USF Libraries
USF Digital Collections

An analysis of household vehicle ownership and utilization patterns in the United States using the 2001 National Househo...

MISSING IMAGE

Material Information

Title:
An analysis of household vehicle ownership and utilization patterns in the United States using the 2001 National Household Travel Survey
Physical Description:
Book
Language:
English
Creator:
Pinjari, Abdul Rawoof
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
multinomial logit model
cars
SUVs
vans
structural equations model
descriptive analysis
socio-demographics
pickup trucks
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: Vehicle ownership and utilization have a profound influence on activity-travel patterns of individuals, vehicle emissions, fuel consumption, highway capacity, congestion and traffic safety. The influence could be further skewed by the diversity of the vehicle fleet. This thesis presents a detailed analysis of the 2001 National Household Travel Survey data to understand the vehicle ownership patterns, fleet mix, allocation and utilization in the context of household and person socio-demographic characteristics. Along with a rich descriptive analysis, models of vehicle ownership and utilization are estimated to distinguish four vehicle types; cars, SUVs (sport utility vehicles), vans and pickup trucks based on their ownership by households and utilization patterns by household members. The primary driver level vehicle utilization analysis provides insights into the extent of allocation of a vehicle to a single person. In addition to confirming many perceptions about the ownership, acquisition and utilization patterns of different types of vehicles, this analysis brings out some subtle differences and similarities among the vehicle types. The analysis results indicate a greater propensity to acquire and use larger vehicles such as minivans, sports utility vehicles and pickup trucks among certain socio-demographic segments of population. Increased ownership and use of vans and SUVs, and their usage as personal vehicles rather than just work vehicles warrants a need to revise vehicle type specific policies, transportation planning and control measures.
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 Abdul Rawoof Pinjari.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 66 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 - 001469378
oclc - 55644124
notis - AJR1132
usfldc doi - E14-SFE0000280
usfldc handle - e14.280
System ID:
SFS0024975:00001


This item is only available as the following downloads:


Full Text

PAGE 1

An Analysis of Household Vehicle Ownership and Utilization Patterns in the United States Using the 2001 National Household Travel Survey by Abdul Rawoof Pinjari 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: April 1, 2004 Keywords: cars, SUVs, vans, pickup trucks, socio-demographics, descriptive analysis, structural equations model, multinomial logit model Copyright 2004, Abdul Rawoof Pinjari

PAGE 2

ACKNOWLEDGEMENTS I express my sincere gratitude to my advi sor, Dr. Ram M. Pendyala, with whom it has been a wonderful experience to work. I tha nk him for the invaluable ideas, advice and availability that made work ing on my thesis an enjoya ble undertaking. The constant support, active encouragement, inspiration and an excellent and flexible academic atmosphere he provided through out the masters program has given me a great opportunity to learn and experience research. Thank to Drs. Steven E. Polzin and Jian J. Lu for serving in my master’s thesis committee. I also thank Juan Pernia in this regard. Interaction with Dr. Polzin in and outside the coursework has helped me learn and analyze better and augment the contents of this work. I would like to thank Drs. Edward Mierzejewski, Jian J. Lu and Xuehao Chu for the courses they have taught. Interacti on with them and researchers from CUTR on many occasions, especially through the ITE ac tivities, has been very valuable. I also thank the Department of Civil and Envir onmental Engineering for providing excellent facilities and research atmosphere. I am thankf ul to Sean Gilmore, Ingrid hall and all the people of the department office. They have been very friendly and he lpful. Thanks to my colleague Ashish Agarwal, who has worked with me for the descriptiv e analysis part of this work. He has also been my team partne r in many other projects I also thank Xin Ye for having helped me out with understand ing the modeling concepts. It has been a pleasure to work with all of my colleagues in the Tran sportation Systems Research Group at USF. I thank all of my friends for their wonderful company and su pport. I take this opportunity to thank my amazing family for their love, a ffection and nurturing. My parents have been a constant source of inspiration for me. My wonderful brother and lovely sister have always given me a warm companionship.

PAGE 3

i TABLE OF CONTENTS LIST OF TABLES iii LIST OF FIGURES v ABSTRACT vi CHAPTER 1: INTRODUCTION 1 1.1 Background 1 1.2 Motivation 1 1.3 Objectives of the Thesis 2 1.3 Organization of the Thesis 3 CHAPTER 2: LITERATURE REVIEW 4 2.1 Importance of Vehicle Ownership and Utilization 4 2.2 Trends in Vehicle Ownership and Utilization 4 2.3 Factors Affecting Vehicle Ownership and Utilization 5 2.4 Differences Among Vehicle Types 5 2.5 Modeling Vehicle Ownership and Utilization Patterns 6 2.5.1 Previous Research Reviews 6 2.5.2 Important Modeling Efforts in the Past 7 CHAPTER 3: DATA DESCRIPTION 10 3.1 The National Household Travel Survey 10 3.2 Data Preparation 11 3.2.1 Original Data Sets 11 3.2.2 Vehicle File Preparation 11 3.2.3 Primary Driver File Preparation 11 3.2.4 Household File Preparation 12 CHAPTER 4: DESCRIPTIVE ANALYSIS 13 4.1 Background 13 4.2 General Findings from the 2001 Nati onal Household Travel Survey 13 4.3 Vehicle Ownership 16 4.3.1 Vehicle Ownership by Type 17 4.3.2 Length of Ownership a nd Age of Vehicles 18 4.3.3 Recent Vehicle Acquisitions 20 4.3.4 Leased Versus Owned Vehicles 21

PAGE 4

ii 4.4 Vehicle Utilization 21 4.4.1 Vehicle Utilization Patterns 22 4.4.2 Primary Driver Vehicle Alloca tion and Utilization Patterns 27 CHAPTER 5: MODELING METHODOLOGY 31 5.1 Structural Equations Modeling 31 5.1.1 Structural Equatio ns Representation 31 5.1.2 Estimation 31 5.1.3 Asymptotically Distribution Free – Weighted Least Squares Estimation 32 5.1.4 Evaluation 33 5.2 Multinomial Logit Model 34 5.2.1 Random Utility Approach 34 5.2.2 Estimation 35 5.2.3 Evaluation 35 CHAPTER 6: MODELS OF VEHICLE OWNERSHIP AND UTILIZATION 37 6.1 Background 37 6.2 Structural Equations Model of Vehi cle Ownership and Da ily Utilization 37 6.2.1 Vehicle Ownership 39 6.2.2 Vehicle Utilization 40 6.3 Multinomial Logit Model of Recent Vehicle Acquisitions 45 6.4 Multinomial Logit Model of Driver’s Vehicle Type Choice for a Trip 46 CHAPTER 7: CONCLUSIONS AND FUTURE RESEARCH 50 7.1 Conclusions 50 7.2 Future Research 52 REFERENCES 54

PAGE 5

iii LIST OF TABLES Table 4.1 Household Characteris tics of the 2001 NHTS Data 14 Table 4.2 Person Characteristics of the 2001 NHTS Data 15 Table 4.3 Vehicle Type Dist ribution by Household Type 17 Table 4.4 Distribution of Single Vehi cle Households by Vehicle Type 18 Table 4.5 Distribution of Two vehicl e Households by Fleet Combination 18 Table 4.6 Cross Tabulation of Two Vehicl e Households by Age of Vehicles 19 Table 4.7 Distribution of Number of Years of Owne rship by Vehicle Type 20 Table 4.8 Distribution of Vehi cle Age by Vehicle Type 20 Table 4.9 Distribution of Households by Recently Owned Vehicle Type 20 Table 4.10 Distribution of Annua l Mileage by Vehicle Type 22 Table 4.11 Annual Mileage Af ter Controlling for Vehicl e Availability and Age 23 Table 4.12 Daily Travel Characteristi cs by Vehicle Type on Weekdays 24 Table 4.13 Daily Travel Characteristi cs by Vehicle Type on Weekends 25 Table 4.14 Primary Driver’s Socio-Demogr aphic Characteristics by Vehicle Type 28 Table 4.15 Vehicle Utilization by Primary Drivers 29 Table 6.1 Direct Effects, Structural E quations Model of Vehicle Ownership and Utilization 42 Table 6.2 Total Effects, Structural E quations Model of Vehicle Ownership and Utilization 43 Table 6.3 Estimated Variance-Covariance Matrix of the Disturbances of the Equations for Endogenous Variables 44

PAGE 6

iv Table 6.4 Multinomial Logit Model for th e Recently Acquired Vehicle Type 48 Table 6.5 Multinomial Logit Model for Driv er’s Vehicle Type Choice for a Trip 49

PAGE 7

v LIST OF FIGURES Figure 4.1 Trip Rates by Purpose by Vehicle Type 26 Figure 6.1 Structural Equations Framew ork of Household Vehicle Ownership Trends and Daily Ut ilization Patterns 44

PAGE 8

vi AN ANALYSIS OF HOUSEHOLD VEHI CLE OWNERSHIP AND UTILIZATION PATTERNS IN THE UNITED ST ATES USING THE 2001 NATIONAL HOUSEHOLD TRAVEL SURVEY Abdul Rawoof Pinjari ABSTRACT Vehicle ownership and utiliza tion have a profound influence on activity-travel patterns of individuals, vehicle emissions, fuel c onsumption, highway capacity, congestion and traffic safety. The influence could be further skewed by the diversity of the vehicle fleet. This thesis presents a detailed analysis of the 2001 National Household Travel Survey data to understand the vehicle ownership patter ns, fleet mix, allocation and utilization in the context of household and person socio-dem ographic characteristic s. Along with a rich descriptive analysis, models of vehicle ownership a nd utilization are estimated to distinguish four vehicle types; cars, SUVs (sport utility vehicl es), vans and pickup trucks based on their ownership by households and utilization patterns by household members. The primary driver level vehicl e utilization analysis provides insights into the extent of allocation of a vehicle to a single person. In addition to confirming many perceptions about the ownership, acquisition and utilizatio n patterns of different types of vehicles, this analysis brings out some subtle differences and simila rities among the vehicle types. The analysis results indicate a greater propensity to acquire and use larger vehicles such as minivans, sports utility vehicles and pi ckup trucks among certain socio-demographic segments of population. Increased ownership and use of vans and SUVs, and their usage as personal vehicles rather than just work ve hicles warrants a need to revise vehicle type specific policies, transportation planning and control measures.

PAGE 9

1 CHAPTER 1 INTRODUCTION 1.1 Background Household vehicle ownership and utilization is an important facet of revealed travel behavior. Over the past few decades, vehicl e ownership levels and vehicle utilization levels have consistently grown both on a macroscopic (population) and a microscopic (individual) level. Increasing levels of comp lexity in people’s activity and trip chaining patterns have also contributed to changes in vehicle ownership and utilization patterns. In addition, new vehicle technology and design ha s made it possible for people to own and use different types of vehicles including cars vans, sport utility vehicles (SUV), trucks, and so on. However, cars have continued to lose thei r market share of private vehicles; the share of cars has gone down from 80 per cent in 1977 to 65 percent in 1995. In the meantime, minivans and sport utility vehicles (SUVs) claimed a larger market share. (Hu, et. al. 1999). The percentage of cars present in today’s fleet is about 60 while minivans, SUVs and pickup trucks have grown in the market share to about 39 percent. The NHTS data indicates that out of a ll the household vehicles that were acquired in the past oneyear (with respect to the NHTS survey time of April 2001 through May 2002) 55.5 percent were cars, 9 percent vans, 14.8 pe rcent SUVs and 16.6 percent were pickup trucks. Percentage of cars of the remaining vehicles in the fl eet is 58.2 indi cating that cars are losing their market shar e to light duty vehicles. Industry sales data also shows an increas e in the light duty vehicles (vans, SUVs, pickup trucks). 1999 data of Polk Company indi cates that Light Duty Trucks capture 51 percent of new passenger vehicle sales (kockelman, et. al. 2000). Between 1975 and 2003, market share for new passenger cars decreas ed from 81 to 52 percent. Growth in the light truck market has been led recently by the increase in the ma rket share of SUVs. The SUV market share increased by more than a factor of ten, from less than two percent of the overall new light vehicle market in 1975 to 24 percent of the market in 2003. Over the same period, the market share for vans in creased by 80 percent, while that for pickups remained relatively constant. (Hellman, et. al 2000). These trends, s howing an increasing share of vans, SUVs and pickup trucks may have several implications to transportation planning, policy and perhaps regulatory actio n from the perspective of fuel economy, emission standards and transportation safety. 1.2 Motivation Increasing vehicle ownership and utilization can influence the trans portation system in terms of its performance a nd the externalities it coul d cause to environment and community. The diversity in the vehicle fleet can skew this influence. In addition, the differences in ownership and utilization of di fferent vehicles can further influence these

PAGE 10

2 effects. The increased ownership and use of di fferent types of vehicl es like vans, SUVs and pickup trucks can have several implica tions to energy consumption, fuel economy, emission levels, highway capacity and safety. Th ese vehicles consume more fuel per mile than ordinary automobiles. They are huge and occupy more roadway and parking space, but are still measured equivalent to cars in capacity considerations Accident frequencies and injury severity could substantially differ across different types of vehicles (Chang, et. al. 1999, Ulfarsson, et. al. 2004, Kockelman, et al. 2002). Most of these differences across different types of vehicles have severa l important implications to transportation planning and policy in the context of fuel economy and energy use, vehicle emissions, traffic congestion, and safety. Light Duty Truck classification protects SUVs, pickup trucks and vans from various stringent regul ations. Pickup trucks were owned and used primarily for work purposes and blue collared jobs. The perceived difficulty of these vehicles in meeting the stringent emi ssion standards was the reason behind their classification as LDTs (Light Duty Trucks). Minivans and SUVs were also classified as LDTs, based on structural similarities, along with pickup trucks. These vehicles enjoy a variety of regulatory protect ions like higher emission caps and do not endure luxurygoods or gas-guzzler taxes. (Kockelman, et. al 2000). All of these factors entail a closer look at the ownership and utilization patterns of these vehicle types. An in-depth analysis is required to assess the differences in owne rship and utilization pa tterns of different types of vehicles. It is also important to understand the demography of current vehicle fleet. In other words, we have to take stock of the current diversity of vehicle fleet. Households own different types of vehicles to utilize them for daily travel. The travel needs of a household depend upon its soci o-demographic characteristics. Hence it is the socio-demographics of a household that shape its vehicle ownership type and level. So it is important to understand the isolated effect of each of the socio-demographic factor on household vehicle owne rship and utilization patterns To state succinctly, it is necessary to understand ‘who owns, what type s of vehicles, why, a nd who drives, what types of vehicles, for going where, and for wh at purposes. Essentially, it is important to analyze and distinguish different types of vehicles (cars, vans, SUVs and pickup trucks) based on their ownership utilization pattern s in the context of socio-demographic attributes of households that own th em and persons that drive them. 1.3 Objectives of the Thesis As vehicle ownership and utilization by vehicle t ype is of much interest to transportation planners and policy makers, this thesis aims to provide a rather detailed analysis of vehicle ownership, fleet mix a nd utilization pattern s of different vehicle types in the United States, using the data recently ava ilable from the 2001 National Household Travel Survey (NHTS). An extensive analysis is pr ovided in the context of socio-demographic attributes to distinguish the four vehicle types; cars, vans SUVs and pickup trucks, based on their ownership patterns, trends in rece nt acquisitions, allocation and utilization patterns. While it would certainly be inte resting to study ve hicle ownership and utilization patterns over time using a series of nationwide personal transportation survey (NPTS) data sets, it was considered beyond th e scope of this thesis which is aimed at taking stock of the current situation. A l ongitudinal analysis of vehicle ownership and utilization should undoubtedly be undertaken and that remain s as a subsequent research

PAGE 11

3 effort. This study considers cars, SUVs, vans and pickup trucks as the major vehicle types. The objectives of the study are briefly listed below: To carry out a detailed desc riptive analysis of the 2001 NHTS data in order to distinguish cars, vans, SUVs and pic kup trucks based on their ownership patterns in terms of market share, ag e and length of ownership and recent vehicle acquisitions. To perform a detailed desc riptive analysis of the 2001 NHTS data in order to distinguish cars, vans, SUVs and pic kup trucks based on their utilization patterns in terms of annual mileag e, daily travel, weekday-weekend differences, trip characteristics (trip purpose, occupancy, trip length etc.) and allocation to primary drivers. To understand the structural relationshi ps between socio-demographic factors and vehicle ownership and u tilization patterns in a uni fied framework. (A joint structural equations model of household vehicle ownership a nd utilization is developed in this context, which can enable the isolation of each sociodemographic factor in a simultaneous and multivariate framework.) To understand the vehicle type choice be havior in recent vehicle acquisitions. (A multinomial logit model for the recently owned vehicle type is developed to understand the differences in choi ce making and preferences across vehicle types in the recent vehicle acquisitions.) To analyze the vehicle type choice be havior in trip making. (A multinomial logit model for the vehicle type chosen by a driver for his/her trip is developed in order to understand the vehi cle utilization patterns.) 1.4 Organization of the Thesis The rest of this thesis is organized as follo ws. Next chapter provides an extensive review of the literature available, highlighting the importance of the topic. Various research efforts in the direction of distinguishing different vehicl e types and several important modeling efforts of vehicle ownership and u tilization are reviewed. The third chapter describes the 2001 NHTS data and provides a detailed descripti on of the data preparation process for further analysis. The fourth chap ter presents the results of an extensive descriptive analysis of vehicle ownership, utilization a nd allocation patterns. The fifth chapter furnishes details of the methodology of the modeling frameworks used in the study. Sixth chapter focuses on the models of vehicle ownership, ut ilization, and choice behavior. It includes a structural equations m odel for vehicle ownership and utilization in a unified framework, a multinomial logit model of a household’s choice of the type of recently acquired vehicle and a multinomial logit model of driver’s vehicle type choice for his/her trip. Finally, conclusions and imp lications of the findings for policy measures and planning practice are discussed in final chapter (7) of the thesis along with further extensions of the topic for future research.

PAGE 12

4 CHAPTER 2 LITERATURE REVIEW 2.1 Importance of Vehicle Ownership and Utilization Vehicle ownership and utilization has been th e subject of substantia l amount of research in the past. As these concepts are central to transportation planni ng and decision-making, researchers and data analysts have spent considerable time and attention to these issues. Vehicle ownership and av ailability are the key determinan ts of mode choice. Pucher, et. al. (2003) showed how even a single car owne d would affect the tr avel behavior of a household. Considering the share of transit, Polzin, et. al. (2003) found a drop from 19.1 percentage of total tr ips by households with no car to only 2.75 percentage of trips by households with one car. Vehicle ownership and utilization patterns can have profound impact on the disaggregate travel characterist ics and thus on the ove r all travel demand. Liss, et. al. (2003) analyzed the 2001 NPTS/N HTS data and found that the annual miles for each individual vehicle declined slightly as the vehicle travel by household members is spread over more vehicles but after co ntrolling for income, households having more vehicles than drivers accounted for more tr ips and mileage than households with fewer vehicles than drivers, also shown by McGu ckin, et. al. (2003). Increased ownership and use of different types of vehicles certainly has implications in the context of fuel consumption, vehicle emissions and air qualit y, crash injury severity, accident rates, highway safety and general he alth issues. Matthew (2003) presented a possibility of relationship between the extent of vehicle owne rship, availability and use to the extent of active walking and gene ral health issues. 2.2 Trends in Vehicle Ownership and Utilization Researchers have also concentrated on the l ongitudinal aspects of vehicle ownership and utilization. Polzin, et. al. (2003) showed th e increased availability of vehicles through changes in the ratios of vehicles to adu lts, drivers and workers since 1969. They also made an important observation that the number of zero vehicle households has only declined from 11.4 million in 1969 to 10.9 in 2001, even though the share of zero car households has declined. This reveals the im portance of vehicle ownership at different levels, especially zero vehicles. Murakami, et. al. (1999) analyzed the vehicle availability of persons with low income and pointed out that despite having fewer vehicles, people in low income households make most of their trips in private vehicles owned by someone else. Hu. (2003) presented the trends in incr easing vehicle ownership, increasing share of SUVs, vans and pickup (P/U) trucks, and the increased use of older vehicles in U.S. Pickrell, et. al. (1999) used 1969-95 NPTS data to offer insights into the changing patterns of household vehicle ownership by analyzing the growth in personal motor vehicle travel; changes in the number, type and age distribution of household motor vehicles; and the determinants of households' vehicle utiliz ation patterns. Hu, et. al.

PAGE 13

5 (1999) presented the changes in the availability and utiliza tion of household vehicles in their ‘Summary of Travel Trends’ repo rt utilizing the 19691995 NPTS data. They showed the continued loss of market share of automobiles (out of all private vehicles) from 80 percent in 1977 to 65 percent in 1995, while mini vans and SUVs gained the market share. They also presented the significant increase in the length of time vehicles were held and operated by households in 1995 when compared to that of 1969. Pisarski (1994) presented the trends a nd emphasized the implications of ageing of the vehicle fleet and increased travel on older vehicles ba sed on an analysis of 1969-90 NPTS data. 2.3 Factors Affecting Vehicl e Ownership and Utilization A wide variety of factors affect the vehi cle ownership and util ization patterns of a household. Pisarski (1996) emphasized the skew ing of auto ownership and usage by race, ethnicity and immigrant population. Pisarski (2003) pointed out the increased vehicle ownership by minorities could have profound impact on national transportation patterns and growth. He also pointed out the possible growth in travel as a result of increased access to and use of personal vehicles by young people, older population, women and racial and ethnic minorities. Gardenhire, et. al. (2001) found behavioral differences in the factors affecting auto ownership of low in come households compared to medium and high income households. Their analysis reve aled that poor househol ds convert income into automobiles at a higher rate and convert larger adult household size into automobiles at a lower rate than non-poor households. Hess. et. al ., (2002) tested a model, for Portland, Oregon, that explained automob ile ownership on the basis of household, neighborhood, and urban design ch aracteristics. They foun d a strong evidence of the effect of mixed land use on automobile ownership; as land use mix changed from homogeneous to diverse, the probability of owning an automobile decreased, ceteris paribus Karlaftis, et. al. (2002) in vestigated the effect of tr affic and network efficiency parameters on automobile ownership. They pointed out that traffic network and efficiency parameters did not, on the one hand, affect autolessness (zero vehicle ownership), but they did, on the other hand, affect the number of automobiles owned by a household. Purvis (1994) estimated auto owne rship models using the 1990 Census Public Use Microdata Sample (PUMS) and discusse d the strengths and weaknesses of using PUMS versus household travel survey data for aggregate auto ownership forecasting purposes. Choo, et. al. (2002) analyzed the dependence of vehicle type choice on person’s attitudes, personality, lif estyle and mobility choices. 2.4 Differences Among Vehicle Types Research in the recent past has also concentrated on di stinguishing various types of vehicles based on their ownership and utiliz ation. Hu (2003) empha sized the need to understand how various types of vehicles are being owned and used; specifically the need to address the question “Who owns what type of vehicle, going where, when, and for what purpose?” Hu, et. al. (1999), in their ‘Su mmary of Travel Trends’ report provided the changes in the distribution of vehicles by type utilizing the NPTS data from 1977 to 1995. They found that “Automobiles continued to lose their market share of private vehicles, from 80 percent in 1977 to 65 per cent in 1995. In the mean time, minivans and sport utility vehicles (SUVs) claimed a larger market share. Regardless of vehicle type,

PAGE 14

6 all vehicles were in operation longer in 1995 than in the past.” Pickre ll, et. al. (1999) utilized NPTS data from 1969 to 1995 to anal yze the growth in personal vehicle travel, changes in the number, types, and age dist ribution of household motor vehicles, and the determinants of household vehicle use patterns Kockelman, et. al. (2000) characterized and distinguished the ownership patterns and use of light duty trucks from that of passenger cars using the 1995 NPTS data. They used the NPTS 1995 data to estimate WLS (weighted least squares) models of VMT on each vehicle, negative binomial regression models of the numbe r of person trips carried by the vehicle on a travel day, ordered probit models for vehicle occupancy for a trip, multinomial logit models for the vehicle type chosen for trip by driver, and multinomial logit models for the newest vehicle type owned by the household. They found the socio-economic attributes and vehicle prizes to be the key determinants of vehicle type choice, ownership and utilization. They found that “the average LD T (Light Duty Truck) is used over long distances with more people aboard and is purchased by wealthier households living in less dense neighborhoods.” Anderson, et. al. (2 001) distinguished th e ownership and use characteristics of pickup trucks in the 1995 NPTS. They observed that households with more vehicles, rural households, single-fa mily dwelling unit and mobile home households, and middle-income households t ypically owned pickup trucks. Men, drivers with less education and full-time workers we re more likely to drive a pickup truck on a travel day than their counterparts. They obser ved that a “higher porti on of trips to work, work-related trips, long trips, and trips w ith fewer people were by pickup truck.” Anderson, et. al. (1999) also characterized pickup truck drivers with respect to demographic factors, and their behavior from safety point of view. Niemer, et. al. (2001) used 1995 NPTS data to analyze the vehicle fl eet with respect to who were driving the vehicles, what types of trips were the vehi cles being used for, and where the primary accumulation of vehicle miles of travel (VMT) was occurring. Kockelman. (2000) characterized light duty trucks and passenger cars based on emissions, safety, and fuel economy and examined household usage differe nces among the vehicle types. The paper pointed out that LDTs are used in ways very similar to passenger cars but enjoy lenient regulation. 2.5 Modeling Vehicle Ownership and Utilization Patterns This section presents an ex tensive review of previous work that involves modeling various aspects of vehicle ownership, vehi cle type choice, and vehicle utilization. 2.5.1 Previous Research Reviews A research review paper by Tardiff (1980) cla ssified the models in the research by the kind of vehicle choice under th e study (Vehicle ownership le vels, purchased new vehicle type choice, joint ownership le vel and mode choice etc) discu ssed the models on the basis of function forms, explanatory variables and results. The author highlighted the advantages of joint models of vehicle owners hip and combination and vehicle type choice over individual conditional ch oice models. He also emphasized the need to use the previous vehicle ownership as important factor in de ciding the recent ownership decisions. This thesis incor porates some of these impr ovements by presenting a joint model system of household vehicle owne rship and utilization and by considering

PAGE 15

7 previous vehicle ownership level and comb ination in the choice making behavior of recent vehicle purchases. Mannering, et. al. (1985) presented a research review with respect to relationship of number and type of autos owned, usage, VMT (Vehicle Miles traveled) on each vehicle, market equili brium and dynamic components of vehicle demand. The above two reviews also sugges ted some directions for automobile ownership, utilization and dema nd models. Given that the a bove reviews are relatively former in time, Choo, et. al. (2002) discussed the above two research reviews in detail and also provided an excellent literature re view of vehicle type choice models and vehicle use models estimated in current research. They re viewed and assessed various analyses present in the research in contex t of various aspects; modeling, explanatory variables included, and significant results of the efforts. They also provide different vehicle type classifications present in the academic literature and various statistical reports. Though many extensive literature review s exist in the context of our topic, this section also reviews of some of the important modeling efforts in the past in the context of the current topic. 2.5.2 Important Modeling Efforts in the Past Lave, et. al. (1979) estimated a multinomial logit model of vehicle type choice for households buying a new car fo r a stratified random sample of 541 new car buyers in 1976. The estimates indicated that larger househ olds were more likely to buy subcompact cars while households with more miles driven were more likely to choose larger cars. Manski, et. al. (1980) presented multinomial logit models of vehicle type choice conditional on the number of vehicles owne d (joint choice model foe two-vehicle households) for a nationwide sample of 1,200 households. Their models had 25 randomly selected alternative vehicle types, out of 600 different t ypes by make, model and vintage in the universal set, along w ith the chosen alte rnative. They found that seating and luggage space positively aff ected vehicle choice in larg er single-vehicle households. Scrappage rate showed a negative effect for the vehicle type choice. Transaction cost variable showed a negative affect on the choice probability due to the inertia or propensity to retain existin g vehicle. Hocherman, et. al (1983) estimated two-stage nested logit model of vehicle type purchase d, conditional on a purchase being made. The upper level was for a choice between buying a fi rst car or replacing an existing car and the lower level choice making was for the chos en alternative plus 19 randomly selected alternatives from the universal set of 950 vehi cle types. They found that the attributes engine size of previous car, brand loyalty, number of same type of cars present along with income showed a positive effect on the vehicle type choice. Berkovec, et. al. (1985) developed a nest ed logit model of vehicle type held households for a U.S nation wide sample of 237 single-vehicle households with the upper level having three vehicle age group cate gories and the lower level having 5 vehicle classes based on size. Their analysis suggested that number of seats had a positive effect and the vehicle size attributes like turning ra dius in urban areas had a negative effect perhaps due to parking issues. Berkovec (1985) presented a simulation model to forecast automobile demand under various gas price po licies. He estimated log linear model of scrappage rate and then developed a nested logit model of vehicle type choice conditional on household vehicle ownership. The simulation model results indicated that households

PAGE 16

8 were less likely to change vehicle types ow ned, as gas price increa ses. Thus the total sales of new vehicles would decrease and scrappage rates of ol der vehicles would increase due to fuel inefficiency as the gas price increased. Mannering et. al. (1985) developed a dynamic model of household vehicle ownership and utilization behavior in whic h they estimated models of vehicle type choice, utilization, and quantity choice for single-vehicle households and two-vehicle households. They used lagged utilization variab les of a vehicle type were taken as brand loyalty variables which showed positive effect on the vehicle type choice. The estimates of the choice probability with respect to inco me and capital cost were less elastic for two vehicle households than for si ngle vehicle households. Manne ring, et, al. (2002) present a nested logit model of vehicle type choi ce conditional on different vehicle acquisition methods such as leasing financing etc. Th e results indicate that regardless of the acquisition type, households are more likely to choose a vehicle with higher brand loyalty and residual values. Households leasing a ve hicle tend to place high value on attributes such as passenger side air bag and horsepow er and are more likel y to choose larger vehicles and SUVs. Kitamura, et. al. (2000) es timated ordinary least squares models of annual vehicle mileage for the vehicle last acquired by a household as a function of primary driver and secondary driver attribut es, vehicle attributes household attributes, and residential attributes such as accessibi lity indices and residential density. The selectivity bias correction terms they incorpor ated to deal with the potential correlation between error terms of vehicle type choi ce and vehicle use were found to be not significantly contributing to the model improvement. Mannering (1983) estimated a simultaneous equation system for a sample of two vehicle households to study vehicle use in multi-vehicle households. The results highlighted that income and vehicle fuel e fficiency are crucial in the allocation of household travel among vehicl es. Mannering (1986) furthe r extended this work and showed the potential bias in results due to the use of vehicl e attributes (endogenous variables) as exogenous variab les in the household vehicle ut ilization models. Golob, et. al. (8) estimated structural equation mode ls of household annual VMT (vehicle miles traveled) by vehicle type fo r single-vehicle households and two-vehicle households separately for a sample of 4,747 California hous eholds. The results indicate that women tend to drive less, while workers tend to driv e more. Households that own mini or small cars drive less and households wi th older heads tend to drive less, while those with more children or high income drive more. Golob (1990) formulated a structural equations model linking car ownership, travel time by car, public transit and non-motorized modes at two points of time for Netherlands da ta. The model specification included car ownership as ordered-response probit variables and all travel times as censored (tobit) continuous variables. Golob, et al. (1996) formulated and es timated a structural driver allocation and usage model for two vehicle households to study household vehicle usage behavior. Hensher (1985) developed six simultane ous equation models for one-, two-, and three-vehicle households for household vehicl e use in short and long run using three stage least squares method for a sample of 1,436 households from the first wave of a household panel survey in Sydney, Austra lia. These simultaneous equations model systems generally found household and pers on attributes, vehicle attributes, and

PAGE 17

9 residential attributes to be significant dete rminants of vehicle ownership and utilization. Hensher, et. al. (1985) developed a series of discrete-choice models to explain household automobile fleet: its com position and changes over time for a panel of 354 Sydney households. This dynamic model system allowe d for prior decisions, brand loyalty, and costs of transacting, which were found to be important. Bhat, et. al. (1998) compared a series of discrete choice modeling specifi cations and found that the unordered response model structure is the most appropriate fo r household auto owners hip modeling. Zhao, et. al. (2002) estimated a multivariate ne gative binomial model of household vehicle ownership by vehicle type for the 1995 NPTS data. The estimates suggest that household size, income, population density, and vehicle pr ice affect the vehicle ownership decisions of a household. SUVs are preferred most, and pickup trucks the leas t, by high income, large size households. In summary, household vehicle ownershi p and fleet combination models are all generally estimated in a simultaneous equatio ns framework. Least squares, structural equations models or discrete choice models we re used for the vehicle ownership and fleet combination and utilization patterns. Discrete choice models were also formulated for vehicle ownership combination. Various disagg regate vehicle type choice models are generally used for the vehicle type choice, in which vehicle and household characteristics are generally used as explanatory variables. Two types of vehicle type choice models; vehicle holdings and vehicle purchase models are generally formulated. The above is by no means a comprehensive review of the literature as it is truly quite vast. However, this section amply illust rates the importance that the profession has given to the study of vehicle ownership, util ization, allocation, and vehicle type choice. While this thesis does not provide new me thodologies for analyzing vehicle ownership and utilization, it provides a detailed descrip tive analysis of vehicl e type distribution, allocation and usage and carefully formulated models of vehicle ownership and usage using the most recent 2001 NHTS data. From that standpoint, it is useful in that it takes stock of the current situation and offers comparisons across demographic groups and vehicle types that may be useful in a po licy context. Next se ction provides a brief description of the National Household Travel Survey data sets used for this study.

PAGE 18

10 CHAPTER 3 DATA DESCRIPTION 3.1 The National Household Travel Survey The 2001 National Household Travel Survey (NHT S) data is used for the analysis of vehicle ownership and utilization patterns in this study. The NHTS, sponsored by the Federal Highway Administration (FHWA), the Bu reau of Transportation Statistics (BTS) and the National Highway Traffic Safety Admi nistration (NHTSA), is an integration of the two national travel surveys. They were previously called as the Nationwide Personal Transportation Survey (NPTS) and the American Travel Survey (ATS). The data sets, corresponding documentation and relevant info rmation can be accessed from the website developed by Oak Ridge Nati onal Laboratory (ORNL). One can also make use of the web-based analysis tools that are designed for preliminary analysis. The purpose of the NHTS interviews, conducted from April 2001 through May 2002, is to take an inventory of the daily a nd long-distance travel (over 50 miles from home) in the United States. There are approximately a to tal of 66,000 households in the final 2001 NHTS dataset. This analysis uses the sample of 26,000 house holds that are in the national sample, while the remaining 40,000 households from nine add-on areas are not used for this study. The study also exclude s the long-distance travel data. Essentially, this study makes use of the daily travel data of the nationally repres entative sample of 26,000 households that was released in Ja nuary 2003. The daily travel survey was conducted using Computer-Assisted Telephone Interviewing (C ATI) technology. Each household in the sample was assigned a specifi c 24-hour “Travel Day” and kept diaries to record all travel by all hous ehold members for the assigned day. The basic sampling method used for this survey is the stra tified random sampling technique with each stratum of random sample from each state in th e United States. The data is collected from a sample of the civilian, non-institutionali zed population of the United States. Hence, People living in college dormitories, nursing homes, other medical institutions, prisons, and military bases were excluded from the sample. This is the only data available at the national level, which includes the demographics of households, household member s, the vehicles owned by the households and detailed trip based inform ation on the daily and long-dist ance travel for all purposes by all modes. Hence, NHTS 2001 provides the opportunity to study the current vehicle ownership, fleet combination, allocation a nd utilization pattern s through linking and combining the vehicle travel with the demogr aphics of the travelers, the household and the vehicles owned by them. This analysis pr ovides a better understa nding of activity and travel patterns on personal vehicles, which can assist the planners and decision makers to effectively plan and formulate policies in th e context of transporta tion safety, energy consumption, and environmental impact and ge neral health. The next section provides a detailed description of the data prepar ation process for the proposed analysis.

PAGE 19

11 3.2 Data Preparation This section describes the process of building the data sets required for the analysis from the available 2001 NHTS data sets. 3.2.1 Original Data Sets The 2001 NHTS data contains four different da ta files; household fi le, person file, trip file and vehicle file. The household file, prepared based on th e household interview contains variables describing the househol d characteristics and household member characteristics that include the socio-demogr aphics, and geographic characteristics of the household and the demographic, and work ing status of all household members. The household file contains information on all members of the household (such as age, gender, and employment and driver’s license status) regardless of whether all of the members responded to th e Person Interview. The person file prepared based on the Person Interview, contains the demographic, dr iving, travel to work, travel evaluation and Internet use information of 60,282 members from the 26,038 households. The trip file contains the purpose, mode, distance and duration, temporal, occupancy, origin and destination characteristics of all the daily trips (248,517) made by all the persons in the trip file. All the four file s are linked through common variab les called identification (ID) variables, which enable the data combini ng and preparation for further analysis. 3.2.2 Vehicle File Preparation Each record in the vehicle file provides information about a part icular vehicle. The original vehicle file from the 2001 NHTS has variables describing th e vehicle attributes (make, model, type and year), ownership length, mileage, household attributes, and the person IDs of primary driver Additional variables desc ribing the primary driver characteristics (Age, Gender, Employment status etc) are added to this file from the 2001 NHTS person file based on the common househol d ID and person ID of primary driver. Now the vehicle file contains the attr ibutes of primary drivers as well. A set of variables for the total household trips carried by the ve hicle on the travel day is created for all trip purposes. These va riables were created for both the person trips and vehicle trips (or driver trips). Similarl y, additional variables were appended for daily mileage (VMT or Vehicle Miles of Travel) and duration the vehicle was driven. These variables describe the total household utilizatio n of the vehicle on the travel day. Another set of variables is created for the primary driver’s utilization of the vehicle. This set has variables for trip frequencies, total travel duration, and tr avel distance (VMT) of the primary driver on his/her vehi cle for all the purposes. 3.2.3 Primary Driver File Preparation Each person in the person file is appended w ith his/her trip frequenc ies, travel durations and the VMTs on the travel day for all purposes. Each person (record) in the person file is flagged, if he/she is a primary driver. For every primary driver’s record, variables are appended for the household vehicle number and the type of the vehicle he/she is a primary driver of. A separate primary drive file is also created in which each record is a primary driver. This has the information abou t his demographic characteristics like age, sex, working status etc and th e vehicle(s) number of the ho usehold and the type of the

PAGE 20

12 vehicle (s) that he is the primary driver of. Th is file also has all his/her travel information as a set of variables for his/ her trip frequency, total travel duration and the VMT on travel day. 3.2.4 Household File Preparation The household file is appended with the household vehicle fleet ownership and utilization variables. The variables for the vehicle fl eet combination are essentially the number vehicles of each type owned. The vehicle u tilization variables include the daily VMT on vehicles of each type and also the total daily household VMT. Household vehicle number and the characteristics of the most recent type of vehicle owned by the household are also appended. This includes the vehicle type its age, and the time when the vehicle was bought. Thus, the NHTS data is appended with required variable sets and ready for an extensive analysis. The next se ction presents a detailed de scriptive analysis of the 2001 NHTS and the findings in the context of vehicle ownershi p and utilization.

PAGE 21

13 CHAPTER 4 DESCRIPTIVE ANALYSIS 4.1 Background This section presents a detailed descrip tive analysis of the vehicle ownership and utilization patterns in the 2001 NHTS. First subsection provides general findings from NHTS of the current vehicle ownership and demographics of the households and the population in United States. The next two subsec tions provide the descriptive analysis for vehicle ownership and vehicle ut ilization patterns respectively. 4.2 General Findings from the 2001 National Household Travel Survey Tables 4.1 and 4.2 give an overview of the p opulation characteristic s in terms of the socio-demographics of households and hous ehold members respectively. Table 4.1 provides weighted analysis for an ove rview of household socio-demographic characteristics. There are a total of 107,368,651 (about 107 million) households of which 92.1 percent households own at least one vehicle. The average household size is 2.56 persons per household. About a quarter of them are single person households and another quarter of the households have more than three persons. On an average there are 0.67 children (<18yrs) and 1.31 employed persons per household. About two-thirds of the households have no children while one-fif th of the households reported having no worker. The average vehicle ownership is 1.9 vehicles per household. Only 7.9 percent of households have no vehicle. When the household attributes of zero-veh icle households are compared to those of other households, they are of smaller si ze (on average 1.7 member s per household). In fact, a huge 61 percent of them are singl e person households. Most of zero vehicle households fall in the lower income cate gory (<$25,000 per year), no children category and no workers category. A huge 90.6 percent of them are from urban areas and 61 percent live in apar tment or condominium types of houses. About 37 percent of households own tw o vehicles and 23.5 percent of the households own three or more vehicles. Interestingly, only about 13 percent of households have three or more licensed dr ivers even though 23.6 percent of households report having three or more vehicles. This suggests that the number of vehicles exceeds the number of drivers in many households. In fact, Only 13.5 percen t of households have less number of vehicles than drivers and th e remaining 86.4 percent of households have either equal number (65.1 percent house holds) or more (21.3 percent households) vehicles than the count of drivers they have. At an aggregate level, the average vehicle ownership of 1.9 vehicles per household is mo re than the average number of licensed drivers per household (1.7). These trends indi cate a high vehicle ownership even at the micro level of a person (driver).

PAGE 22

14 Table 4.1 Household Characteristics of the 2001 NHTS Data Characteristic All Households Households With Vehicles Households Without Vehicles Sample Size 26,038 24,615 1,423 Weighted Population 107,368,651 98,878,005 8,490,646 Household Size 2.56 2.63 1.80 1 person 25.82% 22.79% 61% 2 persons 32.63% 33.76% 19.5% 3 persons 16.53% 17.40% 6.4% 4 persons 25.02% 26.06% 13% No. of Children (under 18) 0.67 0.69 0.39 0 children 64.4% 63% 80.7% 1 child 14.6% 15.1% 7.9% 2 children 13.8% 14.5% 5.7% 3+ children 7.3% 7.4% 5.7% No. of Workers 1.31 1.37 0.6 0 workers 22.9% 20.1% 55.3% 1 worker 34.5% 34.6% 33.7% 2 workers 33.7% 35.8% 8.6% 3+ workers 8.9% 9.4% 2.4% No. of Licensed Drivers 1.75 1.86 0.45 0 licensed drivers 5.38% 0.34% 64.1% 1 licensed driver 31.85% 32.11% 28.9% 2 licensed drivers 49.25% 52.99% 5.7% 3 or more drivers 13.52% 14.56% 1.3% Annual Income $25 K or less 29.1% 25.2% 78% $25 K $50 K 33.3% 34.7% 15.5% $50 K $75 K 17.3% 18.4% 3.4% Greater than $75 K 20.3% 21.6% 3.1% Vehicle Ownership 1.90 2.06 NA 0 auto 7.9% 0% NA 1 auto 31.4% 34.1% NA 2 autos 37.1% 40.3% NA 3 autos 23.6% 25.6% NA Dwelling Unit Type Detached single house 63.7% 67% 26.2% Duplex 4.7% 4.7% 4.7% Row House/Town House 3.6% 3.6% 3.6% Apartment/Condo 22% 18.7% 61% Mobile Home/Trailer 5.7% 5.8% 4.1% Residential area type Urban 79.5% 78.6% 90.6% Non-Urban 20.5% 21.4% 9.4%

PAGE 23

15 Table 4.2 Person Characterist ics of the 2001 NHTS Data All Persons All driv ers Primary Drivers Characteristic Workers NonWorkers Workers Nonworkers Workers NonWorkers Sample Size 29,729 30,433 28,701 14,601 25,674 11,247 Population 138,291,467 138,164,039 131,28 5,676 57,954,332 115,273,322 43,485,432 Age (in years) 39.3 31.9 39.7 53.3 40.1 55.1 0-5 years 0% 14.8% 0% 0% 0.0% 0.0% 6-15 years 0.3% 29.7% 0.1% 1.1% 0.0% 0.3% 16-25 years 16.6% 9.0% 15.5% 13.2% 14.3% 9.8% 26-64 years 80% 23.2% 81.4% 46.7% 82.6% 49.1% >64 years 3% 21.1% 3.0% 39% 3.1% 40.7% Sex Male 54.3% 43.2% 54.7% 38.2% 54.9% 39.6% Female 45.7% 56.8% 45.3% 61.8% 45.1% 60.4% Employment Status Full time 77.4% NA 78.4% NA 84.2% NA Part time 15.4% NA 14.7% NA 15.3% NA Multiple Jobs 0.3% NA 0.5% NA 0.5% NA Highest Education Level Highschool/less 39.4% 55.5% 36.6% 52.4% 36.6% 82.3% Some college 30.4% 22.0 % 30.4% 26.1% 31.8% 45.7% Collegegraduate 21.0% 11.9% 21.3% 14.8% 22.3% 26.8% Post graduate 11.6% 5.4% 11.8% 6.7% 12.2% 12.4% Driver Status Driver 94.9% 74.7% 100% 100 100% 100% Primary Driver 83.3% 31.5% 87.8% 75% 100% 100% Daily Travel Trips/day 4.5 3.7 4.61 4.1 4.66 4.3 Work trips 1.4 0.1 1.4 0.1 1.4 0.1 Non-work trips 3.1 3.6 3.2 4.0 3.2 4.2 Miles Traveled 51.2 29.4 52.5 37.2 51.6 38.5 Minutes Traveled 92.2 65.8 93 76.7 91.8 77.9 Note: Workers are those who indicated that they are employed while non -workers indicated that they are not employed. The above table ignores persons who did not respond to specify their worker status.

PAGE 24

16 79.5 percent of the households live in urban residential locations. About 64 percent of the households live in detached single houses while about 22 percent live in apartments or condominiums. Only about 5.4 pe rcent of the households report having no licensed driver and most of those househol ds do not own a vehi cle as well. Sociodemographic attributes of a household and it s residential location characteristics are important in determining its vehicl e ownership level and combination. Table 4.2 provides weighted analysis for an overview of person sociodemographic characteristics. At a further di saggregate level, characteristics of each person in the household and the inter-personal interactions play an important role in utilization patterns of the hous ehold vehicles for travel to perform various activities. Person characteristics are provided in Table 4.2 for various groups. There are a total of about 277 million persons of which about 50 perc ent are workers. Among all the persons, about 68.5 percent are drivers and a bout 57.5 percent are primary drivers of the vehicles owned by the household they belong t o. Essentially about 85% of all drivers are primary drivers. 83.3 percent of workers ar e primary drivers while only 31.5 percent of non-workers are primary drivers, indicating th at the working population that owns most of the vehicles. It could be either due to the purchasing power given by their income and/or other factors like re quirement of driving to/in work. The working population in general, whether primary drivers or not, show very similar characteristics with average age of about 40 years and more than 80 per cent falling in the 26-64 age range. A little over 50 percent of the workers are males and more than 75% percent are employed full time. Nearly all workers are licensed to drive and most of them are primary drivers. On an average each working person makes about 4.5 trips per day, travels about 50 miles per day and spend a total of about 90 minutes on the road. Non-workers on the other hand, show some differences based on primary driver status. Pres umably, the population of all non-workers includes a large proportion of sc hool-going children. He nce, we compare only those non-workers who are drivers to thos e who are primary drivers. There seem to be no huge differences due to most of th e drivers are primary drivers even among nonworkers. As expected, non-workers who declar ed that they are drivers (includes primary drivers) tend to be elderly and retired folks. As such the average age of non-worker primary drivers is about 55 year s and 40 percent of the primar y driver non-worker sample is 65 years or above. A majority of non-worker s are female. In addition, it is found that non-workers exhibit a lower level of education achievement than workers, even among primary drivers. Primary driver non-workers make about 4.3 trips per day and are found to make a few more trips than driver non-wo rkers (4.1). They travel about 38 miles per day and spend about 78 minutes per day on the road. These trends seem to indicate that most of the licensed drivers are primary driv ers and in fact About 84 percent of the licensed drivers are primary drivers. The prim ary drivers’ utilizati on trends are further discussed in the vehicle utilization section. The next section provide s a detailed analysis and discussion of vehicle ownership patterns. 4.3 Vehicle Ownership Unites States is an auto dependent nation. Currently there are a bout 203 million vehicles owned by all households in the U.S, which turns out to be an average of 1.9 vehicles per household. The distribution of such a huge vehi cle fleet definitely warrants closer look

PAGE 25

17 and a careful analysis. The ownership of vehi cles by type can skew the fuel consumption patterns of the fleet, which can have signifi cant implications in the emissions and air quality related issues, thus making it important to analyze the vehi cle ownership patterns by type. This section provides a detailed desc riptive analysis of the household vehicle ownership patterns with an emphasis on the dist inctions by vehicle type. First subsection presents a discussion on the distribution of vehicles by type in the overall vehicle fleet. An analysis of the trends in the recent acquisition of hou seholds by vehicle type is presented in the next subsection. 4.3.1 Vehicle Ownership by Type The widespread ownership of household vehi cles makes it important to understand the trends by type. This section provides such an analysis of household vehicle ownership by type. In order to understand the ownership of vehicles by type, table 4.3 shows the distribution of vehicles by type for different household types and lifecycles. Though a major share of about 56.9 percent of the 203 million vehicles in the U.S. are cars, a considerable portion of about 12 percent are SUVs, 18.2 percent are pickup trucks and 8.9 percent are vans. A 4.1 percent of other types of vehicles in clude medium/heavy trucks, recreational vehicles, motorcycles and other vehicle types. As households go through various lifecycle stages, one can expect the vehicle ownership patterns to differ. The working couple with children population se gment shows the least percentage of cars and highest percentage of vans and SUVs The population of working couple with no children shows the highest perc entage of pickup trucks. Households with children and working status show a greater propensity to own vans and SUVs relative to other single person households, households without childr en, and households with no worker. As expected, the distributions show that househol ds in rural areas are more likely to have pickup trucks than urban area households Multi-vehicle households show higher ownership of vans, SUVs and pickup trucks. Considerable presence of larger vehicles such as vans, SUVs and pickup trucks ma kes it important to understand the household vehicle ownership by number vehicles of each vehicle type. Table 4.3 Vehicle Type Distri bution by Household Type Characteristic Households Cars Vans SUVs Pickup Other All Household Lifecycle Single person 22.8% 69.7% 4.4% 8.3% 14.3% 3.3% 13.8% Single parent 3.9% 65.5% 8.7% 12.1% 10.5% 3.1% 2.6% Working couple/no child 30.9% 58.4% 6.1% 11.3% 19.8% 4.4% 36.4% Working couple/child 32.5% 49.5% 13.2% 14.7% 18.5% 4.1% 38.3% Multiperson/no worker 9.9% 60.4% 9.3% 6.8% 19.1% 4.3% 8.9% Household Residential Area type Urban 78.6% 60.9% 9.1% 12% 14.6% 60.9% 74.8% Rural 21.4% 45.2% 8.6% 11% 29.1% 45.2% 25.2% No of vehicles 1 vehicles 34.1% 77.6% 6.4% 8.2% 7.7% .2% 16.5% 2 vehicles 40.3% 56.9% 10.4% 13.0% 18.5% 1.2% 39.1% >2 vehicles 25.6% 49.3% 8.6% 12.0% 21.9% 8.2% 44.4% Total 98,878,005 56.9% 8.9% 11.8% 18.2% 56.9% 203 Million

PAGE 26

18 The tables 4.4 and 4.5 show the distributions of singl e vehicle households and two vehicle households by the vehicle fleet combinat ion. It can be seen that almost 78 percent of the households with single vehicle have cars. Only 31.2 percent of two vehicle households have only cars. These trends also indicate that non-car ty pes of vehicles are owned more by multi-vehicle households. Table 4.4 Distribution of Single Vehi cle Households by Vehicle Type Households with children Households w/o child All Households Sample 1392 5675 7067 Population 7520806.199 26164018 33684824.32 Car 70.30 79.56 77.5 Van 12.19 4.78 6.4 SUV 11.34 7.31 8.2 Pickup Truck 6.00 8.21 7.7 Other 0.17 0.15 0.2 Table 4.5 Distribution of Two Vehicle Households by Fleet Combination Households with children Households w/o child All Households Sample 4224 6371 10595 Population 17260061.57 22617995.44 39878057 Car-Car 24.18 36.56 31.20 Car-Van 18.52 7.46 12.25 Car-SUV 17.40 13.31 15.08 Car-Pickup 17.16 27.31 22.92 Car-Others 0.33 1.52 1.00 Van-Van 1.22 0.79 0.98 Van-SUV 3.36 0.90 1.97 VanPickup 6.84 3.00 4.66 Van-Others 0.17 0.19 0.18 SUV-SUV 2.21 1.25 1.67 SUVPickup 6.89 4.31 5.43 SUV-Others 0.20 0.37 0.30 PickupPickup 1.25 2.09 1.73 Pickup -Other 0.14 0.78 0.51 Other-Other 0.13 0.16 0.15 4.3.2 Length of Ownership and Age of Vehicles Table 4.7 shows the distribution of the number of years that a vehicle has been held by a household by vehicle type. Essentia lly this is an anal ysis of the time th at has elapsed after the acquisition of the vehicle, based on the respondent’s answer to the question about how long the vehicle was owned. On average, the number of years that a household has held a vehicle is 4.2 years. The average for SUVs is the least at 3.2, followed by vans at 3.8, cars at 4.3 and pickup trucks at 4.8 years. This indicates that SUVs and vans are recently acquired vehicles, while pickup trucks were acquired much former than cars. An

PAGE 27

19 examination of the distribution for each vehicl e type shows that about 45 percent of the vehicles have been held for 1-2 years wh ile another 21.5 percent have been owned for about 3-4 years. About 13.5 per cent of the vehicles in all ca tegories have been held or owned for 9 years or more with the exception of vans and SUVs that tend to be a bit newer with lower percentage of them falling in this category. While it would be certainly interesting to analyze the differences by vehi cle type in total ow nership length from purchase to disposal, the NHTS data cannot be used for such an analysis for it does not provide such information. The length of time a vehicle is held may also be different from the actual age of the vehicle, which is based on its model year. He nce it is useful to discuss the age of the vehicles that households own and utilize. Ta ble 4.8 looks at the distribution of vehicle age (obtained from the model y ear of the vehicle) by vehicle type. The average age is about 9 years for cars, 7.6 years for van, 6.5 years for SUV, and 10.1 years for truck. Thus, vans and SUVs appear to be newer vehicl es in comparison to cars and trucks. It is possible that households like to keep newer, reliable, and good-looking vehicles as their family vehicles and do not mind having older vehicles for personal transportation. More than 55 percent of SUVs and more than 40 pe rcent of vans are in the 0-5 year range. Only about 2-3 percent of these vehicles ar e more than 20 years old. On the other hand, about 5 percent of cars and about 10 percent of trucks are more than 20 years old. Combining table 4.4 with table 4.5 points to an interesting pattern where vans and SUVs tend to be newer vehicles and held/owned for sl ightly fewer years. It appears that people may be purchasing SUVs and vans in the more recent past as they dispose of older vehicles and purchase newer vehicles. This pattern is simply a manifestation of the vehicle acquisition proce ss coupled with people’s vehicle type choice. The table 4.6 shows a cross tabulation of households by age of vehicles. There are about 4.9 percent of the two vehicle households that have both new vehicles and about 10.9 percent with both old vehicles. About 21 percent of two vehicle households have both vehicles with in 5 years of age. Table 4.6 Cross Tabulation of Two Vehicl e Households by Age of Vehicles Age of older vehicle 0-2yrs 3-5yrs 6-10yrs 11+yrs Total Count 1917672 3715976 3997504 2329700 11960852 0-2yrs % of Total 4.9% 9.4% 10.1% 5.9% 30.3% Count 2844618 5812588 4235220 12892426 3-5yrs % of Total 7.2% 14.7% 10.7% 32.7% Count 4220508 6099470 10319978 6-10yrs % of Total 10.7% 15.5% 26.1% Count 4300998 4300998 Age of newer vehicle 11+yrs % of Total 10.9% 10.9% Count 1917672 6560594 14030600 16965388 39474254 Total % of Total 4.9% 16.6% 35.5% 43.0% 100.0%

PAGE 28

20 Table 4.7 Distribution of Number of Ye ars of Ownership by Vehicle Type Years Owned Car Van SUV Pickup Truck Other Total 1-2 years 43.6% 45.2% 55.1% 41.8% 43.1% 44.8% 3-4 years 21.7% 22.9% 22.9% 19.8% 19.7% 21.5% 5-6 years 13.1% 13.9% 9.6% 12.9% 11.4% 12.6% 7-8 years 8% 8.1% 5.5% 7.6% 7.4% 7.6% 9+ years 13.7% 9.9% 6.9% 17.8% 18.5% 13.5% Mean 4.3 3.8 3.2 4.8 5.3 4.2 Total vehicles 57.4% 9.2% 11.9% 17.6% 3.8% 172,199,884 Table 4.8 Distribution of Vehi cle Age by Vehicle Type Age Model year Car Van SUV Pickup Truck Other Total 0-5 years 1997-2002 35% 41.3% 55.5% 34.5% 34% 37.9% 6-10 years 1991-1996 31.2% 33.4% 25% 26.1% 16.9% 29.2% 11-15 years 1985-1990 21.2% 17.6% 11.8% 18.7% 11.2% 19% 16-20 years 1989-1984 7.4 % 5.4% 5.1% 10.8% 15.1% 7.9% 21+ years Before 1984 5.2% 2.3% 2.6% 9.8% 23.7% 6.1% Mean age 9.0 7.6 6.5 10.1 12.7 8.9 Total vehicles 195,854,080 57.2% 8.9% 11.9% 18.2% 3.7% 100% 4.3.3 Recent Vehicle Acquisitions Households add a new vehicle to their fleet either to replace an older vehicle or to accommodate for an increase in their long-term travel requirements. Analysis of the recent vehicle purchases by households helps understand the trends in addition of new vehicles to the fleet. Table 4.9 shows the di stribution of the households by the type of recent vehicle acquired. About 40 percent of the households have owned a new vehicle not former than a year. Only about 15% of the households (with vehicles) have not acquired a new vehicle in past 5 years. This indicates the increasing vehicle acquisition by households in the recent past. Though cars ar e being added most to the vehicle fleet, the share of SUVs, vans and pickup trucks be ing added to the fleet is increasing. Out of all the households with cars as their recent acq uisitions, the percentage that acquired not former than a year is only about 37 percent, while it is 40 percent for vans, 44.7 percent for pickup trucks and 47% for SUVs. The cha nges in the vehicle ac quisitions over time and the recent acquisition trends definitely warrant a closer look at the recent vehicle acquisition patterns. Table 4.9 Distribution of Households by Recently Owned Vehicle Type Newly Owned Vehicle Car Van SUV Pickup Truck Total With in 1 year before 37% 40% 47% 44.7% 39.7% Before 1-2 years 19.6% 23.2% 23.5% 20.4% 20.6% Before 2-5years 24.6% 26.3% 22% 23% 24.1% Before 5+years 18.7% 11.6% 7.6% 12.5% 15.6% Total Households 62.1% 9.5% 13.2% 15.2% 91,448,602

PAGE 29

21 4.3.4 Leased Versus Owned Vehicles One of the developments related to vehicle hol ding, over the past decade, is the leasing of vehicles. There could be many differences in the way people own and use leased vehicles because of differences in the initial payment and the salvage values etc. In addition to this, there could also be ma ny differences in the way peopl e own and use new cars to used cars. Bureau of Labor Statistics reports that leased vehicles are more likely to be luxurious in terms of having air conditioning, automatic transmissions, four wheel drive etc when compared to vehicles owned by consumers. They are also much more likely to be new when obtained by the consumer. Accord ing to their figures from the Consumer Expenditure Survey, 93 percent of the leased vehicles were new where as only 37percent of the owned vehicles were new when obtained by consumers in 1996. Leases now represent as much as 1 in 3 new car acquisitions by consumers. Aizcorbe et. al., 1997 used the Federal Reserve Board’s Survey of Consumer Finances (SCF) and the Bureau of Labor St atistics Consumer Expenditure interview Survey (CE Survey) for the year 1992 to anal yze the vehicle holdings in terms of timing and financial terms of acquisitions and dispos als of vehicles. They present evidence on the growth of auto leasing in recent years. Th eir analysis showed that the percentage of households that leased a vehicle was about 3.0 from SCF and 1.9 from CE survey. Both surveys indicated that the av erage number of leased vehi cles was at 1.1 per household with leased vehicles. The av erage age of leased vehicles was about 1.6 years in 1992 when these surveys were conducted. Their anal ysis of the SCF shows that 43.6 percent of all vehicles owned by househol ds were new when first ac quired. The figure from CE survey is similar at 42.6 percent. Out of the vehicles that were bought used, the average age at acquisition was 6.5 years in SCF and 6.6 years in CE survey. The vehicle acquisition trends may be in fluenced by lease-versus-buy, and newversus-old decisions. Identif ying and controlling for fact ors like lease-versus-buy, and new-versus-old can provide us better insight s into the vehicle ownership and utilization patterns. 4.4 Vehicle Utilization A total of 410,969,163,093 (about 411 billion) pers on trips are taken annually in United States, out of which 79% are on household vehi cles. Household vehicles are utilized for a huge 96 percent of all the 235,506,624,828 (about 235.5 billion) vehicle trips driven annually. An analysis of utilization patterns of household vehicles is important to understand what type of vehicl es are being driven by whom, how much and for what type of trips (length, purpose, occupancy, day of week, time of day). This section describes vehicle utilization patterns in the United States in the context of trip characteristics, daily travel, annual mileage and corresponding differences among the vehicle types. The primary driver allocation and utilization analys is provides insights in to distinctions of vehicles by type based on their driving owne rs. It also gives an understanding of the extent to which primary drivers accomplish their travel needs in the primary vehicle and the extent to which households allocate/devot e vehicles to different household members.

PAGE 30

22 4.4.1 Vehicle Utilization Patterns In order to understand the ex tent of vehicle usage, Table 4.10 provides the distribution of vehicles by type and annual mileage driven (estimated by a household respondent). Vans and SUVs show slightly higher average a nnual mileages than cars and trucks. Table 4.10 Distribution of Annua l Mileage by Vehicle Type Vehicle Type No. of Vehicles Average Miles 0-5000 miles 500110000 miles 1000115000 miles 1500120000 miles >20000 miles Car 56.9% 10696 35.5% 27.0% 19.5% 9.3% 8.7% Van 8.9% 12706 24.9% 26.8% 23.9% 12.3% 11.6% SUV 11.8% 12869 22.3% 25.8% 25.5% 13.7% 12.6% Pickup Truck 18.2% 11644 34.2% 25.0% 18.7% 10.0% 12.1 % Other 4.1% 4920 82.3% 10.6% 3.2% 1.20% 2.7% Total Vehicles 2,03,266,300 11,118 34.5 % 25.9% 19.9% 10% 9.8% Average annual mileage on SUVs is th e highest at about 12,900 miles followed by vans, pickup trucks and cars in that orde r. The differences are very clear in the percentage distributions. More than 10 percent of SUVs, vans and pickup trucks are driven more than 20,000 miles per year. The corresponding figure for ca rs is just 8.7%. Whereas 35.5 percent of cars are driven 05000 miles, only 22 pe rcent of SUVs are driven in this small annual mileage range. These figures indicate that vans, SUVs and pickup trucks are driven more than cars on an annual mileage basis. When the differences are observed among different vehicle types, th e annual mileage by vehi cle type in the all households segment of table 4.11 shows that vans, SUVs, and pickup trucks are driven more per annum than cars. However, a vehicle’s utilization depe nds upon how old it is. Generally older vehicles are driven less comp ared to newer ones, which is quite evident from the annual mileage figures in the table 4.11. Hence it is pos sible that vans and SUVs being relatively new vehicles are being driven more because they are relatively new in the fleet. Given that cars are relatively new in the vehicle fleet it is possible that thes e vehicles are driven less than relatively young vans and SUVs. However, given that pickup trucks are relatively older than cars and are still being driven more, one cannot expect that vans and SUVs may be driven less afte r controlling for age. Hence Table 4.11 compares the annual mileage differences among vehicl e types after control ling for age as well as the vehicles to drivers ratio. The annual mileages of vans and SUVs are higher even after controlling for the age of the vehicle. This trend coul d be observed in all categories of household vehicle availability. Households with larger number of vehicles than the drivers show the least mileage on vehicles of age 6 years or more. This may be because they have these older vehicles as contingency vehi cles that may be used less compared to newer vehicles. Hence, the vehicle availability also plays an important role in the individual vehicle utilization.

PAGE 31

23 Table 4.11 Annual Mileage After Controllin g for Vehicle Availability and Age Annual Mileage on the Vehicle Type (Miles) Vehicles to Drivers Ratio Vehicle Age (years) Car Van SUV P/U Other Total 0 to 2 13877 14343 16733 13368 11000 14240 3 to 5 12998 14972 16579 12706 36263 13784 6 to 10 11005 16018 16432 12675 NA 12347 11 or more 9508 8965 8526 10902 96 9548 < 1 All ages 11439 13694 15144 12236 20551 12188 0 to 2 13725 15257 14661 18154 6641 14717 3 to 5 11868 13906 13729 14671 7203 12754 6 to 10 11201 11699 13033 12171 6802 11548 11 or more 8041 10833 9429 8652 5080 8392 = 1 All ages 10844 12865 13256 12744 6034 11590 0 to 2 14447 17889 15002 17269 5396 14504 3 to 5 13717 15910 14175 13782 7017 13424 6 to 10 11016 11927 11236 11905 6398 11053 11 or more 6882 7600 7407 6755 3087 6569 > 1 All ages 10150 12132 11816 10507 4798 10139 0 to 2 13918 15950 14881 17531 5486 14612 3 to 5 12408 14526 14050 14221 7212 13030 6 to 10 11136 12243 12637 12085 6424 11461 11 or more 7723 9094 8329 7625 3222 7621 All Households All ages 10696 12706 12869 11644 4920 11118 Tables 4.12 and 4.13 provide daily travel characteristics by vehicle type. Table 4.8 offers a detailed analysis of vehicle us age on a daily basis for weekdays (Monday to Friday). The first set of rows shows the to tal utilization of hous ehold vehicle by both household and the non-household members. The di fferences are quite apparent in that vans and SUVs show higher person trip rates than cars and pickup trucks. This is presumably due to the vans and SUVs serving multi-person family-oriented trips contributing to a larger average number of person trips on those vehicles. Indeed, a comparison of average vehicle occupancy (at the bottom of the table) shows that the average occupancy is 2.0 persons per vehicl e in vans and 1.58 persons per vehicle in SUVs. Pickup trucks show the lowest ve hicle occupancy level and correspondingly the lowest person trip rate. As expected, the bulk of the difference in person trip rates on vehicles occurs for non-work trip purposes. In order to control for vehicle occupancy, vehicle trip rates were compared across ve hicle types. After controlling for vehicle occupancy rates, it is found that the diffe rences among vehicle types are less pronounced due to a considerable decrease in non-work trips rates. Nevertheless, vans and SUVs show higher average vehicle tr ip rates and mileage when compared to cars and pickup trucks. Again, non-work trip rates showed th e bulk of difference. As expected, Pickup trucks show much lower non-work trip rates compared to all othe r types of vehicles.

PAGE 32

24 Table 4.12 Daily Travel Characteristic s by Vehicle Type on Weekdays Characteristic Car Van SUV P/U Truck Other Total Total person trips 4.47 6.76 5.42 3.23 0.17 4.42 Work 1.05 1.03 1.22 1.10 0.15 1.05 Non-Work 3.42 5.73 4.19 2.13 0.17 3.37 Total driver trips 3.48 4.08 3.89 2.71 0.28 3.32 Work 1.00 0.95 1.17 1.06 0.15 0.99 Non-Work 2.48 3.13 2.72 1.66 0.13 2.33 VMT (miles) 29.45 31.49 33.87 28.40 4.57 29.04 Work 11.84 10.24 13.79 14.29 3.05 12.03 Non-Work 17.61 21.26 20.09 14.11 1.52 17.01 VTT (Vehicle Time Traveled -min) 60.46 65.40 65.58 51.63 7.82 57.96 Work 22.15 19.96 25.26 24.49 4.89 22.08 Total Utilization of the Vehicle Non-Work 38.31 45.44 40.32 27.15 2.94 35.88 Total vehicle trips 2.94 3.28 3.26 2.40 0.23 2.81 Work 0.88 0.79 1.05 0.96 0.13 0.88 Non-Work 2.05 2.49 2.22 1.44 0.10 1.93 VMT (miles) 24.71 24.43 28.26 24.73 3.66 24.31 Work 10.52 8.18 12.30 12.89 2.25 10.63 Non-Work 14.19 16.25 15.96 11.84 1.41 13.68 VTT (Vehicle Time Traveled -min) 50.97 50.93 55.15 44.91 6.36 48.67 Work 19.60 16.07 22.65 22.09 3.77 19.48 Vehicle Utilization by Primary Driver Non-Work 31.37 34.86 32.50 22.82 2.59 29.19 Total Vehicle trips 3.79 4.38 4.20 3.90 3.78 3.89 Work 1.16 1.11 1.36 1.50 1.46 1.24 Non-Work 2.63 3.27 2.84 2.40 2.32 2.65 VMT (miles) 35.21 36.16 39.99 45.52 51.75 37.38 Work 15.28 12.90 18.07 23.01 25.35 16.66 Non-Work 19.93 23.26 21.92 22.51 26.40 20.72 VTT (Vehicle Time Traveled -min) 69.27 72.84 74.86 79.44 85.30 71.40 Work 27.31 24.53 31.41 38.07 40.22 29.29 Total Primary Driver’s Travel Non-Work 41.96 48.31 43.45 41.37 45.08 42.11 Average Vehicle Occupancy 1.44 2.0 1.58 1.37 1.15 1.51 Average Trip Length (miles) 8.94 8.17 9.12 10.96 15.60 9.21 Average Trip Length (min) 17.9 16.6 17.5 20.0 27.56 18.0 Average Speed (Miles/Min) 0.413 0.404 0.439 0.465 0.483 0.423

PAGE 33

25 Table 4.13 Daily Travel Characteristic s by Vehicle Type on Weekends Characteristic Car Van SUV P/U Truck Other Total Total person trips 4.18 7.08 5.29 3.05 0.51 4.24 Work 0.30 0.26 0.30 0.33 0.08 0.30 Non-Work 3.88 6.81 5.00 2.72 0.43 3.94 Total driver trips 2.83 3.23 3.11 2.11 0.40 2.68 Work 0.27 0.22 0.26 0.31 0.07 0.27 Non-Work 2.56 3.01 2.84 1.80 0.33 2.41 VMT (miles) 24.44 28.02 32.18 22.94 7.54 24.72 Work 2.76 2.39 3.85 4.77 1.03 3.15 Non-Work 21.68 25.63 28.32 18.16 6.51 21.57 VTT (Vehicle Time Traveled -min) 48.01 54.42 58.61 41.04 14.37 47.24 Work 5.25 4.30 6.35 7.40 1.82 5.54 Total Utilization of the Vehicle Non-Work 42.76 50.12 52.26 33.63 12.54 41.70 Total vehicle trips 2.23 2.12 2.30 1.77 0.31 2.07 Work 0.23 0.16 0.24 0.27 0.06 0.23 Non-Work 2.00 1.96 2.06 1.50 0.25 1.84 VMT (miles) 18.74 17.42 23.41 19.21 6.09 18.74 Work 2.29 1.79 3.58 4.25 0.89 2.69 Non-Work 16.46 15.63 19.83 14.96 5.21 16.06 VTT (Vehicle Time Traveled -min) 37.06 34.34 42.89 34.64 10.65 36.01 Work 4.37 3.17 5.86 6.56 1.48 4.71 Vehicle Utilization by Primary Driver Non-Work 32.70 31.17 37.02 28.08 9.17 31.30 Total Vehicle trips 3.05 3.05 3.08 3.27 3.40 3.06 Work 0.30 0.27 0.30 0.42 0.42 0.31 Non-Work 2.75 2.78 2.78 2.85 2.98 2.75 VMT (miles) 27.71 27.80 35.98 38.89 51.94 30.43 Work 3.46 4.74 6.78 6.96 7.47 4.26 Non-Work 24.26 23.06 29.20 31.93 44.47 26.17 VTT (Vehicle Time Traveled -min) 53.31 52.90 63.05 66.74 78.23 56.15 Work 6.29 7.76 9.86 10.23 11.92 7.19 Total Primary Driver’s Travel Non-Work 47.02 45.14 53.19 56.51 66.31 48.97 Average Vehicle Occupancy 1.87 2.89 2.14 1.77 1.21 2.01 Average Trip Length (miles) 9.65 11.10 12.15 11.43 22.21 10.50 Average Trip Length (min) 18.24 19.00 20.86 20.11 38.04 19.11 Average Speed (Miles/Min) 0.419 0.436 0.453 0.471 0.510 0.435

PAGE 34

26 Figure 4.1 Trip Rates by Purpose by Vehicle Type 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40Work and Related School Shopping Family/Personal Social Recreation Serve Passenger Other Return HomeTrip PurposeTrip Rate Car van SUV Pickup Other All Rows at the bottom of table 4.12 have aver age vehicle occupancy and trip lengths on weekdays. On a per trip basis, it is f ound that vans and SUVs exhibit higher vehicle occupancy rates as mentioned earl ier. It is also f ound that the average tr ip length for vans is slightly lower than that for cars, SUVs and pickup trucks, presumably due to the higher non-work trip rate in these ve hicle types. Generally on week days, non-work trips tend to be of shorter length than work trips due to the time constraints. More over, many nonwork trips on a weekday tend to be shorter li nks of a trip chain. Th at trend was seen both in duration and distance. The average trip leng ths for SUVs and pickup trucks are slightly on a higher side compared to th at of cars. The average trip durations follow similar trends as trip lengths except that the SUVs show sm aller trip durations than that of cars while they show higher trip distances than cars. This is perhaps due to their higher average speed. Table 4.13 provides the same kind of information for weekends. Many of the trends seen on weekdays are once again seen on weekend days. As expected, all vehicles are used substantially less for the work purpose on weekend days. Also, it is seen that the

PAGE 35

27 difference between SUVs and vans on the on e hand and cars and pickup trucks on the other is further amplified on weekends. As many weekend ac tivities tend to be household oriented and joint activities, it is likely that va ns and SUVs are used more relative to cars and pickup trucks on weekends It is found that vehicle o ccupancy rates on weekends are so high relative to weekdays that, although th e number of person trips on weekends and weekdays are quite similar, the number of vehi cle trips on weekends is less than that on weekdays. Total vehicle VMT and travel dura tion are substantially larger for SUVs and vans on weekends. Unlike on weekdays, the aver age vehicle trip length of vans is higher than that of cars because the non-work trips on a weekend tend to be longer due to lower time constraints. However, it is primarily the number of trips and not the trip lengths that contribute to the differences across vehi cle types in daily duration and VMT on weekends. (similar to weekdays). Figure 4.1 shows the daily vehicle trip ra tes by purpose for each vehicle type (for all days of a week). The graph indicates that vans and SUVs have the highest trip rates for shopping, family/personal, social recrea tion and serve passenger trip purposes. Work and related trip purpose trip are carried the most by SUVs and pickup trucks. The trends indicate that vans and SUVs are used the most for non-wo rk kind of trip purposes because of their use as personal vehicles a nd family vehicles. SUVs are also used for work trips indicating their highe r use for all trip purposes. 4.4.2 Primary Driver Vehicle Allocation and Utilization Patterns Vehicles are allocated to drivers for their use to pe rform household and individual activities. Only about 12 percent of all the household vehicles are not allocated to any specific primary driver. Rest of the 88 percent of the househol d vehicles are allocated to primary drivers. Primary driver characteristics are likely to be play ing an important role in shaping the differences across the vehicle ty pes. Hence it is useful to understand who primarily drives the vehicles in particular Table 4.14 provides such an analysis through description of primary driver characteristics by vehicle type. In this table, the characteristics of those who reported themselv es as primary drivers for different vehicle types are summarized. Although th e average age of the differe nt primary driver groups are similar across vehicle types, it is important to note the differences in age distributions. 17 percent of the car primary drivers are in the 16-25 years categor y, the corresponding percentage for vans is only 3.7 percent and th at for SUVs and trucks is about 11 percent and 12 percent respectively. While about 85 pe rcent of the van and SUV primary drivers are in the 26-64 year age group, only 66 percen t of car primary drivers are in this age group. In general, the elderly show a greater propensity to be primar y drivers for cars as opposed to larger vans, SUVs, and trucks. In terestingly, it was found that the majority of primary drivers for cars, vans, and SUVs are female. The big difference in gender distribution is noted for trucks where 88 percent of the primary drivers are male. The van shows a slightly higher percentage of females as primar y drivers at about 60 percent as opposed to cars at about 56 percent and SUVs at about 53 percent. In general, a larger percentage of SUV and pickup truck primary drivers tend to be workers. About threefourths of these drivers are workers; the corres ponding percentage for cars and vans is about two-thirds. While the highest educat ion level distribution shows similar trends across car, van, and SUV primary drivers, it is found that the pickup truck driver group is

PAGE 36

28 less educated, possibly suggesting that they may be using the truck for blue-collar service occupations. The SUV primary driver group exhibits the highest education level distribution. Many of these va riables are correlated with on e another. As SUVs cost more, individuals must be in higher paying jobs and have higher incomes to afford SUVs. Individuals in higher paying j obs and having higher income are likely to have higher education levels as well. Overall, there se em to be some clear patterns of vehicle allocation and choice that em erge from this table. Table 4.14 Primary Driver’s Socio-Demograp hic Characteristics by Vehicle Type Characteristic Car Van SUV Pickup Truck Other Total Vehicles 115,723,09318,186,47123,949 ,10737,054,676 8,090,527 203,266,300 Primary drivers 98,370,224 15,771,17021,933,75331,711,971 5,910,345 158,795,853 Age Average age (yrs) 45.04 45.62 41.64 44.73 44.97 44.24 0-5 years 0% 0% 0% 0% 0% 0% 6-15 years 0.07% 0% 0.09% 0.05% 1.21% 0.11% 16-25 years 17.18% 3.68% 11.08% 12.09% 9.23% 14.76% 26-64 years 66.18% 85.26% 83.36% 75.98% 78.16% 71.74% >=65 years 16.57% 11.07% 5.47% 11.88% 11.39% 13.39% Gender Female 56.03% 60.81% 53.07% 11.77% 8.78% 49.32% Male 43.97% 39.19% 46.93% 88.23% 91.22% 50.68% Employment Status Unemployed 29.37% 33.66% 22.75% 22.17% 21.59% 27.38% Employed 70.63% 66.34% 77.25% 77.83% 78.41% 72.62% Education High school 37.39% 37.39% 31.80% 52.43% 44.99% 39.28% Some college 29.98% 30.59% 30.56% 29.17% 31.72% 29.94% College graduate 20.91% 20.59% 25.30% 12.91% 16.63% 20.14% Post graduate 11.72% 11.42% 12.34% 5.49% 6.66% 10.64% Out of about 235 billion vehicle trips that are made annually in the United States, about 96 percent (225.5 billion) vehicle trips are driven on household vehicles. On an average 80 percent of these driver trips on household vehicles ca rry primary drivers. Remaining 20 percent of the household vehicle tr ips are with out the primary diver in the vehicle. These figures give an idea of the ex tent to which househol d vehicles are being used for carrying primary drivers. Corresponding percentages for each vehicle type are given in table 4.15, which shows that pickup trucks are above aver age with 85.5 percent of the trips carrying primary drivers. Cars ar e closer to the averag e figure with 80.4% of the trips carrying primary drivers. Vans and SUVs are below the average with vans being

PAGE 37

29 the lowest at only 72.7 percent of the trips ca rrying primary drivers and SUVs at a below average figure of 78.5 percent carrying primary drivers. Another way to look at the primary driver’s utilization of the vehicle is the extent of driving done by the primary driver when he/s he was in his/her vehicle on the trip. This gives an idea of the extent to which driv ers other than primary drivers are driving vehicles even when the primary driver is actu ally present in the vehicle. In other words, this is an analysis of the role of primary dr iver in multi-occupant trips. Out of all the 411 billion trips made annually in the Unite d States, 193 billion trips are done by primary drivers in their primary vehicl es and 67.5 billion trips of th em are multi-occupant trips. 82.5 percent of these trips are driven by the primary drivers them selves, implying that the remaining 17.5 percent of the multi-occupant trips that carried primary drivers were driven by other drivers. The corresponding percentages of prim ary driver’s role in multioccupant trips for each vehicle type are s hown in figure 4.15. Pickup trucks have the highest percentage (91.9) of multi-occupant trip s that carried primary driver were driven by the primary driver. Cars are 82.1 percent (c lose to the all vehicles figure of 82.5) and SUVs are at 80 percent (little lower than 82.5) The percentage of vans trips driven by primary driver is only 79.6 indicating that the remaining 20.4 percent of multi-occupant van trips that carried primary driver were driven by some on e other than primary driver. The primary driver vehicle utilization analys is from table 4.15 indicates that pickup trucks are allocated the most (shared the least) to the primary drivers, while vans are shared the most. This is perhaps due to higher tendency of fe males and unemployed persons being the primary drivers of vans than any other vehicle type. Vans are used more on weekends and for high occupancy trip s and it is possible th at the household head (mostly an employed male) drives the vehicle irrespective of the primary driver status. Table 4.15 Vehicle Utilization by Primary Drivers Vehicle Type Percentage of household vehicle trips that carried primary drivers Primary driver’s role in multi-occupant trips (Percentage driven by primary driver) Car 80.4% 82.1% Van 72.7% 79.6% SUV 78.5% 80% P/U truck 85.5% 91.9% All vehicles 80% 82.5% Total Trips 225,626,347,382 67,500,040,642 Tables 4.12 and 4.13 provide primary driver utilization analysis on a daily basis. Table 4.12 offers a comparison of total utilizati on of the vehicle to the primary driver’s usage to gain an understandi ng of the extent to which primary drivers are using the vehicles for weekdays on a daily basis. Primar y driver’s total travel is also compared to his/her vehicle utiliza tion. The first set of rows shows the total utilization of household vehicle by both household and the non-househol d members. The second set of rows shows the primary driver’s util ization of their vehicles on weekdays. When the number of driver trips of primary drivers on their vehicle is compared to the total driver trips carried

PAGE 38

30 by the vehicle, most of the vehicle trips car ried by the vehicles are those of primary drivers. Overall, the vehicles are driven, on average, about 30 mile s per day with about 60 percent of the mileage for non-work travel. Of the 30 miles that a vehicle is driven, it appears that the primary driver accounts for a large percentage of about 84 percent. A similar trend is found with re spect to travel duration. Th e vehicles average about 58 minutes in travel time per day. The primary driver accounts for about 85 percent of that time. These findings show that vehicle alloca tion in a household tends to be quite strong and that people rarely deviate from the vehi cle allocation pattern (a t least on weekdays). However, The primary driver’s utilization, in terms of vehicle mileage, varies from about 77% for vans to about 87% for pi ckup trucks. This show s that pickup trucks are being strongly allocated to primary driv ers and vans show comparatively weaker allocation to a single person. The third set of characteristics rev eals the total primary driver’s travel on weekdays. In general, pr imary drivers of vans and SUVs are found to have taken/driven slightly higher number of tr ips. When the total vehicle trips of primary driver are compared to the trips made on th eir vehicle (from primary driver’s vehicle utilization) it is found that on an average not more than 70 percent of the primary drivers’ travel (mileage) is accomplished using their ow n vehicle. This may be because either they are primary drivers of more than one vehicle or they are using the vehicle that they are not primary drivers of. This analysis has indi cated a strong allocation of vehicles to their primary drivers on weekdays. One could exp ect a weaker allocation of vehicles to primary drivers on weekends due to lower constraints of work. Hence, same kind of analysis is done for weekends in table 4.13. It is found that the primary driver once again plays a large role in using the vehicle, but not as much as it was on weekdays. The primary driver accounts for about 75 percent of the total daily mileage on his vehicle on a weekend day compared to the 83 percent on weekdays. The percentage by vehicle type varies from about 62 percent for vans to 83 percent for pickup trucks. The primary driver utilization patterns suggested very strong allocation of vehicles, even on weekends, to an individual rather than shar ing that one could expect in a household. Much of shared usage of a vehicle appears to be in the form of shared riding but the primary driver did most of the driving on the vehicle. While it would certainly be interesting to understand the factors or constr ains that imply these strong allocations even on weekends, it is beyond the scope of this study. In summary, these descriptives have suggested the possibility of differences in ownership and utilization patterns of cars, SUVs, vans and pickup trucks. The forthcoming models will explore the trends of vehicle ownership, fleet combination and utilizati on in detail in a multivariate setting.

PAGE 39

31 CHAPTER 5 MODELING METHODOLOGY This section describes the underlying mathematical framework, methodology of estimation and application of the model struct ures of structural equations modeling and discrete choice modeling frameworks that are used for the analysis. 5.1 Structural Equations Modeling Structural Equations Model (SEM) systems have n been widely adopted in activity pattern and travel behavior Research and vehicle ownership and utiliza tion patterns. (Golob. 2003). SEM offers a very high extent of flex ibility in terms of handling simultaneous multivariate out comes, in other words, multiple inter-dependent exogenous variables. 5.1.1 Structural Equations Representation G GB X Y Y Y . . .1 1 A typical structural equations model (with ‘G’ number of endogenous variables) is defined by a matrix equation system as shown in the equation above. This can be rewritten as, YBYX (or) YIBX ()()1 where Y is a column vector of endogenous variables, B is parameters matrix associated with right-hand-side endogenous variables, X is a column vector of exogenous variables, is a matrix of parameters associ ated with exogenous variables, and is a column vector of error terms a ssociated with the endogenous variables. 5.1.2 Estimation Single equations estimation methods like the OLS (Ordinary Least Squares), ILS (Indirect least Squares), 2SLS (Two-Stage least Squares) canno t be used for the estimation of simultaneous equation’s parameters. Even the LIML (Limited Information Maximum likelihood) method of estimation cannot be used for these are not suitable to deal with limited dependent variables with different underlying distributions. Though the FIML (Full Information Maximum Likelihood) is an ideal method of estimation of the

PAGE 40

32 parameters of for system of equations, this tends to be computati onally intensive hence making larger model systems almost impracti cal to estimate. More over, even a small mis-specification in any part of the model syst em affects all of the estimates rather than just a related few. SEM uses covariance-based structural analysis for the parameter estimation. Essentially, the difference between the samp le covariances and the model predicted covariances are minimized (Bollen, 1989). Th e fundamental hypothesis for this approach is that the covariance matrix of observed variab les is a function of a set of parameters as shown below: = () where, is the population covariance matrix of observed variables, is a vector that contains the model parameters, and () is the covariance matrix written as a function of The above equation implies that each element of the covariance matr ix is a function of one or more model parameters. The relation of to () is basic to an understanding of identification, estimation, and assessmen ts of model fit. The matrix () has three components, namely, the covariance matrix of Y the covariance matrix of X with Y and the covariance matrix of X Let = covariance matrix of X and = covariance matrix of Then it can be shown that: () ()()()() () IBIBIB IB111 1 After ensuring that the specified model syst em is mathematically identified (Bollen, (1989), Judge et al (1985) and Johnston et al (1997).), The unknown parameters in B, , and are estimated so that th e implied covariance matrix, is as close as possible to the sample covariance matrix, S. In order to achieve this, a fitting function F(S, ( )) which is to be minimized is defined. Th e fitting function has the following properties: F(S, ( )) is a scalar; F(S, ( )) 0; F(S, ( )) = 0 if and only if ( ) = S, and F(S, ( )) is continuous in S and ( ). 5.1.3 Asymptotically Distribution Free – Weighted Least Squares Estimation There are four widely used fitting met hods; Maximum Likelihood (ML), Unweighted Least Squares (ULS), Generalized Least Squa res (GLS) and Asymptotically DistributionFree (ADF) fitting method. Any of the above methods of fitting can be used for the estimation for obtaining consistent paramete rs. However, the possibility that the endogenous variables specified in the system may have different underlying theoretical distributions precludes the use of ML, UL S and WLS fitting functions. Hence ADF-WLS method is employed to get consistent and as ymptotically efficient parameter estimates (Golob 2003, Amos User’s Guide, 1997)

PAGE 41

33 The ADF-WLS estimation method proceeds in three distinct step s. First, it is assumed that each observed endogenous variable is generated by an unobserved normally distributed latent variable. If the latent variable is greater than a censoring level, it is observed; otherwise the censoring level is observed. Each latent variable is assumed to be conditional on the other variables in the sy stem. The problem is to determine the conditional unknown mean and variance of each censored latent variable. This can be done using the Tobit model. An appropria te maximum likelihood estimation procedure for the Tobit model is described in Maddala (1983). Second, estimates of the correlations between the latent censored endogenous variable s, and the correlations between each of the latent variables and the continuous e xogenous variables in the system are derived. Finally, parameters of the structural equati on model are estimated such that the modelimplied correlation matrix is as close as possible to the sample correlation matrix, where the sample correlation matrix is determined in the previous steps. The fitting function is then: FWLS = [s ( )]’ W-1[s – ( )] where, s is a vector of censored correlation coefficients for all pairs of endogenous and exogenous variables, ( ) is a vector of model-implied co rrelations for the same variable pairs, and W is a positive-definite weight matrix. Minimizing FWLS implies that the parameter estimates are those that minimize th e weighted sum of squa red deviations of s from ( ). This is analogous to we ighted least squares regre ssion, but here the observed and predicted values are variances and covari ances rather than raw observations. The best choice of the weight matrix is a consistent estimator of the asymptotic covariance matrix of s: W = ACOV(sij, sgh) Under very general conditions: ) ( 1gh ij ijghs s s N W is a consistent estimator, where sijgh denotes the fourth-order moments of the variables around their means, and sij and sgh denote covariances. Browne (1998) demonstrated that FWLS with such a weight matrix will yield consistent estimates, which are asymptotically efficient with correct parameter test statisti cs. These properties ho ld for very general conditions, and consequently such estimators are known as arbitrary distribution function, or asymptotically distribution free (ADF) es timators. ADF-WLS estimators are available in several structural equation model estima tion packages including AMOS (Arbuckle, 2000) and LISREL (Joreskog et. al., 1993). 5.1.4 Evaluation Many criteria are available for assessing overa ll goodness-of-fit of a Structural Equations Model. Most of these evaluati on criteria are based on the chisquare statistic given by the product of the optimized. Fitting functi on and the sample size (Golob 2003). The asymptotic distribution of (N-1) FADF is 2 distribution with {(1/2) (G+K) (G+K+1)}-t degrees of freedom, where t is the number of free paramete rs. The null hypothesis of the chi-square test is H0 = = ( ). This implies that the overidentifying restrictions for the model are correct. Rejection of H0 suggests that at least one re striction is in error so that ( ). The objective is to attain a non-si gnificant model chi-square, since the

PAGE 42

34 statistic measures the difference between the observed and reproduced variancecovariance matrices. The level of statistical significance indicates the probability that the differences between the two matrices are due to sampling variation. One rule of thumb for good fit is that the chi-squa re should be less than two times its degrees of freedom (Ullman, 1996). However, for large samples it ma y be very difficult to find a model that cannot be rejected due to the direct influe nce of sample size (Golob 2003). For such large samples, Critical N (Hoetler, 1983) gives the sample size for which the chi-square value would correspond to p = 0.05. A rule of thumb is that cr itical N should be greater than 200 for an acceptable model (Tanaka, 1987). One of the several other ways to calibrate the match of the variance-covariance matrices is the Goodness-of-Fit Index proposed by Joreskog., et. al. (1986). The goodness of fit index measures the rela tive amount of the variances a nd covariances in S that are predicted by The Adjusted Goodness-of-Fit Index (AGFI) adjusts for the degrees of freedom of a model relative to the number of variables. Both the indices reach a maximum of one when S = 5.2 Multinomial Logit Model This section provides an overview of the methodology of multinomial logit models that are most widely used for modeling disc rete choice phenomenon in transportation, economics, marketing and many other fields. 5.2.1 Random Utility Approach Discrete choice models are based on random utility maximization hypothesis. The random utility theory which assumes that the decision-maker’s preference for an alternative is captured by the value of an i ndex, called utility. A decision-maker selects the alternative from the choice set that has th e highest utility value. Probability of choice ‘i’ is equal to the probability th at the utility of alternative ‘i’ is greater than or equal to the utilities of all other altern atives in the choice set. (or) P(i|C n ) = Pr [U in U jn all j C n ] where, C n is the set of alternatives available for the nth choice maker (choice set). Random utility models assume that decision-makers have perfect discriminating capability. However, the analyst will have limited information about an individual’s utility level. The uncertainty introduced by intr oducing an error term in the utility of each alternative. Hence, the u tility of an alternative ‘Ui’ is split into a deterministic term ‘Vi’ and a random term ‘ i’. Then, P(i|C n ) = Pr [V in + in V jn + jn all j C n ] The deterministic utility Vin expressed as linear function of explanatory variables (Xk) is given by: in k ki in i i inX X V ...1 1 0 Alternative distributional assu mptions about the joint probability distribution of the full set of disturbances (error te rms) yield different probabilis tic choice models. Assumption that the disturbances are ‘Gumbel’ distribut ed leads to the multinomial logit model with the ‘Independence of Irrelevant Alternatives’ (IIA) property. Multinomial logit model is

PAGE 43

35 the most popular form of discrete choice mode l in practical applica tions. It can be shown that the probabil ity of individual n choosing the alternative i is given by logit formula n jn inC j V V ne e i P ) ( (or) n jn inC j x x ne e i P' ') ( 5.2.2 Estimation Maximum Likelihood method is used to esti mate the parameters (coefficients) of multinomial logit model. Maximum likelihood es timates are the value of the parameters for which the observed sample is most likely to have occurred. The likelihood function for a genera l multinomial choice model is N nC i y nn ini P L1 *) ( where, N denotes the sample size and, yin = 1 if choice maker n chooses alternative ‘i’ = 0 otherwise Where, for the linear in parameters model: n jn inC j x x ne e i P' ') ( Taking the logarithm provides th e log-likelihood function as: We can then solve to estimate the parameters, which maximize L. 5.2.3 Evaluation Likelihood ratio test is used to compare the specified mode l with the baseline model, which assumes that the probability of an al ternative being chosen by all the decision makers is equal its the market share. This is the case of all coefficients in the model except the constants being equal to zero. Under the null hypothesis that the all the coefficients, except the constants are zero, the statistic given by: –2[L(c) – L(ˆ )] is 2 distributed with K-J+1 degrees of freedom. Where, L(c) is the Log-likelihood at market share L(ˆ ) is the Log-likelihood at convergence of the specified model K is the number of parameters J is the number of alternatives. L(c) can be obtained by estimating a model w ith J-1 alternative specific constants: J i i iN N N c L1ln ) ( N nC iC j x in innn jne x y L1ln

PAGE 44

36 One can also look at the goodne ss-of-fit measure given by: ) ( ) ˆ ( 12c L K L This measure does not have any statis tical interpretation unlike that of R 2 measure in linear regression. This measure is called the adjusted Rho-Square and give s an idea about the improvement of the log-likelihood functi on from the base model. Having provided a brief methodology of the models the next sect ion describes the model estimation results.

PAGE 45

37 CHAPTER 6 MODELS OF VEHICLE OWNERSHIP AND UTILIZATION 6.1 Background This section is devoted to models of hous ehold vehicle ownershi p and utilization and vehicle type choice behavior of households and drivers in the c ontext of household and person attributes. First subs ection presents a Structural Equations Model of vehicle ownership and utilization in a unified framewor k. This joint model of vehicle ownership, combination and utilization is definitely advantageous over individual models of vehicle fleet combination and utilization. The next subsection has a multinomial logit model of vehicle type chosen by the households in thei r recent purchases. This model incorporates previous vehicle ownership le vel and type as explanator y factors, which are very important along with the socio-demographics in the choice making behavior of recent vehicle purchases. The last subsection presents a multinomial logit model of the type of vehicle chosen by the driver for a trip. These mo dels are important in that they provide an interpretation of the eff ects of each factor in a ceteris paribus situation, hence separating the correlated factors to avoid confounding effects. Apart from confirming certain common perceptions about various vehicle types, the models also bring out some subtle differences that cannot be identified through simple descriptive analysis. 6.2 Structural Equations Model of Vehicle Ownership and Daily Utilization The purpose of this model is to understand the effect of socio-demographic attributes on the household vehicle ownership and utilization patterns in a unified framework. The hypothesis of the model structure is shown figure 6.1, according to which the vehicle ownership and the fleet combination of a household is explained by its sociodemographic attributes. The vehicle utili zation patterns are explained by the sociodemographics as well as the vehicle ow nership and the fleet combination of the household. Analyzing the vehicle utilization patterns along with the ownership trends enables us to control for the vehicle owners hip level and fleet combination, which can definitely influence the utilization patterns. This joint model system of model system of household vehicle ownership and utilization es timated in a simultaneous equations setting is definitely advantageous over analyzing the vehicle use patterns conditional upon the vehicle fleet. The exogenous variables in this model sy stem are socio-demogr aphics attributes; number of adults in the household, number of children, number of working persons, annual household income, indicators for urba n/rural area location of the household, housing type and the weekend travel day indi cator. These exogenous variables are chosen based on previous literature review, prelim inary exploratory analysis and judgment. Vehicle ownership, fleet combination and utilization are endoge nous in the system. Vehicle ownership (endogenous) is taken as the number of vehicles owned by the

PAGE 46

38 household. Vehicle fleet combina tion or the ownership by type is the number of vehicles of each type owned by the household. The tota l daily Vehicle Miles Traveled (VMT) and the VMT on vehicles of each type give the vehicle utilization of the household. The socio-demographics are taken in as 6 exoge nous variables; number of adults in the household, number of children, number of workers, number of drivers, annual household income, dummy variables for urban area locati on and detached single type of housing. The error terms of the equations for number of vehicles of each type are correlated among each other and similarly the error terms of the equations for the daily VMT on vehicles of each type are also correlated. The endogenous variables specified in th e model have diffe rent underlying theoretical distributions, t hus precluding the possibility of the normal distribution assumption of the variables. Hence, this pa per employs a structural equations estimation methodology that accommodates skewed non-normal endogenous variables. The ADFWLS (Asymptotically Distribution-Free – Wei ghted Least Squares) method available in the Structural Equations Modeling Framework is used for the estimation, which reduces to Generalized Least Squares (GLS) estimation in the absence of non-normality and Maximum Likelihood (ML) estimation in the absence of heteroskedastcity or autocorrelation. The ADF-WLS procedure corrects for any source of non-normality like skewness, kurtosis and censoring of the vari able distributions. Essentially the results presented use consistent estimators that are asymptotically efficient and provide corresponding test statistics that are asymptoti cally valid. However, the software package called AMOS (Analysis of Moment Structures) used for the estimation does not take into take into consideration the pot ential ordered-discrete natu re of some of the endogenous variables (vehicle ownershi p) which is possible through extended forms of ADF-WLS estimation available in other packages availa ble Hence, there is an implicit assumption here that all the endogenous variables are continuous. The proposed model is estimated for the sa mple of households having at least one vehicle. The tables show the direct effects and total effects that constitute relationships among variables. The model estimates showed excellent goodness-of-fit measures with the2statistic indicating that the model cannot be rejected with a high degree of confidence (95 percent or higher) The critical N is above 200 hence avoiding the pitfall of not rejecting the hypothesis due to larger sample size. The Goodness-of-fit index (GFI) equal to unity and the Adjusted GFI is e qual to 0.098. Thus the model framework is capable of capturing key relationships among the variables. The indications provided by the model are consistent with the expect ations and are otherwise plausible. The regression coefficients significant at the 95 percent confidence level are retained. These regression coefficients show the direct effect of one variable on ot her and hence are also called direct effects. An arrow linking the tw o variables in the pa th diagram depicts a direct effect. On the other hand, an indirect effect is one where a variable influences another variable through a mediating variable In some cases, a variable may have both a direct and an indirect effect on another variab le. Then the total effect is the sum of the direct and indirect effects. The results presen ted for the model are from the final validated model after a series of exploratory analyses. The discussion is divide d into two separate sections for vehicle owne rship and utilization.

PAGE 47

396.2.1 Vehicle Ownership The estimates of direct effects (Table 6.1) show that vehicle ow nership is higher in households with higher income, or larger nu mber of workers, or larger number of licensed drivers. Households with larger num ber of adults show higher vehicle ownership while the households with more children tend to have lower vehicle ownership. Households in urban areas tend to have lesser number of ve hicles while h ouseholds living in detached single houses tend to have highe r vehicle ownership. Both, housing type and residential location, have been described in the literature as a measure of the activity travel environment in the immediate neighborh ood and the area the ho use is situated in respectively. In other words, these variable s can also be interpreted as surrogates for urban form, neighborhood design and parking av ailability. Househol ds in urban areas tend to have higher car ownership than any ot her vehicle type while households in urban areas show the least tendency to own pickup trucks with a negative coefficient indicating that rural area households tend to own mo re pickup trucks. SUV ownership is not significantly affected by the urban/rural area location of the househol d. Surprisingly, Van ownership is negatively affected by the urba n area location of the household. Detached single type of housing favors higher pickup truck ownership followed by vans and SUVs. They show a negative effect on the car ow nership. This may be because of the availability of parking space with detached houses that enables the ownership of larger vehicles like pickup trucks, Vans and SUVs. Households with higher annual income te nd to own more SUVs and less number of pickup trucks, which is consistent with expectations. Household income doesn’t show a significant effect on the ownership of cars and vans. Nu mber of adults in the household does not show a significant effect on the car ow nership while it shows a positive effect on van ownership and a negative effect on SUV ow nership and pickup truc k ownership. This indicates that larger households show a higher tendency to own vans rather than SUVs and pickup trucks. Number of children in the household has a positive effect on the van ownership and SUV ownership while it has a negative effect on the car ownership and pickup truck ownership. This indicates the ow nership of vans and SUVs for their use as family vehicles. Households with higher num ber of workers show higher ownership of SUVs and pickup trucks and a lower owners hip of cars and vans. This indicates the ownership of SUVs and pickup tr ucks for their use by workers. Households with more number of drivers tend to have higher car ownership and lower van and pickup truck ownership. Total vehicle ownership also sh ows significant effect on the individual ownership of each vehicle type. The coefficien ts are positive and are less than unity as expected. These coefficients represent how an additional vehicle ownership contributes to the ownership of each vehicle type. The coeffi cients indicate that, keeping all else the same, households with larger number of vehi cles have more cars followed by pickup trucks, SUVs and vans in that order. In othe r words, any extra vehi cle in the household is most likely to be a car followed by a pic kup truck, SUV and a van in that order. In summary, cars are owned the most by households in urban areas, and households with more number of drivers. Mo re over, each extra vehicle in a household is most likely to be a car. Households in ur ban areas and households living in detached single houses, with larger number of adults and children have higher van ownership. Households with larger number of workers and drivers tend to have lower van ownership.

PAGE 48

40 Each extra vehicle in a household is least li kely to be a van, indicating the smaller presence of vans in the households. Larger SUV ownership is in th e households of higher income, households living in detached houses and households with more children and working people. Pickup trucks are owned more by rural ar ea households and households living in detached single houses and households with lower income, smaller size (adults and children), smaller number of drivers and larger number of workers. The intercepts indicate that, keeping everything else the same ; the ownership of pickup trucks is most prevalent followed by cars and the ownershi p of vans and SUVs in less prevalent. 6.2.2 Vehicle Utilization The exogenous variables explaining vehicl e utilization include an additional dummy variable for weekend day for the utilization is on a daily basis. The direct effect of the weekend day on overall household VMT is ne gative indicating a lower household VMT on a weekend day compared to that of a weekday. However, the weekend effect on the VMT of the vans of a household is positive and that of pickup trucks is negative indicating that vans are used more and pickup trucks are used less on weekends. This is quite consistent with the expected results because households te nd to make high occupancy trips of recreation and other pur poses on more on week ends causing a higher mileage on vans and lower mileage on pickup trucks. Urban area households show lower total VM T compared to those of rural areas. The total effects (Table 6.2) of the urban area dummy variable on the VMT by each type vehicle indicate that all type s of vehicles; cars, SUVS, vans, and pickups is lower in urban areas. This may be because of the lower overall mileage by urban area households. But surprisingly, the urban area households show a positive direct effect on the VMT by pickup trucks. This indicates that even after controlling for all other relevant factors, including total household VMT and vehicle ow nership and fleet combination, an urban area household tends to have more miles trav eled on its pickup trucks compared to a household in a rural location. Households in urban areas show lower VMT on cars compared to those of rural area households, while the urban /rural area effect is not significant in the case of vans and SUVs. Hous eholds living in detached single houses do not show significant effect on total daily VMT. However, the total effect of this variable on the daily VMT is positive due to the higher vehicle ownership of the households living in detached single houses. The direct eff ects of this variable on VMT by individual vehicle type show lower car VMT and highe r van VMT. The corresponding total effects indicate higher daily VMT on vans, SUVs and pick trucks and lower daily VMT on cars. As expected, the direct effects of the higher income households show higher daily total VMT, van VMT and SUV VMT. Number of adults doesn’t have any significant effect either on the total VMT or on the individual vehicle type VMT. The number of children shows a positive effect on the total VMT of the household while it does not have any significant effect on the VMT by each ve hicle type. However, number of adults and children directly affect the vehicle ownership and hence indirectly a ffect the utilization. Number of workers in the household shows a positive effect on the total VMT of the household while it shows a nega tive effect on the VMT on vans of the household. This is perhaps due to every worker in the house hold requiring a separate vehicle for work. Number of drivers shows a pos itive effect on the total VMT and the pickup truck VMT of

PAGE 49

41 the household. Naturally, households with la rger number of vehicl es show higher total VMT but lower car VMT and pickup truck VMT. These households travel more on their vans and SUVs. However, this model doesn’t control for the age of vehicles. Vans and SUVs may be relatively new in the vehicle fleet and henc e being used more. The total VMT and the split among the vehicle types can also be affected by the fleet combination owned by the household. After controlling fo r the number of vehicles owned by a household, cars do not significan tly increase the total dail y vehicle mileage of the household on the other hand vans, SUVS and pi ckup trucks have a positive effect on the household VMT. This indicates lager amount of travel by households with SUVs, vans and pickup trucks even after all other factors are controlled for. Same argument holds for the higher positive effect of number of vans, SUVs and pickup trucks on their own VMTs while number of cars has a smaller positive effect on the car VMT of a household. As expected, the ownership of each vehicle t ype shows a negative effect on the VMT of other vehicles. In summary, the VMT on cars is lower on households living in detached houses and urban area households. However, owing to a larger presence of cars in the households, the total VMT on cars tends to be higher. VMT on vans is higher on weekends and in households living in detach ed houses. Households with higher number of workers tend have lower VMT on vans. Hi gher annual income increases the van VMT and also the SUV VMT of a household. Pic kup truck VMT is higher in urban households and households with larger number of driver s, while it is lower on weekends. The intercepts indicate that, keeping every thing else the same, the VMT on vans is lower than that of any other vehi cle type. Vans, SUVs and pic kup trucks show higher VMT on them selves as well as total household VMT when compared to that of cars.

PAGE 50

42Table 6.1 Direct Effects, Structural Equation s Model of Vehicle Ownership and Utilization Endogenous variables Ownership (Number of vehicles) Endogenous variables -Utilization (Daily VMT – Vehicle Miles Traveled) Total no of Vehicles No of Cars No of Vans No of SUVs No of Pickup Trucks Household VMT VMT on Cars VMT on Vans VMT on SUVs VMT on Pickup Trucks Intercept 0.538 0.090 -0.095-0.1010.180 13.007 0.000 -1.338 0.000 0.000 Weekend (d) 0.000 0.000 0.000 0.000 0.000 -10.504 0.000 1.573 0.000 -0.826 Urban Household (d) -0.395 0.249 0.021 0.000 -0.286 -13.772 -1.549 0.000 0.000 1.070 Detached House (d) 0.232 -0.0930.043 0.028 0.051 0.000 -1.136 1.070 0.000 0.000 Annual income 0.004 0.000 0.000 0.003 -0.002 0.191 0.000 0.010 0.011 0.000 No of Adults 0.054 0.000 0.068 -0.025-0.026 0.000 0.000 0.000 0.000 0.000 No of Children -0.041 -0.1260.115 0.025 -0.018 1.590 0.000 0.000 0.000 0.000 No of Workers 0.141 -0.015-0.0280.025 0.038 10.263 0.000 -0.444 0.000 0.000 Exogenous Variables No of Drivers 0.674 0.202 -0.0240.000 -0.041 11.772 0.000 0.000 0.000 0.954 Total no of Vehicles 0.000 0.319 0.059 0.089 0.289 4.685 -2.222 0.000 0.000 -1.810 No of Cars 0.000 0.000 0.000 0.000 0.000 0.000 16.65 -2.714 -3.502 -4.405 No of vans 0.000 0.000 0.000 0.000 0.000 2.395 -12.363 26.916 -3.831 -5.139 No of SUVs 0.000 0.000 0.000 0.000 0.000 3.386 -12.164 -3.521 25.968 -4.726 No of Pickup Trucks 0.000 0.000 0.000 0.000 0.000 3.953 -9.432 -2.490 -3.631 20.765 Endogenous variables Household VMT 0.000 0.000 0.000 0.000 0.000 0.000 0.465 0.103 0.121 0.170 No of observations = 19,360 Households. 2[44] = 48.971, Prob[ 2>value] = 0.280. Adjusted Goodness-of-Fit Index = 0.998. All effects significant at 95% level Notes: (d) Dummy variable Row variables affect column variables.

PAGE 51

43Table 6.2 Total Effects, Structural Equation s Model of Vehicle Ownership and Utilization Endogenous variables Ownership (Number of vehicles) Endogenous variables -Utilization (Daily VMT – Vehicle Miles Traveled) Total no of Vehicles No of Cars No of Vans No of SUVs No of Pickup Trucks Household VMT VMT on Cars VMT on Vans VMT on SUVs VMT on Pickup Trucks Weekend (d) 0.000 0.0000.0000.000 0.000 -10.504 -4.889 0.493 -1.268 -2.609 Urban Household (d) -0.395 0.123-0.003-0.035-0 .400 -17.330 -2.447 -1.068 -1.969 -9.834 Detached House (d) 0.232 -0.0190.0560.049 0.118 1.854 -3.507 2.355 0.918 1.915 Annual income 0.004 0.0010.0000.003 -0.001 0.217 0.082 0.026 0.118 -0.014 No of Adults 0.054 0.0170.071-0.021-0.010 0.311 -0.221 1.995 -0.792 -0.600 No of Children -0.041 -0.1390.1130.022 -0.030 1.625 -2.840 3.578 0.923 -0.338 No of Workers 0.141 0.029-0.0200.038 0.079 11.315 4.481 -0.219 2.036 3.096 Exogenous Variables No of Drivers 0.674 0.4170.0160.060 0.154 15.775 10.418 0.319 1.376 3.405 Total no of Vehicles 0.000 0.3190.0590.089 0.289 6.269 1.468 0.330 0.667 3.135 No of Cars 0.000 0.0000.0000.000 0.000 0.000 16.650 -2.714 -3.502 -4.405 No of vans 0.000 0.0000.0000.000 0.000 2.395 -11.248 27.162 -3.541 -4.732 No of SUVs 0.000 0.0000.0000.000 0.000 3.386 -10.588 -3.173 26.377 -4.151 No of Pickup Trucks 0.000 0.0000.0000.000 0.000 3.953 -7.592 -2.084 -3.154 21.436 Endogenous variables Household VMT 0.000 0.0000.0000.000 0.000 0.000 0.465 0.103 0.121 0.170

PAGE 52

44Table 6.3 Estimated Variance-Covariance Matrix of the Disturbances of the Equations for Endogenous Variables No of cars No of vans No of SUVs No of Pickup tucks No of Cars owned 0.482 No of Vans owned -0.090 0.166 No of SUVs owned -0.138 -0.036 0.222 Ownership variables No of Pickup tucks owned -0.190 -0.031 -0.039 0.281 VMT on Cars VMT on Vans VMT on SUVs VMT on Pickup Trucks VMT on Cars 1081.578 VMT on Vans -210.255 400.533 VMT on SUVs -241.315 -55.813 462.473 Utilization variables VMT on Pickup Trucks -340.708 -72.473 -91.962 624.689 Figure 6.1 Structural Equations Framewor k of Household Vehicle Ownership T rends and Daily Utilization Patterns Total Vehicle Utilization (Household VMT) Utilization by Vehicle Type (VMT by vehicles of each type) Total Vehicle Ownership (Number of Vehicles) Ownership by Vehicle Type (Number of vehicles of each type) Socio-Demographic Attributes of the Household

PAGE 53

456.3 Multinomial Logit Model of Recent Vehicle Acquisitions In addition to the vehicle fleet ownership by combination of number of each type, it is important to understand the recent vehicle acqui sition trends. It is crucial to consider vehicle type choice behavior of households for recent vehicle purch ases to understand the recent trends in vehicle ownership. Unders tanding of choice making behavior is also important in the assessment of demand for each type of vehicle and in the development of vehicle transactions models. This section pr ovides such analysis in the context of household attributes. To distinguish the four types of vehicles based on their addition to the household vehicle fleet, a multinomial logit model (Table 6.4) was estimated for the choice of the type (car, van, SUV and pickup truck) of ve hicle recently acquired by the household. Model was estimated only for a sample of ho useholds that have acquired their recent vehicle not former than one year. This is because for the households that have purchased new vehicle former than a ye ar, the socio-economic and the residential location and type attributes when they actually purchased the vehicle, which would have actually influenced the choice making, are not likely to be the same as current attributes. Given that the data has information of only curren t attributes, it is appropriate to model the vehicle acquisition of only t hose households that have re cently bought vehicles. The observed utility was specified as a functi on of household struct ure and demographic attributes, annual income status, residentia l type and location va riables and previous vehicle ownership. The explan atory variables set for previo us vehicle ownership, which is very important in the new vehicle type c hoice, is generally not found in the literature. The utility of buying a car was taken to be the base case with zero value. The model estimates show that households in urban areas show higher propensity to add a car to their vehicle fleet compared to any other ve hicle type, whereas households living in detached single type of houses show the least tendency to add a car to their vehicle fleet. This is consistent with the results of the joint ownership and utilization model in the previous section. Households living in houses that are not detached and single houses do tend to own cars perhaps due to the parking space issues. These trends indicate the effect of urban form a nd immediate neighborhood design and dense environments on the vehicle type choice of the households. This also suggests a possibility of transportation control measures like parking pricing etc. Higher average annual income favors addi tion of SUV followed by van. They are least likely to add a pickup truc k to their vehicle fleet. This reiterates the fact that SUVs belong to higher income households. Children in the household incr ease the tendency of households to buy a van followed by SUV, pickup truck and car in that order. Households with retired people are most likel y to acquire a car and are less likely to add SUV or pickup truck. Househol ds of higher size are most lik ely to add van or SUV and least likely to add pickup truck. Higher number of workers in a household makes it least inclined towards buying a van. This is probably because each worker in the household might necessarily need a separate vehicle for work. The effect of number of workers in the household is not significantl y different for cars, vans an d SUVs. The first vehicle (if the household was previously in zero ve hicle state) acquired by any household was mostly likely to be a car. It can also be inferred that multi-vehicle households are more likely to possess pickup trucks, vans and SUVs. Presence of van/SUV/pickup truck deters

PAGE 54

46 the acquisition of the same vehicle type. Howe ver, presence of a car doesn’t deter the acquisition of another car. Presence of cars) does not show any statistically significant difference in the addition of van/SUV/car to the fleet. So, households with car are likely to buy any given type of vehicle. However, th ey are most likely to acquire a pickup truck. Households with car/van/SUV are more likely to add a pickup truck to their vehicle fleet. Previous vans or SUVs deter the addition of either type of them as a new vehicle in vehicle fleet. Presence of pickup truck doesn ’t significantly aff ect the acquisition of van/SUV in comparison to addition of a ca r. Thus, apart from the socio-economic and residential area attributes of a household, previous vehicle ownership of the household plays a vital role in the choice of the vehicle type to be adde d to the fleet. The alternative specific constants indicate that given all el se the same, cars are given higher intrinsic preference by the households while making a pu rchase decision. The alternative specific constant term for pickup truck choice is not significantly different from that of car. Essentially, the discrete choice model for the type of recently owned vehicle has shown the behavioral differences in choice making in the context of socio-demographic attributes and previous vehicle ownership co mbination. The model estimates have also confirmed the findings of the joint model of vehicle use ownership and utilization in the previous section. The ownership of different type s of vehicles is for di fferent types of trip making represented by the socio demographic a ttributes in this model. The forthcoming model of the type of vehicle chosen by a driv er for his/her trip distinguishes the four vehicle types based on the attri butes of the trips they are be ing driven for and also on the characteristics of the trip maker. 6.4 Multinomial Logit Model of Driver’s Vehicle Type Choice for a Trip Analysis of the vehicle type choice behavior of drivers for their trips can give further insights into the trends of vehicle utiliza tion and the preferences of people amongst the opportunities provided and constraints imposed by their socio-demographics. It will also provide a better understanding of the vehicle utili zation patterns in the context of driver attributes in a ceteris paribus situation. Hence, a multinomial logit model (Table 6.5) is estimated for the vehicle type (car, van, SUV, Pickup truck) chosen by a driver for his/her trip based on the trip attributes and driver attributes. The choice set was varied for trips from each household; i.e. the alternatives considered for a trip from a particular household were only the types of vehicles possessed by that household. Cases with no choice (only one alternative) were remove d for they wouldn’t provide any useful information in the logit model estimation. The base case utility of c hoosing car for a trip was specified as zero for id entification purposes. Given that the outcome is just the probability of the type of the vehicle and not of the individual vehicle, the possibility of error term correlations due to common unobserved attributes of simila r types of vehicles is eliminated. The sign of the intercepts in the utility e quations indicate that given all else the same, cars are still driven the most compared to all other vehicle types. The coefficients indicate a higher tendency among drivers to choose comparatively new SUVs and older pickup trucks, perhaps due to the differences in the holding durations for these vehicles. Moreover, SUVs are comparativ ely newer vehicles of the fl eet. The tendency to prefer vans, SUVs and pickup trucks increases with ag e of the driver. Howeve r, as indicated by

PAGE 55

47 the coefficients of age square term, drivers st arting from the age of la te 40s tend to revert back to cars. This implies that drivers in their mid age tend to drive non-car type of vehicles and elderly drivers tend to drive cars. On the other hand, one might question whether the elderly of tomorrow will behave like the elderly of today or like themselves today. That is, if people are accustomed to using minivans and SUVs during their working years, perhaps they will continue to do so in the future when they retire as well. Employed drivers choose to drive cars more compared to any other vehicle type. Employed drivers might choose to drive cars when compared to pickup trucks and vans. However, the tendency of unemployed drivers choosing to drive SUVs is counter intuitive. Female drivers tend to choose va ns and male drivers tend to choose pickup trucks and SUVs. Drivers from high averag e income households tend to choose SUVs. Drivers from households with hi gh vehicle to driver ratio are more like ly to drive a car than any other vehicle. This is perhaps due to a larger presence of cars in the vehicle fleet. As expected, there is a high tendency th at trips of high vehicle occupancy are made in vans and SUVs rather than in cars. Pickup trucks are more likely to be used to make trips of lower occupancy. However, pi ckup trucks are most likely, and SUVs and vans are less likely to be dr iven for joint trip making with non-household members. There is also a high tendency that pi ckups are chosen for work trips and that they are chosen the least for recreational trips and for trips on weekend. SUVs and vans show positive coefficients for weekend trips but are statisti cally not significant. They are also more likely to be used for non-work trips, but the coefficients are not sta tistically significant. The coefficients of trip length in terms of duration are also negative and statistically significant for both vans and SUVs. They are mo re likely to be driven for comparatively shorter trips (trip time). The significant negativ e coefficient of trip length (minutes) on SUV choice was not expected as the descrip tive analysis showed higher average trip lengths for SUVs when compared to cars. The tendency of vans to be driven for shorter trips, given that they are also being driven more (in terms of number of trips), has important implications to policy from the pers pective of emissions and fuel consumption. Pickup trucks are more likely to be used in a non-urban setting. Even the coefficients of Vans and SUVs show negative signs for the urban indicator variab le but they are not statistically significant. Vans are least likely to be used for trips that involve longer time spent at the destination. This might indicate the lower usage of vans for driving to perform activities of longer durations. As e xpected, there is no si gnificant difference in the type of the vehicle driven for trips in di fferent times of day. The vehicle type choice decisions are not such short-term decisions to be influenced by the time of day. Thus the vehicle type choice for a trip is dependent upon the type of th e trip and the characteristics of the trip maker. The general trip making patterns of the household members shaped by the socio-demographics of the hous ehold will in turn play an important role in the vehicle ownership patterns.

PAGE 56

48Table 6.4 Multinomial Logit Model for the Recently Acquired Vehicle Type Variable Coefficient S.E B/S.E P Constant -1.87217 0.21351 -8.769 0.000 Urban HH (d) -0.17600 9.03E-02 -1.949 0.051 Detached single house (d) 0.29761 9.75E-02 3.052 0.002 High income HH (d) 0.00000 3.29E-06 -1.307 0.191 No of children (<18yrs) 0.36056 8.14E-02 4.427 0.000 Retired people in HH (d) 0.09974 0.134845 0.740 0.460 Household size 0.17418 7.66E-02 2.275 0.023 No of workers in HH -0.28964 6.13E-02 -4.727 0.000 Previously zero vehicle HH (d) -0.57669 0.152194 -3.789 0.000 HH owns a car (d) -0.10732 0.109913 -0.976 0.329 HH owns a van (d) -0.75020 0.137313 -5.463 0.000 HH owns a SUV (d) -0.47705 0.127639 -3.737 0.000 Van HH owns a pickup truck (d) -0.10531 0.10236 -1.029 0.304 Constant -1.80841 0.168469 -10.734 0.000 Urban HH (d) -0.25802 7.31E-02 -3.529 0.000 Detached single house (d) 0.31354 7.95E-02 3.945 0.000 High income* HH (d) 0.00002 2.06E-06 9.703 0.000 No of children (<18yrs) 0.17556 6.78E-02 2.591 0.010 Retired people in HH (d) -0.47263 0.116503 -4.057 0.000 Household size 0.10498 6.38E-02 1.645 0.100 No of workers in HH -0.02400 5.20E-02 -0.461 0.645 Previously zero vehicle HH (d) -0.81720 0.128662 -6.352 0.000 HH owns a car (d) -0.05448 8.43E-02 -0.646 0.518 HH owns a van (d) -0.43223 0.103452 -4.178 0.000 HH owns a SUV (d) -0.12911 9.00E-02 -1.435 0.151 SUV HH owns a pickup truck (d) -0.14461 7.87E-02 -1.838 0.066 Constant -0.05339 0.157518 -0.339 0.735 Urban HH (d) -0.88050 6.46E-02 -13.627 0.000 Detached single house (d) 0.35685 7.57E-02 4.716 0.000 High income* HH (d) -0.00001 2.32E-06 -3.227 0.001 No of children (<18yrs) 0.14439 6.56E-02 2.200 0.028 Retired people in HH (d) -0.37967 0.104064 -3.648 0.000 Household size -0.17503 6.18E-02 -2.832 0.005 No of workers in HH -0.02188 4.90E-02 -0.446 0.655 Previously zero vehicle HH (d) -0.96006 0.127844 -7.510 0.000 HH owns a car (d) 0.25480 7.92E-02 3.218 0.001 HH owns a van (d) 0.09584 9.14E-02 1.049 0.294 HH owns a SUV (d) 0.34809 8.20E-02 4.243 0.000 Pickup Truck HH owns a pickup truck (d) -0.38434 7.47E-02 -5.147 0.000 #obs = 8651, Log-Likelihood function = -9581.852, 2[36] = 1061.076, Prob [ 2>value] =.000 Log-L fn R2 R2 Adj Model -9581.852 No coefficients -11992.833 0.201 0.199 Constants only -10112.390 0.052 0.051

PAGE 57

49Table 6.5 Multinomial Logit Model for Dr iver’s Vehicle Type Choice for a Trip Variable CoefficientS.E B/S.E P Constant -3.96361 0.151 -26.207 0.000 Number of years vehicle owned -0.00634 0.005 -1.319 0.087 Drivers Age 0.17498 0.007 26.603 0.000 Square of drivers age -0.00173 0.000 -24.340 0.000 Driver is employed (d) -0.49307 0.042 -11.845 0.000 Male driver (d) -0.33169 0.033 -9.999 0.000 Vehicle occupancy 0.46439 0.019 24.360 0.000 Non-household passenger on trip (d) -0.44134 0.060 -7.408 0.000 Trip Length (min) -0.00184 0.001 -2.150 0.032 Time spent at destination of trip -0.00036 0.000 -3.326 0.001 Van Vehicles to drivers ratio -0.12396 0.028 -4.418 0.000 Constant -2.07397 0.113 -18.313 0.000 Number of years vehicle owned -0.07261 0.004 -16.646 0.000 Drivers Age 0.10180 0.005 19.042 0.000 Square of drivers age -0.00106 0.000 -17.275 0.000 Driver is employed (d) -0.28025 0.036 -7.803 0.000 Male driver (d) 0.22272 0.027 8.297 0.000 Vehicle occupancy 0.22926 0.018 12.484 0.000 Trip length (min) -0.00224 0.001 -3.435 0.001 Non-household passenger on trip (d) -0.30950 0.051 -6.103 0.000 Non-work trip (d) 0.05738 0.032 1.807 0.071 Vehicles to drivers ratio in HH -0.18872 0.024 -7.898 0.000 SUV Average household income 0.00001 0.000 6.368 0.000 Constant -3.35042 0.108 -31.024 0.000 Number of years vehicle owned 0.01105 0.004 3.141 0.002 Drivers Age 0.11209 0.005 24.275 0.000 Square of drivers age -0.00123 0.000 -24.385 0.000 Driver is employed (d) -0.21907 0.037 -5.943 0.000 Male driver (d) 2.95597 0.031 96.678 0.000 Vehicle occupancy -0.45727 0.021 -21.545 0.000 Non-household passenger on trip (d) 0.35430 0.051 6.974 0.000 Trip length (min) -0.00097 0.001 -1.688 0.092 Work trip (d) 0.22006 0.032 6.786 0.000 Recreational trip (d) -0.06896 0.033 -2.098 0.036 Trip on a weekend (d) -0.10531 0.030 -3.521 0.000 Urban area household (d) -0.12776 0.026 -4.997 0.000 Pickup Truck Vehicles to drivers ratio -0.32036 0.021 -15.410 0.000 #obs = 71360, Log-L fn = -43952.678, 2 [34] = 96042.162, P[ 2>value] = 0.000 Log-L R2 R2 adj Model -43952.678 No coefficients -98925.9656 .555 .555 Constants only -66309.949 .454 .455

PAGE 58

50 CHAPTER 7 CONCLUSIONS AND FUTURE RESEARCH 7.1 Conclusions This thesis utilized the 2001 National Household Survey (NHTS) data to study the patterns of household vehicle ownership and uti lization in the United States. Four vehicle types; cars, vans, SUVS and pickup trucks are distingu ished based on how households own and use them for their travel needs. The NHTS data proved to be a rich source for such an analysis. The thesis provided an extensive descrip tive analysis of the data to assess the differences among the vehicle types in a univa riate setting. The relevance and importance of socio-demographic factors is assessed in the context of the ownership, use and allocation of each vehicle type. The ownership analysis probed into aspects like vehicle fleet combination, length of ownership and trends in vehicle acquisitions across vehicle types. Vehicle utilization patte rns are discussed in terms of the trip attributes and the extent of use and primary dr iver allocation. Vehicle mile s traveled, person trips and driver trips served, occupancies and trip lengths were analyzed for weekdays and weekends to illuminate any differences in use across vehicle types. The primary driver allocation and utilization analysis offered in sights into extent to which a vehicle is devoted to a primary driver and also the differences across vehicle types. This study also analyzed the structural relationships among socio-demographics, ownership and utilization patterns of each ve hicle type in a unified framework. The structural equations model developed in this context provides uncondi tional estimates by considering ownership and uti lization simultaneously. The m odel considered the vehicle ownership and fleet combina tion as endogenous, which is better than an individual ownership model that considers the vehi cle ownership as exogenous. The multinomial logit model for the type of recently acquired vehicle provided insights into the recent trends in the vehicle ownership patterns. The model estimates illustrated the importance of considering previous vehicle ownership in the decision making of the new vehicle type to be bought. Finally the multinomial logit model of the type of vehicle chosen by the driver for a trip offered an understanding of the choice making of the drivers in the context of person attributes a nd trip attributes. Essentially, this model has offered an understanding of who drives wh at type of vehicles for wh at purposes when and where. Results of the analysis indicate that car s are still the dominant types of vehicles owned and recently purchased. Each additional vehicle in the household is more likely to be a car rather than any other vehicle. This may be because of their prevalence right from the beginning of the automobile era. Cars are s till the prevalent type of vehicles present in the fleet. Urban area households have higher car ownership than that of a rural area household. Younger and elderly people and employed persons tend more to drive cars than any other vehicle.

PAGE 59

51 Vans show the least presence in the vehi cle fleet. These vehicles are being owned more by households in urban areas, households living in detached houses, and households of larger size and children count. Households with larger number of workers show lower van ownership tendency. Vans ar e relatively new vehicles in the fleet, essentially being used as family vehicles for trips of high occ upancy, shorter length and are driven more by females and are being used more on week ends. These findings could have several implications. Vans, being owned more by ho useholds with children for their use as family vehicles, need to be better designed from safety perspective with child restraints etc. Given that they are being driven more by female drivers who are more prone to higher accident severity, vans ne ed to be better designed for co mfort of driving and safety with female driver as a basis. SUVs are also relatively new vehicles in the fleet, and are being owned by affluent households SUVs are similar to in th at they are also owned by households living in detached houses and la rger households and are used as family vehicles for trips of higher occupancy. Pickup trucks are being owned more by hous eholds in rural ar eas, households of smaller size and households living in de tached houses, and households with lower income. Pickup trucks are not used as family vehicles; they are rather used as work vehicles. These vehicles are driven more by male drivers and are owned for longer time when compared to other vehicles. Pickup trucks are similar to vans a nd SUVs in that they are being driven more than cars and that people of mid-age are driving them more. In addition to confirming many perceptions about the patterns in ownership and use of the four vehicle types, this study has brought out so me subtle differences. In many ways, this analysis of the diverse personal vehicle fleet of United States has several implications to the transportation systems planning, policymaking and perhaps regulatory action. The finding that vans, S UVs and pickup trucks are dr iven over larger distances than cars can have several implications to emissions and energy consumption. These vehicles being larger than cars can actually take up more car equiva lents on a road and on parking spaces hence pose capacity constraint s on roads, at intersections and in parking lots. An important observation that could be made throughout the analysis is that the usage of vans and SUVs is similar to that of cars in similar ways. There is no indication of additional work travel on Vans and SUVs; th ese vehicles are in fact being used being used more than cars as personal vehicles. They are also being used as family vehicles for trips of high vehicle occupancy to vari ous purposes. This evidence warrants the reconsideration of the light du ty truck classification. Evidence of higher trip rates of primary drives of vans, SUVs and pickup trucks than that of cars, and the possibility of differences in the daily household VMT differe nt vehicle fleet combination indicates that it may be necessary to incorporate the effect of vehicle fleet combination in addition to the total vehicle ownership in determining trip generation rates. However, a closer study is warranted in order to determine the causa lity of the relation between the amount of household travel and the vehicle fleet combin ation. Higher ownership of vans by female drivers, and their usage for high occupanc y trips and by households with children indicates that these vehicles may need to be designed female driver friendly and be made more safer with better children restraints etc.

PAGE 60

527.2 Future Research This study offers a detailed analysis of th e cross sectional data from the 2001 NHTS. However, patterns, preferences, and mark ets do change over time. Analysis of the temporal trends of vehicle ownership and util ization should certainly be undertaken for a panel data set in order to obtain better insi ghts. Such study can illuminate much about the market share trends and inform policy in a better way. A thor ough policy oriented research is required to unders tand the implications of the current vehicle ownership and utilization patterns in order to suggest appr opriate regulatory action. Models of stated preference survey data are necessary in order to understand consumer and market responses and evaluate proposed policy measures. The analysis in this study can be extende d in many ways. It may be necessary to econometrically refine the models presented in this study. The structural equations model of vehicle ownership and utilization doesn’t take into consideration the censored nature of endogenous variables. Vehicl e ownership and utilization ca nnot be less than zero. The estimates could also be improved by consider ing the ordinal discrete nature of vehicle ownership variables. It would be certainly interesting to analyze the fractional split taken by vehicle of each type of the household’s da ily vehicle mileage (VMT) instead of the mileage itself in the structural equations fr amework. This kind of fractional split model would give insights into the structural re lationships between the socio-demographic variables and the share of the daily househol d vehicle mileage taken by vehicles of each type. This fractional split could be further be extended to the work, non-work fractional split for each vehicle type in order to better understand the vehicle usage in a multivariate setting. The fractional split models in a logistic distribution framework can provide better predictions. The multinomial logit models pr esented could be saddled with the IIA (Independence of Irrelevant Alternatives) property. A random coefficients logit model can incorporate any form of error correlati on across alternatives and heteroskedasticity through a general covariance matrix. Moreover, such general model can account for the heterogeneity of parameters, hence tast e variations in the population. Current methodological advancements must be exploited to better analyze the trends before coming up with policy suggestions. Not all aspects of ownership and utilizati on are covered in this analysis. One such unexplored aspect in the recent past is the ve hicle holding durations. An in-depth analysis of the length of time a vehicle is used befo re it is disposed is necessary because older vehicles in the fleet cause more vehicular emissions. The descriptive analysis of the primary driver allocation can be extended to models of vehicle allocation patterns to primary drivers in order to study the trends in a multivariate setting. This study has concentrated on the socio-demographic factor s, which are important determinants of vehicle ownership and u tilization patterns. However, anal ysis should also be extended to many other important aspects; attitudinal factors, population density and other land use and urban form specific variables, vehicle ow nership and operating co sts, transportation system performance indicators like network leve l of service etc. which certainly have the potential to influence vehicle ownership and use. Future e fforts require exploration of further complex structural relations intr oducing the effects of the above additional factors.

PAGE 61

53 Given this analysis, results and impli cations to planning, policy and perhaps regulatory action, several cauti ons must be exercised before directly using the results. There could be many differences in the wa y households and people own and use leased vehicles and vehicles that are bought old. Vehicle ownershi p decisions may also depend upon the type of ownership; ex: lease-vers us-own and new-versus-old vehicles. A household’s vehicle purchase definitely depends upon th e new-versus-old vehicle decision. Extended use of older vehicles may have implications to policy. Presence of older vehicles in the fleet may mask the e ffectiveness of certain policy and regulatory measures that are actually appropriate for new vehicles. Identifyi ng and controlling for factors like lease-versus-buy, and new-versus-o ld can provide us better insights into the vehicle ownership and utilization patterns. Ma turity of vehicle type is an issue that precludes the direct translation of current tr ends and differences in the ownership and use patterns of different types to future. The pred ictive values and trends may be influenced due to the fact relatively new vehicles like va ns and SUVs have not yet matured in terms of their penetration into the market. Curre nt ownership and use patterns may not be extrapolated to future with out considering th e possibility of growth in market share of these vehicles and their maturity in term s of marker penetration. However, many commodities in the market may not see a matu rity period due to the rapid changes in technology that accelerates the replacements of products. It may be possible that before SUVs see period, the new cross breed of SUVs and pickup trucks can take over the market share and so on. For example, elderly of today show lower ownership of SUVs and vans. However, today’s mid-age drivers, who will be the elderly of tomorrow, may be different because they are currently showing higher ow nership of vans and SUVs. Hence it is possible that the elderly of to morrow continue the use of their vans and SUVs. However, pickup trucks, which are relatively old in th e fleet, have not show n such trend. Their increasing popularity in the recent years can perhaps make them popular among the elderly of tomorrow. Similar arguments can be extended to other implications discussed in this chapter.

PAGE 62

54 REFERENCES Aizcorbe, A. and M. Starr-McCluer. 1997. Vehicle Ownership, Purchases, and Leasing: Consumer Survey Data. Monthly Labor Review 120(6): 34-40. Anderson, C.L., P.F. Agran and D.G. Wi nn. 2001. Pickup Truck Use in the National Personal Transportation Survey. Accident Analysis and Prevention 33(4): 499-506. Anderson, C.L., D.G. Winn and P.F. Agran. 1999. Differences between Pickup Truck and Automobile Driver-Owners. Accident Analysis and Prevention 31(1): 67-76. Arbukle, J.L., 2000. AMOS Users’ Guide, Version 4. Smallwaters Corp., Chicago. Ben-Akiva, M. and S.R. Lerman. 1985. Discrete Choice Analysis: Theory and Application to Travel Demand. The MIT Press, Cambridge. Berkovec, J. 1985. Forecasting automobile demand using disaggregate choice models. Transportation Research B 19B(4): 315-329. Berkovec, J. and J. Rust. 1985. A nested l ogit model of automobile holdings for one vehicle households. Transportation Research B 19B(4): 275-285. Bhat, C.R. and V. Pulugurta. 1998. A Comparis on of Two Alternative Behavioral Choice Mechanisms for Household Auto Ownership Decisions. Transportation Research 32B(1): 61-75. Bollen, K.A., 1989. Structural Equations with Latent Variables. John Wiley, New York. Chang, L.Y. and F. Mannering. 1999. Analysis of injury severity and vehicle occupancy in truckand non-truc k-involved accidents. Accident Analysis and Prevention 31 (5): 579–592. Choo, S. and P.L. Mokhtarian. 2002. The Relationship of Vehicle Type Choice to Personality, Lifestyle, Attitudinal, and Demographic Variables. Publication No. UCDITS-RR-02-06, Institute of Transportation Studies, University of California, Davis. Gardenhire, A.D. and M.W. Sermons. 2001. Understanding Automobile Ownership Behavior Of Low-Income Households: Ho w Behavioral Differe nces May Influence Transportation Policy. Transportation Research Circular E-C026: 179-195.

PAGE 63

55 Golob, T.F. 2003. Structural equation m odeling for travel be havior research. Transportation Research 37 B: 1–25 Golob, T.F. 1990. The Dynamics of Trav el Time Expenditures and Car Ownership Decisions. Transportation Research 24 A (6): 443-463. Golob, T.F., S. Kim and W. Ren. 1996. How Households Use Different Types of Vehicles: A Structural Driver Allocation and Usage Model. Transportation Research 30A(2): 103-118. Golob, T.F., D.S. Bunch and D. Brownst one. 1997. A Vehicle Use Forecasting Model Based on Revealed and Stated Vehicl e Type Choice and Utilization Data. Journal of Transport Economics and Policy 31(1): 69-92. Greene, W.H., 2002. LIMDEP Version 8.0: User's Manual, Econometric Software, Inc., New York. Hellman K.H.,and R.M. Heavenrich. 2000. Li ght Duty Automotive Technology and Fuel Economy Trends: 1975 Through 2003. Executive Summary. Report of Advanced Technology Division, Office of Transportation and Air Quality, U.S. Environmental Protection Agency. Hensher, D.A. 1985. An Econometric Model of Vehicle Use in the Household Sector. Transportation Research 19B(4): 303-314. Hensher, D.A., and V.L. Plastrier. 1985. Towards a Dynamic Discrete-Choice Model of Household Automobile Fleet Size and Composition. Transportation Research 19B(6): 481-495. Hess, D.B. and P.M. Ong. 2002. Traditional Neighborhoods and Automobile Ownership. Transportation Research Record 1805: 35-44, Journal of the Transportation Research Board, National Research Council, Washington, D.C. Hocherman, Irit., J. N. Prashker and Mosh e Ben-Akiva (1983) Estimation and use of dynamic transaction models of automobile ownership. Transportation Research Record 944, pp. 134-141. Hu, P.S. 2003. Americans and Their Vehicles. Presentation from the 2001 National Household Travel Survey website http://nhts.ornl.gov/2001/presentations/americanVehicles/index.shtml Hu, P.S. and Y. Jennifer. 1999. Summary of Travel Trends: 1995 National Personal Transportation Survey, Federal Highwa y Administration, Washington, D.C. Hoetler, J.W., 1983. The analysis of cova riance structures: G oodness-of-fit indices. Sociological Methods and Research 11: 325-344.

PAGE 64

56 Jreskog, K.G. and D. Srbom. 1986. LISR EL 6: Analysis of Linear Structural Relationships by Maximum Likelih ood and Least squares methods. Scientific Software International, Chicago. Jreskog, K.G. and D. Srbom, 1993. LISREL 8 User’s Reference Guide. Scientific Software International, Chicago. Karlaftis, M. and J. Golias. 2002. Autom obile Ownership, Households without Automobiles, and Urban Traffic Parameters: Are They Related? Transportation Research Record 1792: 29-35, Journal of the Transporta tion Research Board, National Research Council, Washington, D.C. Kitamura, R., T.F. Golob, T. Yamamoto and Wu, G. 2000. Accessibility and Auto Use in a Motorized Metropolis. CD-ROM of the 79th Annual Meeting of the Transportation Research Board, National Research Council, Washington, D.C. Kockelman, K.M. 2000. To LDT or Not to LDT: An Assessment of the Principal Impacts of Light-Duty Trucks. Transportation Research Record 1738: 3-10, Journal of the Transportation Research Board, National Research Council, Washington, D.C., pp. 310. Kockelman, K.M. and Y. Zhao. 2000. Behavior al Distinctions: The Use of Light-Duty Trucks and Passenger Cars. Journal of Transportation and Statistics 3(3): 47-60. Kockelman, K.M. and Y-J Kweon. 2002. Driver Injury Severity: An Application of Ordered Probit Models, Accident Analysis & Prevention 34(3): 313-321. Kuppam, A.R., 1999. A limited dependent variable structural equations model of travel behavior for transportation policy evaluation. Unpublished Ph.D. Dissertation, University of South Florida. Liss, S., N. McGuckin and G. Bryant. 2003. Do More Vehicles Make More Miles? A snapshot analysis of the Nationa l Household Travel Survey 2001. Article from the 2001 NHTS website http://nhts.ornl.gov/2001/presentati ons/vehicleMiles/VehicleMiles.html Lave, C.A. and K. Train. 1979. A di saggregate model of auto-type choice. Transportation Research A 13A(1): 1-9. Mannering, F. 1983. An Econometric Analys is of Vehicle Use in Multi-Vehicle Households. Transportation Research 17A(3): 183-189. Mannering, F. 1986. A Note on Endogenous Vari ables in Household Vehicle Utilization equations. Transportation Research 20B(1): 1-6.

PAGE 65

57 Mannering, F. and K. Train. 1985. Recen t Directions in Automobile Demand Modeling. Transportation Research B 19B(4): 265-274. Mannering, F., and C. Winston. 1985. Dyna mic Empirical Analysis of Household Vehicle Ownership and Utilization. The RAND Journal of Economics 16(2): 215-236. Mannering, F., C. Winston., and W. Starkey. 2002. An exploratory analysis of automobile leasing in the United States. Forthcoming in Journal of Urban Economics 52(1): 154-176. Manski, Charles F. and Leonard Sherman ( 1980) An empirical analysis of household choice among motor vehicles. Transportation Research A 14A(5/6): 349-366. Matthew, A.C. 2003. Why Care About Walking? Presentation from the 2001 NHTS website http://nhts.ornl.gov/2001/pres entations/walki ng/index.shtml McGuckin, N. and N. Srinivasan. 2003. A Walk Through Time Changes in the American Commute. Presentation from the 2001 NHTS website http://nhts.ornl.gov/2001/pres entations/commuting/index.shtml Murakami, E. and J. Young. 1999. Daily Travel By Persons With Low Income. Nationwide Personal Transportation Survey Symposium, conducted at Bethesda, Maryland. FHWA-PL-99-003, HPPI/6-9 9(5M) EW: 65-84, Federal Highway Administration, Washington, D.C. Niemeier, D., L. Redmond, J. Morey, J. Hick s, P. Hendren, J. Lin, E. Foresman, and Y. Zheng. 2001. Redefining Conventional Wisdom : Exploration Of Au tomobile Ownership And Travel Behavior In The United States. Transportation Research Circular E-C026 : 207-219. Oak Ridge National Laboratory (ORNL). The 2001 NHTS website http://nhts.ornl.gov/2001/index.shtml Pickrell, D and P. Schimek. 1999. Growth in Motor Vehicle Ownership and Use: Evidence from the Nationwide Personal Transportation Survey. Journal of Transportation and Statistics 2(1): 1-17. Pisarski, A.E. 1994. The Demography of the U. S. Vehicle Fleet: Observations from the NPTS. Special reports on trips and vehicle at tributes, 1990 NPTS publications website http://npts.ornl.gov/npts/1990/ Pisarski, A.E. 1996. Commuting In America II: The Second National Report On Commuting Patterns And Trends. Eno Transportation Founda tion, Inc., Landsdowne, VA.

PAGE 66

58 Pisarski, A.E. 2003. Trends We Cannot Afford To Miss. Public Works Management & Policy 8(1): 3-10. Polzin, S.E., and X. Chu. 2003. Early Findings on Public Transporta tion Travel Trends. Presentation from the 2001 NHTS website http://nhts.ornl.gov/2001/pres entations/polzin/index.shtml Polzin, S.E., X. Chu and L. Toole-Holt. 2003. The case for Moderate Growth in Vehicle Miles of Travel: A Critical Juncture in U.S. Travel Behavior Trends. Article from the 2001 NHTS website http://nhts.ornl.gov/2001/articles/index.shtml Pucher, J. and J.L. Renne. 2003. Socioeconomics of Urban Travel: Evidence from the 2001 National Household Travel Survey. Transportation Quarterly 57(3): 49-78. Purvis, C.L. 1994. Using 1990 Census Public Use Microdata Sample To Estimate Demographic And Automobile Ownership Models. Transportation Research Record, 1443:21-29, Journal of the Transportation Rese arch Board, National Research Council, Washington, D.C. Srinivas Meka, S., R.M. Pendyala, and M. Wasantha Kumara. 2002. Structural Equations Analysis of Within-Household Ac tivity and Time Allocation Between Two Adults. CD-ROM of the 82nd Annual Meeting of the Tran sportation Research Board, Transportation Research Board, Nationa l Research Council, Washington, D.C. Tanaka, J.S., 1987. How big is big enough ? Sample Size and Goodness of Fit in Structural Equation Models with Latent Variables. Child Development 58:134-146. Ulfarsson, G. and F. Mannering (2004) Differences in male and female injury severities in sport-utility vehicle, minivan, pickup and passenger car accidents, Accident Analysis and Prevention 36 (2): 135-147. Ullman. J.B., 1996. Structural equation modeli ng. in B.G. Tabachnick and L.S. Fidell Using Multivariate Statistics, 3rd Ed: 709-819. New York, HarperCollins College Publishers. Wooldridge J.M., 2001. Econometric Analysis of Cross Section and Panel Data. The MIT Press, Cambridge. Zhao, Y. and K.M. Kockelman. 2002. Household Vehicle Ownership by Vehicle Type: Application of a Multivariate Negative Binomial Model. CD-ROM of the 81st Annual Meeting of the Transportation Research Board, National Research Council, Washington, D.C.


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 001469378
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-SFE0000280
035
(OCoLC)55644124
9
AJR1132
b SE
SFE0000280
040
FHM
c FHM
090
TA145
1 100
Pinjari, Abdul Rawoof.
3 245
An analysis of household vehicle ownership and utilization patterns in the United States using the 2001 National Household Travel Survey
h [electronic resource] /
by Abdul Rawoof Pinjari.
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 66 pages.
520
ABSTRACT: Vehicle ownership and utilization have a profound influence on activity-travel patterns of individuals, vehicle emissions, fuel consumption, highway capacity, congestion and traffic safety. The influence could be further skewed by the diversity of the vehicle fleet. This thesis presents a detailed analysis of the 2001 National Household Travel Survey data to understand the vehicle ownership patterns, fleet mix, allocation and utilization in the context of household and person socio-demographic characteristics. Along with a rich descriptive analysis, models of vehicle ownership and utilization are estimated to distinguish four vehicle types; cars, SUVs (sport utility vehicles), vans and pickup trucks based on their ownership by households and utilization patterns by household members. The primary driver level vehicle utilization analysis provides insights into the extent of allocation of a vehicle to a single person. In addition to confirming many perceptions about the ownership, acquisition and utilization patterns of different types of vehicles, this analysis brings out some subtle differences and similarities among the vehicle types. The analysis results indicate a greater propensity to acquire and use larger vehicles such as minivans, sports utility vehicles and pickup trucks among certain socio-demographic segments of population. Increased ownership and use of vans and SUVs, and their usage as personal vehicles rather than just work vehicles warrants a need to revise vehicle type specific policies, transportation planning and control measures.
590
Adviser: Pendyala, Ram M.
653
multinomial logit model.
cars.
SUVs.
vans.
structural equations model.
descriptive analysis.
socio-demographics.
pickup trucks.
0 690
Dissertations, Academic
z USF
x Civil Engineering
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.280