities and trips within those prisms varies considerably from one day to next. Further research should focus on these aspects to develop robust models of activity and travel scheduling under time space prism constrains.

PAGE 79

REFERENCES 1. Hgerstrand, T. What about People in Regional Science? Papers of the Regional Science Association, Vol. 24, 1970, pp. 71. 2. Fujii, S., Y. Otsuka, R. Kitamura, and T. Monma. A Micro-simulation Model System of Individuals Daily Activity Behavior that Incorporates Spatial, Temporal and Coupling Constraints. Infrastructure Planning Review, Vol. 14, 1997, pp. 643. 3. Pendyala, R. M., R. Kitamura, C. Chen, and E. I. Pas. An Activity-Based Micro-Simulation Analysis of Transportation Control Measures. Transport Policy, Vol. 4, No. 3, 1997, pp. 183192. 4. Kitamura, R., C. Chen, R.M. Pendyala and R. Narayanan (2000) Micro-simulation of daily activity-travel patterns for travel demand forecasting. Transportation, 27(1), 25-51 5. Kitamura, R., T. Yamamoto, K. Kishizawa, and R. M. Pendyala. Stochastic Frontier Models of Prism Vertices. In Transportation Research Record: Journal of the Transportation Research Board, No. 1718, TRB, National Research Council, Washington, D.C., 2000, pp. 18-26. 6. Aigner, D., C. A. K. Lovell, and P. Schmidt. Formulation and Estimation of Stochastic Frontier Production Function Models. Journal of Econometrics, Vol. 6, 1977, pp. 21. 7. Pendyala, R. M., T. Yamamoto, and R. Kitamura. On the Formulation of Time-Space Prisms to Model Constraints on Personal Activity-Travel Engagement. Transportation, Vol. 29, 2002, pp. 73-94. 8. Yamamoto, T., R. Kitamura, and R. M. Pendyala. Comparative Analysis of Time-Space Prism Vertices for Out-of-Home Activity Engagement on Working and Non-Working Days. Presented at 82 nd Annual Meeting of the Transportation Research Board, National Research Council, Washington, D.C., 2003. 9. Kondo, K., and R. Kitamura. Time-Space Constraints and the Formation of Trip Chains. Regional Science and Urban Economics, Vol. 17, 1987, pp. 49-65. 10. Nishii, K., and K. Kondo. Trip Linkages of Urban Railway Commuters under Time-Space Constraints: Some Empirical Observations. Transportation Research B, Vol. 26, No. 2, 1992, pp. 334. 11. Waldman, D. M. A. Stationary Point for the Stochastic Frontier Likelihood. Journal of Econometrics, Vol. 18, 1982, pp. 27579.

PAGE 80

12. Greene, W. H. LIMDEP Reference Guide 8.0. Econometric Software Inc., Plainview, New York, 2002


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 001469398
003 fts
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 040524s2004 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000299
035
(OCoLC)55731053
9
AJR1152
b SE
SFE0000299
040
FHM
c FHM
090
TA145
1 100
Nehra, Ram S.
0 245
Modeling time space prism constraints in a developing country context
h [electronic resource] /
by Ram S. Nehra.
260
[Tampa, Fla.] :
University of South Florida,
2004.
502
Thesis (M.S.C.E.)--University of South Florida, 2004.
504
Includes bibliographical references.
516
Text (Electronic thesis) 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 80 pages.
520
ABSTRACT: Recent developments in microsimulation modeling of activity and travel demand have called for the explicit recognition of time-space constraints under which individuals perform their activity and travel patterns. The estimation of time-space prism vertex locations, i.e., the perceived time constraints, is an important development in this context. Stochastic frontier modeling methodology offers a suitable framework for modeling and identifying the expected vertex locations of time space prisms within which people execute activity-travel patterns. In this work, stochastic frontier models of time space prism vertex locations are estimated for samples drawn from a household travel survey conducted in 2001 in the city of Thane on the west coast of India and National Household Travel Survey 2001, United States. This offers an opportunity to study time constraints governing activity travel patterns of individuals in a developing as well as developed country context. The work also includes comparisons between males and females, workers and non-workers, and developed and developing country contexts to better understand how socio-economic and socio-cultural norms and characteristics affect time space prism constraints. It is found that time space prism constraints in developing country data set can be modeled using the stochastic frontier modeling methodology. It is also found that significant differences exist between workers and non-workers and between males and females,possibly due to the more traditional gender and working status roles in the Indian context. Finally, both differences and similarities were noticed when comparisons were made between results obtained from the data set of India and United States. Many of these differences can be explained by the presence of other constraints including institutional, household, income, and transportation accessibility constraints that are generally significantly greater in the developing country context.
590
Adviser: Pendyala, Ram M.
653
travel behavior.
comparisons of behavior.
stochastic frontier models.
household travel survey.
activity based travel demand modeling.
690
Dissertations, Academic
z USF
x Civil Engineering
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.299