SWUTC Research Project Description

Title of Project:  Evolution of Vehicle Fleet: Anticipating Fleet Composition, Use and PHEV Adoption

Project Number:  161023

Principal Investigator:
Kara Kockelman
(512) 471-0210
P.I. Affiliation:  University of Texas at Austin

Project Monitor:
Dr. David Greene
Oak Ridge National Laboratory
P.O. Box 2008 MS6472
Oak Ridge, TN 37831-6472
(865) 946-1314

Project Status:  Active

Date Started:  9/1/09

Estimation Completion Date:  8/31/10

Estimated Cost - Current Fiscal:  $39,000

Estimated Cost - Total Planned:  $39,000

Project Summary:
Project Abstract:
Rising gasoline prices, emerging engine technologies, and changes in fuel-economy policy will result in a variety of behavioral changes, including household holdings (number of vehicles, make and model), vehicle purchase and retirement timing decisions, mode choice and travel distances, speeds, and emissions. This research provides new data on ownership decisions and owner preferences under various scenarios, coupled with calibrated models to microsimulate the nation’s personal-fleet evolution.

Project Objectives:
Evaluating the near- and long-term vehicle ownership decisions and owner preferences under various scenarios, in order to anticipate the qualities of America’s evolving passenger vehicle fleet and carbon emissions, and anticipate PHEV adoption and use policies.

Task Descriptions:
Task 1: Synthesizing Relevant Literature and Experience
To begin this work, the research team will review relevant literature, through a combination of library holdings and on-line resources, computerized databases, and personal communications. Works will include those recognized in the vehicle ownership discussion of Kockelman et al.’s (2008) NAS-commissioned paper (on the topic of greenhouse gas emissions vis-à-vis human activities, emerging technologies, and potential policies) and Musti and Kockelman’s (2009) recent paper. They also are likely to include works by Fisher (2007), Heffner et al. (2007), Chen and Niemeier (2005), Mannering and Winston (1985), Lemp and Kockelman (2008), Dumas et al. (2007), Kockelman and Zhao (2005), Parry et al. (forthcoming), Bomberg and Kockelman (2007), Turrentine and Kurani (2004), Greene et al. (2004, 2006, 2007), and Greene (1994).

Task 2: Initial Data Assembly
A variety of data sets will be examined and analyzed, to get a sense of trends in vehicle holdings, use and life times. These have already been obtained by the PI and include the Austin Vehicle Choice Survey, Toronto Area Car Ownership Study, National Household Travel Surveys, U.S. Panel Study of Income Dynamics, Vehicle Inventory and Use Survey, NHTSA’s General Estimates System, Texas vehicle registrations data base, and recent Austin and Dallas-Ft. Worth travel (and vehicle ownership) surveys. Taken together, these surveys provide information on make and models owned recent vehicle purchases and retirements, vehicle use levels, and vehicle loss through crashes, as well as owner and driver attributes, among other details. The PI has experience working with all of these data sets, and looks forward to a having focused look at vehicle ownership, retirement and use decisions. By ascertaining what details are missing in existing data sets, the project team will fine-tune the project’s new survey instrument, in order to maximize information obtained from project expenditures on Task 3’s data collection.

Task 3: Data Collection
Background data collection on this topic is underway in Austin (thanks to current SWUTC research), which will enable this near-term nation-wide extension of the research, with solicitation of information on current, past and (likely) future vehicle holdings decisions, sensitivity to gasoline prices, strength of vehicle technology preferences, and the like, possibly with follow-up mailed questionnaires (depending on budget constraints). Questions will be added about recharging location opportunities and power rates ($/kWh by time of day) vis-à-vis trip destination and scheduling decisions.   Regional land use variables (e.g., population density, job density, access to a local supermarket or bus route to one’s place of work) will supplement information obtained from the survey on household attributes and vehicle choices. Access to such rich data sets will enhance our understanding of vehicle purchases, retirements and various travel choices – both stated and revealed/actual.

Note: As instructor of record for UT’s only Transportation Data Acquisition and Analysis (graduate) course and chair of the National Academy of Engineering’s Transportation Research Board’s Travel Survey Methods committee, the PI has a very strong track record with survey design and management, and subsequent data analysis. She has managed design and implementation of five different surveys, including one as a series of 1200 phone interviews (for the Texas Department of Transportation on the subject of the public’s perceptions of tolling and other transportation policies).

Task 4: Data Analysis and Model Calibration
Highly experienced in a variety of data analysis techniques (including classical and Bayesian methods, spatial and panel data sets, multiple equations systems, factor analysis, and econometric specifications), the PI looks forward to close analysis of the data from these experiments. Such analysis will include multinomial logit models (see, e.g., Zhao and Kockelman 2000) of vehicle choices, both stated and revealed. The results of these will illuminate the value of gas savings (in the form of fuel economy, and as a function of gas prices) versus vehicle price, performance, age and other attributes. It will be interesting to examine how heavily households discount future gas savings.

The analyses also will include hazard-based and other duration models (see, e.g., Gadda and Kockelman 2007) for vehicle holding period, till loss or resale. Such details will provide insight on how long different types of vehicles are held by different individuals and households. Models will include travel statistics for each vehicle, including miles traveled, number of trips taken and typical travel speeds – all of which impact a household’s carbon emissions. Ordered probit models (see, e.g., Podgorski et al. 2004) of opinions about different vehicle technologies, energy policy, and climate change also will be calibrated, for additional results of interest (e.g., what types of individuals are most likely to adopt new technologies, and what policies may receive the most public support). Newer specifications allowing for modeling joint continuous-discrete responses (e.g., using Bhat’s [2006] multiple discrete-continuous extreme value specification to pivot off of Mannering and Winston’s [1985] work) will also be pursued.

Task 5: Model Applications
Data analysis and model linkages will allow for the full model specification and microsimulation of U.S. household vehicle ownership decisions over time. Existing travel survey data will be mined as well, to get a sense of use patterns and likely use strategies for PHEVs of different range (e.g., PHEV10 vs. PHEV30). The great variety of data and models resulting from this work will allow examination of various transport scenarios and policies, both long-term and short-term. These include investigations of the impact of rising fuel prices, the carbon emissions impacts of gas taxes, cap and trade policies across vehicle manufacturers (to attain certain fleet-level fuel economy standards), benefits of vehicle scrappage (“early retirement”) policies, use of feebates, and the like.

The models estimated above will be applied in order to anticipate the responsiveness of vehicle miles traveled and vehicle fleet composition to energy price fluctuations, the availability of new vehicle technologies and fuels, manufacturers’ vehicle-pricing and design decisions, and so forth. The team will predict vehicle purchase, retirement, and use changes as well as greenhouse gas emissions from the personal-transportation sector as a result of a range of several gas pricing scenarios (e.g., $3, $5 and $7/gallon), over 5, 10, 20, and 30- year horizons. Such vehicle holdings and retirement models are highly innovative, and will permit microscopic evolution of synthetic household (and vehicle) populations, in order to anticipate actual emissions. The PI’s recent and on-going work microsimulating the vehicle fleet in Austin (e.g., Musti and Kockelman [2009]) will prove helpful here.

Task 6: Report Writing & Results Dissemination
Deliverables will include a report analyzing all data assembled; details of models predicting vehicle purchase, retirement, and use changes; and all application results. Information relevant to PHEV purchase and use and their benefit for industry stakeholders and policy makers will be highlighted.

Index Terms: