Peer Review for the Report
"The Rebound Effect from Fuel
Efficiency Standards:
Measurement and Projection to 2035"
&EPA
United States
Environmental Protection
Agency
-------
Peer Review for the Report
"The Rebound Effect from Fuel
Efficiency Standards:
Measurement and Projection to 2035"
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
Prepared for EPA by
ICF International, L.L.C.
EPA Contract No. EP-C-12-011
Work Assignment No. 2-07
NOTICE
This technical report does not necessarily represent final EPA decisions or
positions. It is intended to present technical analysis of issues using data
that are currently available. The purpose in the release of such reports is to
facilitate the exchange of technical information and to inform the public of
technical developments.
&EPA
United States
Environmental Protection
Agency
EPA-420-R-15-013
July 2015
-------
In 2011, EPA contracted with Ken Small of UC Riverside to update and enhance an existing model to
estimate the VMT rebound effect for light-duty vehicles, defined as the change in vehicle miles
traveled resulting from a change in fuel economy. The updates included using more recent state-
level data for travel, as well as methodological enhancements to explore potential asymmetric
responses depending on the direction of fuel cost changes, and to evaluate the role of media
coverage of energy costs on driver's response. The resulting report by Ken Small, with contributions
by Kent Hymel, is entitled "The Rebound Effect from Fuel Efficiency Standards: Measurement and
Projection to 2035."
Prior to the release of the Final Report from Small and Hymel, EPA contracted with ICF
International to conduct a peer review of the Small and Hymel report. The three peer reviewers
selected by ICF were Drs. Kenneth Gillingham (Yale University), David Greene (University of
Tennessee), and James Sallee (University of Chicago). EPA would like to extend its appreciation to
all three reviewers for their efforts in evaluating this survey. The three reviewers brought useful
and distinctive views in response to the charge questions.
This document contains three main components:
I. Peer Review of Small and Hymel Report on the Rebound Effect for Light-Duty Vehicles,
Conducted by ICF International
1. Introduction
2. Selection of Peer Reviewers
3. The Peer Review Process
4. Summary of Reviewer Comments
Appendix A. Resumes and Conflict of Interest Statements
Appendix B. Charge Letter
Appendix C, D, and E. Complete Reviews
II. Draft Report - Peer Reviewed Version, "The Rebound Effect from Fuel Efficiency
Standards Measurement and Projection to 2035"
III. EPA's Response to Peer Review Comments
-------
I.
Peer Review of Small and Hymel Report
on the Rebound Effect for Light-Duty
Vehicles, Conducted by ICF International
-------
Peer Review of December
2013 LDV Rebound
Report by Small and
Hymel
January 31, 2014
Prepared for
Jeff Cherry
U.S. Environmental Protection Agency
Office of Transportation and Air Quality
2000 Traverwood Drive
Ann Arbor, Michigan 48105
Prepared by
Larry O'Rourke
ICF International
100 Cambridge Park Drive
Cambridge, MA 02140
-------
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Contents
1. Introduction 1-1
2. Selection of Peer Reviewers 2-1
3. Peer Review Process 3-1
4. Summary of Review Comments 4-1
4.1. Responses to Charge Questions 4-1
Appendix A. Resumes and Conflict of Interest Statements A-l
Appendix B. Charge Letter B-l
Appendix C. Kenneth Gillingham Review Comments C-l
Appendix D. David Greene Review Comments D-l
Appendix E. James Sallee Review Comments E-l
Tables
Table 2-1. Potential Reviewers 2-1
Table 2-2. Final Reviewers 2-2
ICF International i January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
ICF International ii January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Acronyms and Abbreviations
Acronym / Abbreviation Stands For
3SLS
AEO
CAFE
EPA
FHWA
GHG
ICF
NHTSA
OTAQ
S&H
UKERC
VMT
WAM
Three-Stage Least Squares
Annual Energy Outlook
Corporate Average Fuel Economy
U.S. Environmental Protection Agency
Federal Highway Administration
Greenhouse Gas
ICF International
National Highway Traffic Safety Administration
Office of Transportation and Air Quality
Small & Hymel
United Kingdom Energy Research Center
Vehicle Miles Traveled
Work Assignment Manager
ICF International
January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
ICF International iv January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymellntroduction
The Office of Transportation and Air Quality (OTAQ) of the U.S. Environmental Protection Agency (EPA)
is responsible for developing regulations to reduce the emissions of greenhouse gases (GHG) from light-
duty vehicles in the U.S. The regulatory option of encouraging the adoption of advanced technologies
for improving vehicle efficiency can result in significant fuel savings and GHG emission benefits. At the
same time, it is possible that some of these benefits might be offset by additional driving that is
encouraged by the reduced costs of operating more efficient vehicles. This so called "rebound effect",
the increased driving that results from an improvement in the energy efficiency of a vehicle, must be
determined in order to reliably estimate the overall benefits of GHG regulations for light-duty vehicles.
Dr. Ken Small, an Economist at the Department of Economics, University of California at Irvine, with
contributions by Dr. Kent Hymel, Department of Economics, California State University at Northridge,
have developed a methodology to estimate the rebound effect for light-duty vehicles in the U.S.
Specifically, rebound is estimated as the change in vehicle miles traveled (VMT) with respect to the
change in per mile fuel costs that can occur, for example, when vehicle operating efficiency is improved.
The model analyzes aggregate personal motor-vehicle travel within a simultaneous model of aggregate
VMT, fleet size, fuel efficiency, and congestion formation. The model uses three-stage least squares
(3SLS) in order to account for the endogeneity of explanatory variables. The results contain both short-
run and long-run estimates based upon lagged effects within annual data. For VMT, the behavioral
responses underlying short run effects could include changes in travel mode, discretionary trips,
destinations, or the combining of several trips into a single chain. Long-run responses might include
changes in the vehicle stock, job or residential relocations, and changes in land use.
The model is estimated using a cross-sectional, time series data set with each variable measured for 50
U.S. states, plus District of Columbia, annually for years 1966-2009. Variables are constructed from
public sources, mainly the U.S. Federal Highway Administration, U.S. Census Bureau, and U.S. Energy
Information Administration.
Since the effectiveness of regulatory efforts to reduce GHG emissions is strongly influenced not only by
the technical attributes of vehicles, but also by vehicle usage levels, it is important to assure that the
methodologies considered by the U.S. EPA for estimating VMT rebound have been thoroughly
examined. Comprehensive, objective peer reviews like the one described here are an important part of
that examination process.
This report details the peer review of the subject report, The Rebound Effect from Fuel Efficiency
Standards: Measurement and Projection to 2035 (December 24, 2013). A number of independent subject
matter experts were identified and the process managed to provide reviews and comments on the
ICF International 1-1 January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Introduction
methodology of the report. This peer review process was carried out under EPA's peer review
guidelines1.
This report is organized as follows:
« Chapter 2 details the selection of the peer reviewers
« Chapter 3 details the peer review process
« Chapter 4 summarizes the reviews
« Appendix A provides resumes and conflict of interest statements for the three selected reviewers
« Appendix B provides the charge letter sent to the selected reviewers
« Appendix C, D and E provide the actual reviews submitted by the three selected reviewers
U.S. Environmental Protection Agency, Peer Review Handbook, 3rd Edition with appendices. Prepared for the U.S. EPA by
Members of the Peer Review Advisory Group, for EPA's Science Policy Council, EPA/100/B-06/002. Available at
http://www.epa.gov/peerreview
ICF International 1-2 January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and HymelSelection of Peer Reviewers
2. Selection of Peer Reviewers
The EPA and ICF WAM compiled a list of 14 reviewers who would be capable of reviewing the subject
report. They are listed in Table 2-1.
Table 2-1. Potential Reviewers
Potential
Reviewer
David Greene
Lucas Davis
Joshua Linn
Jonathan Rubin
Sarah West
James Sallee
Kenneth
Gillingham
Chris Knittel
Mark Jacobson
David Rapson
Soren T.
Anderson
HuntAllcott
Available
Yes
No -too
busy
Yes
Yes
Yes
Yes
Yes
No response
Yes
Yes
No -too
busy
Yes
Affiliation
Senior Fellow in the Howard H. Baker,
Jr. Center for Public Policy and a
Research Professor of Civil and
Environmental Engineering, the
University of Tennessee
Associate Professor
University of California, Berkeley
Fellow (indefinite appointment),
Resources for the Future
Professor, Margaret Chase Smith
Policy Center and School of
Economics, University of Maine
Professor, Macalester College,
Economics
Assistant Professor, Harris School of
Public Policy Studies
University of Chicago
Assistant Professor of Economics,
School of Forestry & Environmental
Studies, Yale
William Barton Rogers Professor of
Energy Economics
Massachusetts Institute of Technology
Sloan School of Management
Associate Professor, University of
California
Assistant Professor of Economics, UC
Davis
Assistant Professor
Michigan State University
Department of Economics
Assistant Professor of Economics, New
York University
Degree
Ph.D., Geography and
Environmental Engineering
Ph.D., Economics
Ph.D., Economics
Ph.D., Agricultural Economics
Ph. D., Economics
Ph.D., Economics
Ph.D., Management Science &
Engineering and Economics
Ph.D., University of California,
Berkeley
Ph.D., Economics
Ph.D., Economics
Ph.D., Economics
Ph.D., Public Policy
ICF International
2-1
January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Selection of Peer Reviewers
Steve Sorrell
Yes
Senior Lecturer (SPRU - Science and
Technology Policy Research, The
Sussex Energy Group
Ph.D. by publication - Analyzing
controversies in energy policy:
the evidence for rebound
effects and global oil depletion,
Todd Litman
Yes
Executive director of the Victoria
Transport Policy Institute
Masters of Environmental
Studies
The three selected reviewers are listed in Table 2-2. Each had the necessary expertise, were available to
review the report in a timely manner and had no conflict of interest. All were agreed upon by the EPA
WAM.
Table 2-2. Final Reviewers
Reviewer
Kenneth Gillingham
Contact Information
Yale University
School of Forestry & Environmental
Studies
P: 203-436-5465
kenneth.gillingham@yale.edu
David Greene
University of Tennessee
Howard H. Baker, Jr. Center for Public
Policy
P: (865) 974-3839
dgreen32@utk.edu
Necessary
Expertise
Yes
Yes
Conflict of
Interest
No
No
James Sallee
University of Chicago
The Harris School of Public Policy Studies
P: 773-316-3480
sallee@uchicago.edu
Yes
No
Resumes and conflict of interest statements for the three reviewers can be found in Appendix A.
ICF International
2-2
January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Peer Review Process
3. Peer Review Process
Once the three reviewers had been decided upon and approved by the EPA WAM, a charge letter and
the subject report were sent to each reviewer via secure email. Shortly after distributing the charge
letter (see Appendix B) and supporting materials for the peer review, a teleconference was held
between the selected peer reviewers, the EPA WAM, EPA-identified relevant project-related staff and
ICF staff to clarify any questions the peer reviewers may have regarding the report/written materials. At
the conference call, EPA provided technical and/or background information on the particular report
under review.
During the review process, no reviewers had questions. Each reviewer provided a written peer review in
a timely manner. These were sent to ICF who forwarded them directly to the EPA WAM.
ICF managed the peer review process to ensure that each peer reviewer had sufficient time to complete
their review of the data analysis by the deliverable date specified (January 17, 2014). ICF adhered to the
provisions of EPA's Peer Review Handbook guidelines to ensure that all segments of the peer review
conformed to EPA peer review policy.
ICF International 3-1 January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Peer Review Process
ICF International 3-2 January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and HymelSummary of Review Comments
4. Summary of Review Comments
In this section, review comments from the three reviewers are summarized. Full comments (including
those in addition to the charge questions) can be found in Appendix C for Kenneth Gillingham, Appendix
D for David Greene and Appendix E for James Sallee. Responses are summarized below relative to the
charge questions.
4.1. Responses to Charge Questions
What are the merits and limitations of the authors' approach for estimating the vehicle miles
traveled (VMT) rebound effect for light-duty vehicles? Are key assumptions underpinning the
methodology reasonable? The VMT rebound effect is defined here as the change in VMT
resulting from an improvement in the light-duty efficiency.
The reviewers highlighted a number of merits to the authors' approach. All three reviewers generally
agree that authors' selection of FHWA data to be appropriate for this study. Sallee mentioned that the
aggregate data used in the report suffer from measurement problems, but due to data gaps in other
sources, the data used for this report may be the best available at this time. Other highlighted merits
include the authors' accurate understanding of the direct rebound effect and an understanding of
estimation issues, resulting in a robust and accurate estimate of the VMT rebound effect.
All three reviewers believed that the assumptions underpinning the methodology were generally
reasonable and consistent with the best methods employed in current research in this area. The
reviewers did discuss other factors that could be considered or evaluated in more depth. For example
Greene noted that the analysis omits part of the effect of increased vehicle prices on the long-run cost
per-mile of travel. An increase in the capital cost of a vehicle also affects the long-run cost of vehicle
travel via usage-induced capital depreciation. Sallee noted that the data used provides no way to model
the relationship between vehicle age and VMT.
Is the implementation of the authors' methodology appropriate for producing estimates of
the VMT rebound effect? Specifically, are the input data and the methodology used to
prepare the data appropriate? Are sound econometric procedures used? Does the model
appropriately reflect underlying uncertainties associated with the assumptions invoked and
the parameters derived in the model?
All three reviewers generally thought the authors' approach was appropriate and was representative of
best practices. They noted that the research did suffer from some data limitations that the author and
the literature more broadly were aware of. A number of tests for robustness and points for additional
clarification were suggested.
Sallee noted that most of the independent variables were not independently measured, but imputed
using methodologies that may differ across states and over time. On-road fuel economy may vary over
time, even for the same vehicle, due to changes in driving conditions, such as congestion or degree of
urbanization. While the existing time series data is the best available, there are significant changes that
ICF International 4-1 January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Summary of Review Comments
have occurred over time that affect the interpretation of the results. The authors' have documented
most of these issues.
Greene notes that "the estimates presented by S&H are based on the maintained hypothesis of
economically rational behavior, in the sense that consumers are assumed to respond to changes in fuel
cost per mile in the same way whether caused by changes in fuel price or changes in fuel economy", but
that the research also demonstrates that the consumer response to fuel economy is less than the
response to changes in fuel price, which are more salient to the consumer.
Gillingham notes that standard time series econometric approaches were not used. The paper does
account for first order autocorrelation, but second order autocorrelation was not considered, which
could introduce some bias into the standard errors.
The methodology used in this report attempts to account for asymmetric responses to
increases vs. decreased in per mile fuel costs (and fuel prices). Does the report's finding of an
asymmetric response seem reasonable given the methodology that the author's employed?
In particular, do the authors' preferred model specifications (3.21 b and 4.21 b) seem
appropriate for capturing driver response to an increase in fuel efficiency?
All three reviewers found the authors' finding of an asymmetric response to be reasonable, and that
models 3.21b and 4.21b were well chosen as the preferred models. Gillingham raises the following
question: If asymmetries come about because of the differing salience of increases and decreases in
gasoline prices, should we expect the same effects to apply for changes in vehicle fuel efficiency?
The report describes a methodology for projecting the VMT rebound effect for light-duty
vehicles forward in time. The concept of dynamic rebound is introduced to quantify the
rebound effect over the period of a vehicle lifetime, during which time the variables that
influence the rebound effect are changing. Is this methodology reasonable and appropriate,
given the inherent uncertainty in making projections about how future drivers will respond to
a change in the fuel efficiency of their vehicles?
All three reviewers agree that the dynamic rebound effect should be used to quantify the rebound
effect over the period of a vehicle lifetime. Gillingham suggests that a nonlinear extrapolation (that is
asymptotic with 0) may be more appropriate when extrapolating out as far as 2030. Greene and Sallee
agree with Gillingham that the rebound effect should not go to 0 and suggest truncating at a value
above 0. Sallee notes that it would be instructive to have the authors compare the dynamic rebound
forecast to a forecast that assumes a constant rebound over time.
Refer to Appendix C, D, and E for further details on the all the reviewers' comments.
ICF International 4-2 January 31, 2014
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Resumes and Conflict of Interest Statements
Appendix A. Resumes and Conflict of Interest Statements
ICF International A-l January 31, 2014
-------
-------
Kenneth Gillingham
CONTACT Yale University
INFORMATION School of Forestry & Environmental Studies
195 Prospect Street
New Haven, CT 06511, USA
phone: (203) 436-5465 fax: (203) 436-9135
E- mail: kenneth. gillingham@yale.edu
WWW: www.yale.edu/gillingham
RESEARCH
INTERESTS
Environmental & Energy Economics, Industrial Organization, Public Economics, Econometrics,
Technological Change, Transportation Economics, Energy & Climate Policy Modeling.
CURRENT
POSITION
Yale University, New Haven, CT USA
Assistant Professor of Economics, School of Forestry & Environmental Studies July 2011-present
Secondary appointment, Department of Economics May 2012-present
Secondary appointment, School of Management June 2013-present
EDUCATION Stanford University, Stanford, CA USA
Ph.D., Management Science & Engineering and Economics, 2011
Dissertation: "The Consumer Response to Gasoline Price Changes: Empirical Evidence and
Policy Implications"
Committee: Jim Sweeney, Larry Goulder, Matt Harding, John Weyant, Jon Levin (orals chair)
Fields: Public & Environmental Economics, Industrial Organization, Econometrics
M.S., Statistics, 2010
M.S., Management Science & Engineering (Economics & Finance), 2006
Dartmouth College, Hanover, NH USA
A.B., Economics and Environmental Studies (minor in Earth Sciences), 2002
PREVIOUS California Air Resources Board, Sacramento, CA USA
EMPLOYMENT Economist (Graduate Student Assistant)
Stanford University, Stanford, CA USA
Research Assistant for Prof. Matt Harding, Stanford Economics Department
Research Assistant for Prof. John Weyant, Stanford Energy Modeling Forum
Research Assistant for Prof. Jim Sweeney, Precourt Energy Efficiency Center
Fulbright New Zealand, University of Auckland, Auckland, New Zealand
Fulbright Fellow
White House Council of Economic Advisers, Washington, DC USA
Fellow for Energy and Environment
Resources for the Future, Washington, DC USA
Research Assistant
2011
2008-2010
2008
2004-2006
2007
2005
2002-2004
Dartmouth College, Hanover, NH USA
Research Assistant for Prof. Karen Fisher- Vanden
1998-2002
-------
WORKING PAPERS Gillingham, K. Selection on Anticipated Driving and the Consumer Response to Changing Gasoline
Prices (previously titled: How Do Consumers Respond to Gasoline Price Shocks? Heterogeneity
in Vehicle Choice and Driving Behavior)
Gillingham, K., M. Kotchen, D. Rapson, G. Wagner, The Rebound Effect and Energy Efficiency
Policy, In preparation for Review of Environmental Economics & Policy
WORK-IN-
PROGRESS
Gillingham, K. The Economics of Fuel Economy Standards versus Feebates
Learning-by-Doing in the Solar Photovoltaic Industry (with Bryan Bellinger)
Automaker Responses to Fuel Economy Standards (with Antonio Bento, Kevin Roth, Yiwei Wang)
The Economic Efficiency of Renewable Portfolio Standards in the Presence of Cap-and-Trade (with
Arthur van Benthem)
Consumer Welfare and Environmental Effects of Registration Fees and Driving Fees in Denmark
(with Bertel Schjerning, Fedor Iskhakov, John Rust, and Anders Munk-Nielsen)
HOV Stickers and the Consumer Adoption of Hybrids: Evidence from California (with Calanit
Kamala)
A Dynamic Model of Household Vehicle Choice and Usage (with David Rapson)
Salience and Upstream versus Downstream Cap-and-Trade
The Geographic and Demographic Distributional Effects of Gasoline Taxes
Uncertainty in Integrated Assessment Models of Climate Change Policy (with Bill Nordhaus)
PUBLICATIONS Gillingham, K. and K. Palmer (2014) Bridging the Energy Efficiency Gap: Policy Insights from
Economic Theory and Empirical Analysis. Review of Environmental Economics & Policy, forth-
coming.
Gillingham, K. (2013) Identifying the Elasticity of Driving: Evidence from a Gasoline Price Shock
in California, Regional Science & Urban Economics, forthcoming.
Yeh, S., G. Mishra, G. Morrison, J. Teter, R. Quiceno, and K. Gillingham (2013) Long-Term Shifts
in Lifecycle Energy Efficiency and Carbon Intensity. Environmental Science & Technology, 47(6):
2494-2501.
Gillingham, K., M. Kotchen, D. Rapson, G. Wagner (2013) The Rebound Effect is Over-played.
Nature, 493: 475-476.
Bellinger, B. and K. Gillingham (2012) Peer Effects in the Diffusion of Solar Photovoltaic Panels.
Marketing Science, 31(6): 900-912.
Gillingham, K., M. Harding, and D. Rapson (2012) Split Incentives in Household Energy Consump-
tion. Energy Journal, 33(2): 37-62.
Gillingham, K. and J. Sweeney (2012) Barriers to Implementing Low Carbon Technologies. Climate
Change Economics, 3(4), 1-25.
Gillingham, K. and J. Sweeney (2010) Market Failure and the Structure of Externalities. In: Har-
nessing Renewable Energy, Boaz Moselle, Jorge Padilla, Richard Schmalensee (eds). RFF Press.
Leaver, J. and K. Gillingham (2010) Economic Impact of the Integration of Alternative Vehicle
Technologies into the New Zealand Vehicle Fleet. Journal of Cleaner Production, 18: 908-916.
Gillingham, K., R. Newell, and K. Palmer (2009) Energy Efficiency Economics and Policy. Annual
-------
Review of Resource Economics, 1: 597-619. Reprinted in Italian in Energia (2010).
Gillingham, K. (2009) Economic Efficiency ol Solar Hot Water Policy in New Zealand. Energy
Policy, 37(9): 3336-3347.
Leaver, J., L. Leaver, and K. Gillingham (2009) Assessment of Primary Impacts of a Hydrogen
Economy in New Zealand using UNISYD. International Journal of Hydrogen Energy, 34(7):
2855-2865.
Gillingham, K., R. Newell, and W. Pizer (2008) Modeling Endogenous Technological Change for
Climate Policy Analysis. Energy Economics, 30(6): 2734-2753.
van Benthem, A., K. Gillingham, and J. Sweeney (2008) Learning-by-Doing and the Optimal Solar
Policy in California. Energy Journal, 29(3): 131-151.
Gillingham, K., S. Smith, and R. Sands (2008) Impact of Bioenergy Crops in a Carbon Constrained
World: An Application of the MiniCAM Linked Energy-Agriculture and Land Use Model. Mit-
igation and Adaptation Strategies for Global Change, 13(7): 675-701.
Safirova, E., K. Gillingham, and S. Houde (2007) Measuring Marginal Congestion Costs of Urban
Transportation: Do Networks Matter? Transportation Research A, 41(8): 734-749.
Gillingham, K., R. Newell, and K. Palmer (2006) Energy Efficiency Policies: A Retrospective Ex-
amination. Annual Review of Environment and Resources, 31: 193-237.
Shih, J-S, W. Harrington, W. Pizer, and K. Gillingham (2006) Economies of Scale in Community
Water Systems. Journal of American Water Works Association, 98(9): 100-108.
Safirova, E., P. Nelson, W. Harrington, K. Gillingham, and A. Lipman (2005) Choosing Congestion
Pricing Policy: Cordon Tolls vs. Link-Based Tolls. Transportation Research Record, 1932: 169-
177.
Safirova, E., I. Parry, P. Nelson, W. Harrington, K. Gillingham, D. Mason (2004) Welfare and Dis-
tributional Effects of HOT Lanes and Other Road Pricing Policies in Metropolitan Washington,
DC. Research in Transportation Economics, 9(1): 179-206.
REPORTS fe OTHER Bellinger, B. and K. Gillingham (2012) Do Peer Effects Matter? Assessing the Impact of Causal
PUBLICATIONS Social Influence on Solar PV Adoption, Photovoltaics International, 17: 160-165.
Priedland, A. and K. Gillingham (2010) Carbon Accounting is a Tricky Business. Letter to the
Editor, Science, 327(5964): 411-412.
Sweeney, J., J. Weyant, K. Gillingham, et al. (2008) Analysis of Measures to Meet the Requirements
of California's Assembly Bill 32. Precourt Institute for Energy Efficiency Working Paper.
Gillingham, K. (2007) Hydrogen Internal Combustion Engine Vehicles: A Prudent Intermediate
Step or a Step in the Wrong Direction? Stanford Global Climate and Energy Project Working
Paper.
Gillingham, K., R. Newell, and K. Palmer (2006) The Effectiveness and Cost of Energy Efficiency
Programs. In: The RFF Reader in Environmental and Resource Policy, Wallace Gates (ed).
RFF Press. 193-201.
Safirova, E., W. Harrington, P. Nelson, and K. Gillingham (2004) Are HOT Lanes a Hot Deal?
Analyzing the Potential of HOV to HOT Lanes Conversion in Northern Virginia. RFF Issue
Brief 03-03.
Nelson, P., E. Safirova, and K. Gillingham (2003) Revving up the Tax Engine: Gas Taxes and the
DC Metro Area's Transportation Dilemma. RFF Issue Brief 03-05.
-------
GRANTS
HONORS AND
AWARDS
TEACHING
"The Influence of Novel Behavioral Strategies in Promoting the Diffusion of Solar Energy," US
Department of Energy, PI, 2013-2015 ($1,899,978)
"Density, Walkability, and VMT," Yale Center for Business and the Environment Sobotka Research
Fund, PI, 2013-2014 ($10,100)
"Deep Dive Solar Cost Analysis," Lawrence Berkeley National Laboratory/US Department of En-
ergy, PI, 2013-2015 (Yale budget: $74,924)
"Modeling Household and Transportation Vehicle Choice and Usage," California Air Resources
Board, co-Pi with Dave Rapson, Chris Knittel, and Pat Mokhtarian, 2012-2014 ($300,000)
"Sunrise New England," US Department of Energy, co-Pi with Stuart DeCew, 2012-2014 (Yale
budget: $215,000)
"The Consumer Response to Gasoline Price Changes," Stanford Institute for Economic Policy Re-
search (SIEPR) Grant, 2010 ($10,000)
"The Consumer Response to Gasoline Price Changes," Shultz Graduate Student Fellowship in Eco-
nomic Policy, SIEPR, 2010 ($4,000)
"Economics of New Zealand Solar Distributed Generation," Fulbright Fellowship, 2007
"The Effect of Income and Congestion on the Rebound Effect of CAFE Standards," US Environ-
mental Protection Agency STAR Fellowship, 2006-2009 ($111,000)
Heitz Graduate Fellowship, Stanford University, 2006
Battelle Memorial Institute Fellowship, 2001 ($6,000)
Full Member, Sigma Xi 2011
Dennis O'Brian Best Student Paper Award, US Association for Energy Economics 2010
Thesis and Research Essay Publication Scholarship, University of Auckland 2008
Outstanding Teaching Assistant Award, Stanford Economics 2006
National Science Foundation Graduate Fellowship, Honorable Mention 2006
American Water Works Association Best Paper Award 2006
Departmental Honors, Dartmouth Economics 2002
Departmental High Honors, Dartmouth Environmental Studies 2002
Associate Member, Sigma Xi 2002
First Prize, Dartmouth Sigma Xi Senior Thesis Competition 2002
Yale University
2012-2013: Ph.D. Environmental and Energy Economics, Energy Economics and Policy Analysis
(Masters), Yale Environmental Economics Seminar.
2011-2012: Economics of the Environment (Masters), Yale Environmental Economics Seminar.
Stanford University, Teaching Assistant
2009-2010: Ph.D. Microeconomics, Introductory Econometrics, Natural Resource Economics (Grad-
uate) .
2008-2009: Transportation Policy (Graduate), Energy Policy Analysis (Graduate), Natural Resource
Economics (Graduate).
2007-2008: Energy & Environmental Policy Analysis (Graduate), Climate Policy Analysis (Gradu-
-------
ADVISING
PRESENTATIONS
ate), Natural Resource Economics (Graduate).
2005-2006: Principles of Economics
Ph.D. Primary Advisor
Hao Deng (FES 2nd year), Jesse Burkhardt (FES 3rd year; co-advised with Matthew Kotchen)
Ph.D. Committee Member
Laura Bakkensen (FES 5th year), Peter Christensen (FES 5th year), Nathan Chan (FES 5th year),
Rich Langford (Yale econ 5th year), Alan Jenn (Carnegie-Mellon 4rd year),
Anders Munk-Nielsen (Copenhagen 3rd year), Nikki Springer (FES 3rd year)
M.E.Sc Advisor
Hilary Staver (2nd year), Paige Weber (2nd year)
Masters Independent Research
2012-2013: Vijeta Jangra (MEM '13)
2011-2012: Howard Chang (MEM/MBA '12), Dustin Schinn (MEM '13), Peter Baum (MEM '13)
Undergraduate Senior Thesis Advising
Ana Grajales (economics '13), Daniel Cheng (math-econ '13)
2013 (scheduled): AEA Meetings (discussant); Modeling Uncertainty Project Meeting (New Haven,
CT); FES/SOM Yale Environmental Economics Seminar; Carnegie-Mellon University; Vil-
lanova Law; Arizona State University Economics of Water Conference (Keynote); Interna-
tional Industrial Organization Conference (Cambridge, MA); AERE Summer Conference
(Banff, AB); Empirical Methods in Energy Economics Workshop (Carlton University, Ot-
tawa); DOE Sunshot Kick-off Meeting; EMF Workshop on Climate Change Impacts and In-
tegrated Assessment (Smowmass, CO); Stanford Institute for Theoretical Economics (SITE)
Advances in Environmental Economics Workshop; Columbia University SIPA; Indiana Uni-
versity Kelley School of Business; Tsinghua University Institute of Energy, Environment, and
Economy; Fudan University Economics; Indiana University SPEA; Behavior, Energy, and
Climate Change (BECC) Conference.
2012: AEA Meetings; Yale FES Seminar Series; Lawrence Berkeley National Laboratory/DOE
Sunshot Initiative Workshop; Triangle Resource & Environmental Economics seminar (Duke,
NCSU, UNC); Indiana University Kelley School of Business/Economics/SPEA; University
of Massachusetts Amherst Resource Economics; Rice University Economics; Texas A&M
Economics; UC Santa Cruz Economics; Naval Postgraduate School Economics; UC Santa
Barbara Economics/Bren School; ETH Zurich Economics; University of Lugano Economics;
AERE Summer Conference (Asheville, NC); EMF Workshop on Climate Change Impacts and
Integrated Assessment (Smowmass, CO); Connecticut Clean Energy Finance & Investment
Authority; UC Berkeley/Lincoln Institute of Land Policy Conference; University of Colorado
Boulder Economics; University of Wyoming Economics; US Association for Energy Economics
(Austin, TX); University of Connecticut ARE; University of Copenhagen Economics.
2011: University of Maryland AREc; Indiana University SPEA; UC Davis Economics; University
of Arizona Economics; Arizona State University Economics; University of Illinois Urbana-
Champaign Finance; University of Notre Dame Economics; Yale University FES; Informing
Green Markets Conference (University of Michigan); Cowles Foundation Structural Microe-
conometrics Conference (Yale University); Empirical Methods in Energy Economics Work-
shop (Southern Methodist University); Harvard Seminar on Environmental Economics and
Policy; Re-examining the Rebound Effect in Energy Efficiency Workshop (Environmental
Defense Fund); RFF-Stanford Workshop on the Next Round of Climate Economics and Pol-
-------
REFEREE SERVICE
icy Research (Washington, DC); US Association for Energy Economics (Washington, DC);
Workshop on Environmental and Transportation Policies to Mitigate Climate Change (New
York University); Religare Capital Markets (Singapore); Behavior, Economics, and Energy
Panel (National University of Singapore Energy Studies Institute); University of Copenhagen
Economics.
2010: UC Berkeley Energy Institute; NBER EEE Summer Institute; World Congress Env & Resource
Economists (Montreal); US Association for Energy Economics (Calgary); Behavior, Energy &
Climate Change (BECC) Conference; 12th Occasional California Workshop on Environmental
and Resource Economics; UC Davis ARE; Resources for the Future.
2009: UC Berkeley ARE; Stanford IO Workshop; DOE El A Advisory Council; US Association for
Energy Economics (San Francisco); UC Energy Institute CSEM Conference.
2008: UC Davis ITS; Victoria University, New Zealand.
2007: University of Auckland Energy Centre, New Zealand; Massey University, New Zealand; Inter-
national Association for Energy Economics (Wellington, New Zealand).
2006: Dartmouth College Workshop on Technological Change & Environment.
2004: Transportation Research Board Annual Meeting.
Economics Journals: American Economic Journal-Applied, American Economic Review, Amer-
ican Journal of Agricultural Economics, B.E. Journal of Economic Analysis & Policy, Climate
Change Economics, Economics of Energy & Environmental Policy, Ecological Economics, Energy
Economics, Energy Journal, Journal of the Association of Environmental and Resource Economists,
Journal of Economic Surveys, Journal of Environmental Economics & Management, Journal of
Institutional & Theoretical Economics, Journal of Public Economics, Management Science, Oxford
Economic Papers, Quarterly Journal of Economics, RAND Journal of Economics, Regional Science
& Urban Economics, Resource & Energy Economics, Review of Environmental Economics & Policy,
Scandinavian Journal of Economics, Southern Economic Journal, The Manchester School.
Environment/Engineering/Policy/Science Journals: Building Research, Cityscape, Climatic
Change, Energies, Energy, Energy & Fuels, Energy Efficiency, Energy Policy, Environment Develop-
ment and Sustainability, Environmental Modeling & Assessment, Environmental Research Letters,
Environmental Science & Technology, Global Environmental Change, International Journal of Sus-
tainable Transportation, Journal of Cleaner Production, Journal of Environment & Development,
Journal of Industrial Ecology, Journal of Policy Analysis & Management, Journal of Sustainable
Forestry, Mitigation & Adaptation Strategies for Global Change, Science, Transportation Research
A, Utilities Policy.
REVIEW SERVICE
Alfred P. Sloan Foundation, Alliance for Research on Corporate Sustainability, KU Leuven, MIT
Press, National Academy of Sciences Transportation Research Board, National Science Foundation,
Swiss National Science Foundation.
PROFESSIONAL
SERVICE
US Department of Agriculture NIFA Bioenergy Policy expert review panel (Apr 19-20, 2012), Yale
Climate & Energy Institute (YCEI) Steering Committee (2013-present), co-organizer of Modeling
Uncertainty in Climate Policy Workshop (Feb 4, 2013 at Yale), US Department of Energy Review
Panel (April 26, 2013), co-organizer of Northeast Workshop on Energy Policy & Environmental
Economics (May 10-11, 2013 at Cornell), US Department of Agriculture NIFA Climate Change
Adaptation expert review panel (Aug 1-2, 2013)
PROFESSIONAL
AFFILIATIONS
American Economic Association (AEA), Association of Environmental and Resource Economists
(AERE), United States Association for Energy Economics (USAEE), Econometric Society, Industrial
Organization Society.
-------
RESEARCH CITED IN Wall Street Journal The Numbers Guy Blog: "The Rebound Effect," May 26, 2009.
THE MEDIA Grist: "Making Buildings More Efficient: Looking Beyond Price," Oct 23, 2009.
Grist: "Solar Power is Contagious," Apr 5, 2011.
Energy Matters: "Australia's Home Solar Power Revolution and the Viral Effect," Apr 6, 2012.
Wired: "Solar Panels are Contagious," Apr 12, 2011.
The David Sirota Show AM 760: "Solar Power is Contagious," Apr 25, 2011.
Connecticut Public Radio (WNPR): "Where we Live: Future of Natural Gas" Aug 8, 2011.
Yale Daily News: "City Wins Transportation Grant," Oct 20, 2011.
The Straits Times (Singapore): "To Save the Earth, Know Human Nature," Nov 20, 2011.
Business Times: "Cutting Green Path Via Behavioural Economics," Nov 21, 2011.
Washington Post: "Solar Power is Contagious - But Not Quite Virulent," Dec 5, 2011.
Forbes: "Keeping Up With the Greens: Neighborhood Solar is Contagious," Dec 9, 2011.
Yale Daily News: "Nuclear's Back with New Clarity," Feb 10, 2012.
CO2 Scorecard: "Non-Conundrum of the Prius Fallacy," Mar 26, 2012.
Climate Progress Blog by Joe Romm: "Debunking the Fallacy of the Prius Rebound Effect," Mar
26, 2012.
CleanTechnica Blog: "Prius Rebound Effect Wrong," Mar 28, 2012.
Wall Street Journal SmartMoney: "For Appliances, Does Energy Efficiency Sell?" Oct 16, 2012.
CleanTechnica.com: "If Your Neighbor Has Solar Panels, You're More Likely to Go Solar," Oct 18,
2012.
Wired UK: "Enthusiasm for Solar Panels is Contagious," Oct 19, 2012.
Albuquerque Express: "Use of Solar Panels Popularized by Example," Oct 19, 2012.
Alternative Energy Blog: "Solar Power Tends to Go Viral, New Report Suggests," Oct 19, 2012.
India Talkies: "Use of Solar Panels Popularised by Example," Oct 19, 2012.
CleanEnergyAuthority.com: "Go Solar, it's the Neighborly Thing to Do," Oct 19, 2012.
Solar Industry Magazine: "New Study Shows Solar Installations Are Contagious in Neighborhoods,"
Oct 19, 2012.
EarthTechling.com: "The Solar Power Bug: Has Your Neighborhood Caught it?" Oct 19, 2012.
R&D Magazine: "Study: Solar Power is Contagious," Oct 19, 2012.
Environmental News Network: "Solar Power Adoption is Contagious," Oct 22, 2012.
Huffington Post: "Solar Panel Installations More Likely In Homes With Energy Efficient Neighbors,"
Oct 23, 2012.
ClimateWire: "Is Renewable Energy Contagious: Research Shows a 'Peer Effect'," Nov 5, 2012.
Yale Daily News: "Sandy Link to Climate Change Questioned," Nov 9, 2012.
AOL Energy: "The Psychology of Small-Scale Solar," Nov 19, 2012.
New York Times: "Solar Industry Borrows a Page, and a Party, from Tupperware," Dec 1, 2012.
Yahoo News: "Economist: Rebound Effect of Energy-Efficient Cars is Overplayed," Jan 23, 2013.
Scientific American: "Does Increased Energy-Efficiency Just Spark Us to Use More?" Jan 23, 2013.
Central Valley Business Times: " 'Rebound Effect' Has Little Bounce," Jan 23, 2013.
Sierra Daily: "Energy Efficiency? Why Bother?" Jan 23, 2013.
Arstechnica: "How Badly Does the Rebound Effect Undercut Energy Efficiency?" Jan 24, 2013.
Phys.org: "Researchers Argue Energy Policy Rebound Effect is Overestimated," Jan 24, 2013.
Grist (David Roberts): "Why Are Greens So Defensive About the Rebound Effect," Jan 24, 2013.
Huffington Post: "Nature: The Rebound Effect is Overplayed," Jan 24, 2013.
R&D Magazine: "The 'Rebound' Effect of Energy-Efficient Cars Overplayed," Jan 24, 2013.
Scaling Green: "New Study: Energy Efficiency Negative 'Rebound Effect' Greatly Exaggerated," Jan
24, 2013.
Revkin.net: "Rebound is Real, But Limited," Jan 24, 2013.
The Naked Scientists, Science News: "Energy Efficiency on the Rebound," Jan 24, 2013.
Swiss National Radio: "Rebound Effect," Jan 25, 2013.
New Haven Register: "Yale Receives $1.9 million Solar Grant," Jan 30, 2013.
Connecticut Public Radio (WNPR): "Yale Gets Award to Help Grow Solar Energy," Feb 20, 2013.
Yale Daily News: "Green Expectations: Yale's Energy Investments Struggle," Mar 26, 2013.
Yale Scientific Magazine: "Solar Energy: Sink or Spread-Professor Gillingham's Study on the Seal-
-------
ability of Solar Energy," April 5, 2013.
Yale Scientific Magazine: "Yale Professor Discusses the Economics of Conservation," May 11, 2013.
Connecticut Public Radio (WNPR): "A New Gas Tax, But What's it Paying For?" Jul 1, 2013.
Washington Square News: "Stern, Yale Professors Team Up To Research Solar Energy," October 1,
2013.
-------
Contractor orporvtcd r H^t.
n« Contra
Subcontract/?** R*v»*w«r: K*nn«0) GHUngham
In accordant*
best lowtafet and brtef. that
A No actual rt«t«>Mt»
An acluat or potential conf !.• •/
Subcontractor/
aland off* 'HUCBr*!*
.iracto- Aidant'
po1
Sub>
-------
-------
DAVID L. GREENE
Home: 212 Way Station Trail • Farragut, Tennessee 37922 • (865) 966-0891
Work: Oak Ridge National Laboratory • National Transportation Research Center • 2360 Cherahala Boulevard •
Knoxville, Tennessee 37932 • (865) 946-1310
PERSONAL
Born: November 18, 1949, New York, New York
Married, two children
EDUCATION
THE JOHNS HOPKINS UNIVERSITY
Ph.D., Geography and Environmental Engineering, 1973-78
UNIVERSITY OF OREGON
MA., 1972-73
COLUMBIA UNIVERSITY
B.A., 1967-71
EMPLOYMENT
UNIVERSITY OF TENNESSEE, KNOXVILLE 2010-PRESENT
1/2010-Present Senior Fellow, HowardH. Baker, Jr. Center for Public Policy
10/2013-Present Research Professor, Department of Civil and Environmental Engineering
1/2010-10/2013 Research Professor, Department of Economics
INSTITUTE FOR TRANSPORTATION STUDIES, UNIVERSITY OF CALIFORNIA, DAVIS 2008-2009
9/2008-6/2009 Visiting Research Faculty
OAK RIDGE NATIONAL LABORATORY (ORNL) 1977-PRESENT
1999-Present Corporate Fellow, Oak Ridge National Laboratory
1989-1999 Senior Research Staff Member II and Manager of Energy Policy Research
Programs, Center for Transportation Analysis
1988-1989 Senior Research Analyst, Office of Policy Integration, U.S. Department of
Energy (On assignment from ORNL)
1987-1988 Head, Transportation Research Section
1984-1987 Senior Research Staff Member I
1982-1984 Research Staff Member
1980-1982 Leader, Transportation Energy Group
1977-1980 Research Associate
AWARDS AND HONORS
Distinguished Career Service Award, U.S. Department of Energy, Energy Efficiency and Renewable
Energy, 2013
2012 Roy W. Cram Award for Distinguished Achievement, Transportation Research Board of the National
Research Council
2011 DOE Vehicle Technologies Program R&D Award, U.S. Department of Energy
2011 Edward L. Ullman Award, Association of American Geographers
2009 Alliance to Save Energy, Energy Efficiency Hall of Fame
2008 Science Communicator Award, UT-Battelle
David L. Greene 1 March 2013
-------
Recognition by the Intergovernmental Panel on Climate Change for Contributions to the Award of the 2008
Nobel Peace Prize to the IPCC
2007 Department of Energy Hydrogen Program R&D Award (with P.N. Leiby)
Barry D. McNutt Award for Excellence in Automotive Policy Analysis, Society of Automotive Engineers,
2007
Member Emeritus, Transportation Research Board Committee on Alternative Fuels, 2006
Barry D. McNutt Award for best paper of 2004, Energy Committee, Transportation Research Board
Lifetime National Associate of the National Academies, 2002
UT-Battelle Award for Excellence in Science and Technology, 2001
Oak Ridge National Laboratory Significant Event Award, 2001
Corporate Fellow of Oak Ridge National Laboratory, 1999
Outstanding Paper of 1999, The Energy Journal, International Association for Energy Economics
Lockheed-Martin Significant Event Award, 1999
Member Emeritus, Transportation Research Board Committee on Transportation Energy, 1998
Lockheed-Martin Significant Event Award, 1996
Distinguished Service Certificate, Transportation Research Board, 1993
ORNL Special Achievement Award, 1991
Distinguished Service Certificate, Transportation Research Board, 1989
Energy Specialty Group Paper Award, Association of American Geographers, 1986
ORNL Special Recognition Award, Oak Ridge National Laboratory, 1986
Technical Achievement Award, Martin Marietta Energy Systems, 1985
Pyke Johnson Award, Transportation Research Board, 1984
PROFESSIONAL ACTIVITIES
Board of Directors, American Council for an Energy Efficiency Economy, 2010-2013
Board of Advisors, Institute for Transportation Studies, University of California, Davis
Editorial Advisory Board, Transportation Research Part D, 1996-present
Editorial Board Member, Energy Policy, 2001-present
Editorial Board Member, Journal a/Transportation and Statistics, 2001-2006, 2011 -present
Editorial Board Member, Transportation Quarterly, 1999-2005
Editor-in-Chief, Journal of Transportation and Statistics, 1997-2000
Editorial Board Member, Macmillan Encyclopedia of Energy, 1998-2001
Editorial Advisory Board, Transportation Research A, 1986-1997
National Research Council
Transportation Research Board Standing Committees:
Committee on Transportation and Sustainability, Member, 2006-present
Committee on Energy, A1F01, Chairman 1983-1986, 1986-1990; Member, 1993-1998;
Member Emeritus, 1999-present
Subcommittee on Forecasting Transportation Energy Demand,
A1F01(2), Chairman, 1982-1983
Section F, Energy and Environmental Concerns, Chairman, 1990-1992
Committee on Alternative Fuels, A1F05, Member, 1993-2006,
Member Emeritus, 2006-present
Task Force on Freight Transportation Data, A1B51, Secretary, 1989-1996
Committee on Transportation Information Systems and Data Requirements,
Member, 1983-1986, 1986-1989
Ad Hoc Committees:
Committee on Assessment of Technologies for Improving Fuel Economy of Light-Duty
Vehicles - Phase 2, 2012-2015
Committee for Research Perspectives on Sustainable Energy and Transportation: A
Conference, 2012-2013
Special Task Force on Climate Change and Energy, 4/15/2012-4/14/2015
Committee on Transitions to Alternative Vehicles and Fuels, 2011-2012
Special Task Force on Energy and Climate Change, 2008-2009
Committee on the Assessment of Fuel Economy Technologies for Light-Duty Vehicles,
2007-2010
Planning Group for Workshop on Issues Related to Peaking of Global Oil Production, 2005
Committee on State Practices in Setting Mobile Source Emissions Standards, 2004-2006
David L. Greene 2 March 2013
-------
Chair, Committee for the Symposium on Introducing Sustainability into Surface
Transportation Planning, 2003-2004
Panel on Combating Global Warming through Sustainable Surface Transportation Policy,
TCRP Project Panel H-21A, 2002-2005
Committee on Effectiveness and Impacts of Corporate Average Fuel Economy (CAFE)
Standards, 2001
Committee for the Study of the Impacts of Highway Capacity Improvements on Air Quality
and Energy Consumption, 1993-1994
Committee on Fuel Economy of Automobiles and Light Trucks, Energy Engineering Board,
Commission on Engineering and Technical Systems, 1991-1992
Committee for the Study of High-Speed Surface Transportation in the United States, 1990
Planning Group on Strategic Issues in Domestic Freight Transportation, 1990
Steering Committee for Conference on Transportation, Urban Form, and the Environment,
1990
National Cooperative Highway Research Program, Panel on "Evaluating Alternative Methods
of Highway Finance," 1991-1992
Intergovernmental Panel on Climate Change
Lead Author, Working Group III, Fourth Assessment Report, 2007
Lead Author, Working Group III, Third Assessment, 2001
Lead Author, Working Group III, Aviation and the Global Atmosphere, 1999
Principal Lead Author, Working Group II, Second Assessment Report, 1995
Association of American Geographers
Board of Directors, Transportation Specialty Group, 1989-1991
Secretary-Treasurer, Transportation Geography Specialty Group, 1980-1982
Editor, Transportation Geography Newsletter, 1980-1982
Society of Automotive Engineers, member, 1985-present
International Association for Energy Economics, member
Consulting
MacroSys for U.S. Bureau of Transportation Statistics, 2013
Rand Corporation, 2012-2013
International Council for Clean Transportation, 2011-present
International Transport Forum, 2007
Addx Corporation, 2007
United Nations Framework Convention on Climate Change, 2007
Securing America's Future Energy, 2007
Center for Clean Air Policy, 2007
Pollution Probe Canada, 2006-2007
The Energy Foundation China Project, 2005-2011
The Pew Center on Global Climate Change, 2004-2012
Eno Transportation Foundation, 1991-1996
Transportation Research Board, 1996-1997
BOOKS
and D.W. Jones and Mark Delucchi, eds., The Full Costs and Benefits of Transportation, Springer-Verlag,
Heidelberg, 1997.
Transportation and Energy, Eno Foundation for Transportation, Lansdowne, Virginia, 1996.
and D. J. Santini, eds., Transportation and Global Climate Change, American Council for an Energy
Efficient Economy, Washington, DC, 1993.
ARTICLES IN PROFESSIONAL JOURNALS
D.L. Greene, S. Park and C. Liu, 2013. "Analyzing the Transition to Electric Drive Vehicles in the U.S.",
Futures, published online, 6 November 2013, http://dx.doi.0rg/10.1016/i.futures.2013.07.003 .
David L. Greene 3 March 2013
-------
and Z. Lin and J. Dong, 2013. "Analyzing the Sensitivity of Hydrogen Vehicle Sales to Consumers'
Preferences", International Journal of Hydrogen Energy, October 25, 2013,
10.1016/j.ijhydene.2013.08.099
and D.H. Evans and J. Hiestand, "Survey evidence on the willingness of U.S. consumers to pay for
automotive fuel economy", Energy Policy, vol. 61, pp. 1539-1550, 2013.
Zhenhong Lin, Jing Dong, David L. Greene, Hydrogen vehicles: Impacts of DOE technical targets on
market acceptance and societal benefits, International Journal of Hydrogen Energy, vol. 38 (2013) pp.
7973-7985.
Z. Lin, J. Dong, C. Liu and D.L. Greene, "Estimation of Energy Use by Plug-in Hybrid Electric Vehicles:
Validating Gamma Distribution for Representing Random Daily Driving Distance", Transportation
Research Record, No. 2287, pp. 37-43, Transportation Research Board, Washington, D.C., 2012.
G. Upreti, D.L. Greene, K.G. Duleep and R. Sawhney, "Fuel cells for non-automotive uses: Status and
prospects", International Journal of Hydrogen Energy, volume 37, issue 8, pp. 6339-6348, 2012.
"Rebound 2007: Analysis of National Light-Duty Vehicle Travel Statistics", Energy Policy, vol. 41, pp. 14-
28,2012.
C. Liu, E.G. Cooke, D.L. Greene and D.S. Bunch, "Feebates and Fuel Economy and Emissions Standards:
Impacts on Fuel Use in Light-Duty Vehicles and Greenhouse Gas Emissions," Transportation Research
Recorded. 2252, pp. 23-30, Journal of the Transportation Research Board, Washington, D.C., 2011.
Z. Lin and D.L. Greene, "Assessing Energy Impact of Plug-in Hybrid Electric Vehicles: Significance of
Daily Distance Variation over Time and Among Drivers", Transportation Research Record No. 2252, pp.
99-106, Journal of the Transportation Research Board, Washington, D.C., 2011.
Z. Lin and D.L. Greene, "Promoting the Market for Plug-in Hybrid and Battery Electric Vehicles: The Role
of Recharge Availability," Transportation Research Record No. 2252, pp. 49-58, Journal of the
Transportation Research Board, Washington, DC, 2011.
"What's Greener than a VMT Tax? The Case for an Indexed Energy User Fee to Finance U.S. Surface
Transportation," Transportation Research D-Environment, vol. 16, pp. 451-458,2011.
"Uncertainty, Loss Aversion and Markets for Energy Efficiency", Energy Economics, vol. 33, pp. 608-616,
2011.
Z. Lin and D.L. Greene, Predicting Individual On-road Fuel Economy Using Simple Consumer and
Vehicle Attributes, SAE Technical Paper Series No. 11SDP-0014, Society of Automotive Engineers,
Warrendale, PA, April 12, 2011.
and P.R. Boudreaux, D.J. Dean, W. Fulkerson, A.L. Gaddis, R.L. Graham, R.L. Graves, J.L. Hopson,
P. Hughes, M.V. Lapsa, T.E. Mason, R.F. Standaert, T.J. Wilbanks and A. Zucker, "The Importance of
Advancing Technology to America's Energy Goals," Energy Policy, vol. 38, no. 8, pp. 3886-3890, March
2010.
Rubin, J., P.N. Leiby and D.L. Greene, "Tradable Fuel Economy Credits: Competition and Oligopoly,"
Journal of Environmental Economics and Management, vol. 58, no. 3, pp. 315-328, 2009.
"Measuring Energy Security: Can the United States Achieve Oil Independence?" Energy Policy, 2010, Vol.
38, No. 4, pp. 1614-1621.
"Feebates, Footprints and Highway Safety," Transportation Research Part D, vol. 14, pp. 375-384, 2009.
David L. Greene 4 March 2013
-------
"Vehicles and E85 Stations Needed to Achieve to Achieve Ethanol Goals," Transportation Research
Record No. 2058, pp. 172-178.
and P.N. Leiby, P.D. Patterson, S.E. Plotkin and M. Sing, "Oil Independence: Achievable National Goal or
Empty Slogan?" Transportation Research Record, No. 2017, pp. 47-53, Washington, DC, 2007.
and J.L. Hopson, R. Goeltz and J. Li, "Analysis of In-Use Fuel Economy Shortfall Based on Voluntarily
Reported Mile-per-Gallon Estimates," Transportation Research Record, No. 1983, pp. 99-105,2007.
Leiby, P.N., D.L. Greene, D. Bowman and E. Tworek, "Systems Analysis of Hydrogen Transition with
HyTrans," Transportation Research Record, No. 1983, pp. 129-139,2007.
and J.L. Hopson and J. Li, "Have We Run Out of Oil Yet? Oil Peaking Analysis from an Optimist's
Perspective," Energy Policy, vol. 34, pp. 515-531, 2006.
S. Ahmad and D.L. Greene, "The Effect of Fuel Economy on Automobile Safety: A Reexamination,"
Transportation Research Record No. 1941, pp. 1-7, Washington, DC, January 2005.
and J.L.Hopson and J. Li, "Running Out of and Into Oil: Analyzing Global Depletion and Transition
Through 2050, Transportation Research Record 1880, pp. 1-9, Transportation Research Board,
Washington, DC, 2005.
and P.D. Patterson, M. Singh and J. Li, "Feebates, Rebates and Gas-Guzzler Taxes: A Study of Incentives
for Increased Fuel Economy," Energy Policy, vol. 33, no. 6, pp. 721-827, 2005.
Sheffield, J., et al., "Energy Options for the Future," Journal of Fusion Energy, vol. 23, no. 2, pp. 63-109,
2004.
and J. Hopson, "An Analysis of Alternative Forms of Automotive Fuel Economy Standards for the United
States," Transportation Research Record No. 1842, pp. 20-28, Transportation Research Board,
Washington, DC, 2003.
H.L. Hwang, S.M. Chin and D.L. Greene, "In, Out, Within and Through: Geography of Truck Freight in
the Lower 48," Transportation Research Record, no. 1768, pp. 18-25, Transportation Research Board,
Washington, DC, 2001.
and S.E. Plotkin, "Energy Futures for the U.S. Transportation Sector," Energy Policy, vol. 29, no. 14, pp.
1255-1270,2001.
and N. Tishchishyna, "The Costs of Oil Dependence: A 2000 Update," Transportation Quarterly, vol. 55,
no. 3, pp. 11-32,2001.
H.L. Hwang, D.L. Greene, S.M. Chin, J. Hopson and A.A. Gibson, "Real-time Indicators of VKT and
Congestion: One Year of Experience," Transportation Research Record, no. 1719, pp. 209-214,
Transportation Research Board, Washington, DC, 2000.
and J.M. DeCicco, "Engineering-Economic Analyses of Automotive Fuel Economy Potential in the United
States," Annual Review of Energy and the Environment, vol. 25, pp. 477-536, 2000.
L.A. Greening, D.L. Greene and C. Difiglio, "Energy Efficiency and Consumption—The Rebound
Effect—A Survey," Energy Policy, vol. 28, pp. 389-401, 2000.
R.N. Schock, W. Fulkerson, M.L. Brown, R.L. San Martin, D.L. Greene and J. Edmonds, "How Much Is
Energy R&D Worth as Insurance?" Annual Review of Energy and the Environment, vol. 24, pp. 487-512,
Annual Review, Palo Alto, California, 1999.
David L. Greene 5 March 2013
-------
S.M. Chin, D.L. Greene, J. Hopson, H.L. Hwang and B. Thompson, "Towards Real-Time Indices of U.S.
Vehicle Travel and Traffic Congestion," Transportation Research Record, no. 1660, pp. 132-139, National
Academy Press, Washington, DC, 1999.
and J. Kahn and R. Gibson, "Fuel Economy Rebound Effect for U.S. Household Vehicles, " The Energy
Journal, vol. 20, no. 3, pp. 1-31, 1999.
"Survey Evidence on the Importance of Fuel Availability to Choice of Alternative Fuels and Vehicles,"
Energy Studies Review, vol. 8, no. 3, pp. 215-231, 1998.
"Why CAFE Worked," Energy Policy, vol. 26, no. 8, pp. 595-614, 1998.
and Donald W. Jones and Paul N. Leiby, "The Outlook for U.S. Oil Dependence," Energy Policy, vol. 26,
no. 1, pp. 55-69, 1998.
and Michael Wegener, "Sustainable Transport," in Journal of Transport Geography, vol. 5, no. 3, pp. 177-
190, 1997.
Steven E. Plotkin and David Greene, "Prospects for Improving the Fuel Economy of Light-Duty Vehicles,"
Energy Policy, vol. 25, no. 14-15, pp. 1179-1188, 1997.
"Economic Scarcity: Monopoly, Not Geology, Threatens Global Supply," Harvard International Review,
vol. XIX, no. 3, Summer, 1997.
"Environmental Impacts," Journal of Transport Geography, vol. 5, no. 1, pp. 28-29, 1997.
"Energy for Transportation," Journal of Transport Geography, vol. 5, no. 1, pp. 30-32, 1997.
and Y. Fan, "Transportation Energy Intensity Trends, 1972-1992," Transportation Research Record, no.
1475, pp. 10-19, Energy and Environment, Transportation Research Board, Washington, DC, 1995.
M.A. Deluchi, D.L. Greene and Quanlu Wang, "Motor Vehicle Fuel Economy: The Forgotten Hydrocarbon
Control Strategy?" Transportation Research A, vol. 28A, no. 3, pp. 223-244, 1994.
"Transportation and Energy," Transportation Quarterly, vol. 48, no. 1, pp. 91-101, Winter, 1994.
and K.G. Duleep, "Costs and Benefits of Automotive Fuel Economy Improvement," Transportation
Research, vol. 27A, no. 3, pp. 217-236, May, 1993.
"Transportation and Energy: The Global Environmental Challenge," Transportation Research, vol. 27A,
no. 3, pp. 163-166, May, 1993.
"Energy Efficiency Improvement Potential of Commercial Aircraft," Annual Review of Energy and
Environment, vol. 17, pp. 537-573, 1992.
"Vehicle Use and Fuel Economy: How Big is the Rebound Effect?" The Energy Journal, vol. 13, no. 1, pp.
117-143, April 1992.
"A Note on OPEC Market Power and Oil Prices," Energy Economics, vol. 13, no. 2, pp. 123-129, April
1991.
"The Cost of Short-Run Pricing Strategies to Increase Corporate Average Fuel Economy," Economic
Inquiry,vol. XXIX, no. 1, pp. 101-114, January 1991.
"Fuel Choice forMultifuel Vehicles," Contemporary Policy Issues, vol. VIII, no. 4, pp. 118-137, October
1990.
David L. Greene 6 March 2013
-------
"CAFE or PRICE? An Analysis of the Effects of Federal Fuel Economy Regulations and Gasoline Price
on New CarMPG, 1978-89," The Energy Journal, vol. 11, no. 3, pp. 37-57, September 1990.
"Technology and Fuel Efficiency," Forum for Applied Research and Public Policy, vol. 5, no. 1, pp. 23-
29, University of Tennessee, Spring 1990.
Carmen Difiglio, K.G. Duleep and D.L. Greene, "Cost Effectiveness of Future Fuel Economy
Improvements," The Energy Journal, vol. 11, no. 1, 1990.
"Short-Term Options for Controlling COa Emissions of Light-Duty Vehicles," SAE Technical Paper Series
901111, Society of Automotive Engineers, 1990.
"Motor Fuel Choice: An Econometric Analysis," Transportation Research A, vol. 23 A, no. 3, pp. 243-253,
1989.
"Fuel Choice for Dual-Fuel Vehicles: An Analysis of the Canadian Natural Gas Vehicles Survey," SAE
Technical Paper Series 892067, Society of Automotive Engineers, Warrendale, Pennsylvania, 1989.
J. J. Erickson, D.L. Greene and A. J. Sabadell, "An Analysis of Transportation Energy Conservation Projects
in Developing Countries," Transportation, vol. 15, no. 3, pp. 163-189, 1988.
and J.T. Liu, "Automotive Fuel Economy Improvements and Consumers' Surplus," Transportation
Research A, vol. 22A, no. 3, pp. 203-218, 1988.
"Advances in Automobile Technology and the Market for Fuel Efficiency, 1978-1985," Transportation
Research Record 1155, pp. 18-27, Transportation Research Board, National Research Council,
Washington, DC, 1987.
and Anthony Araya Jacome, Robert Kowalski and Patricia S. Hu, "Road Transport Energy Conservation in
Costa Rica," Energy, vol. 12, no. 12, pp. 1299-1308, 1987.
and P.S. Hu, "A Functional Form Analysis of the Short-Run Demand for Travel and Gasoline by One-
Vehicle Households," Transportation Research Record, no. 1092, pp. 10-15, Transportation Research
Board, Washington, DC, 1986.
and N. Meddeb and J.T. Liu, "Vehicle Stock Modeling of Highway Energy Use: Tunisian and U.S.
Applications," Energy Policy, pp. 437-446, October 1986.
"Efficiency-Related Changes in Automobile and Light Truck Markets," SAE Technical Paper Series, no.
861423, Society of Automotive Engineers, Warrendale, Pennsylvania, September 1986.
"The Market Share of Diesel Cars in the U.S., 1979-83," Energy Economics, vol. 8, no. 1, pp. 13-21,
January 1986.
and P.S. Hu and L. Till, "An Analysis of Trends in Automotive Fuel Economy from 1978 to 1984,"
Transportation Research Record, no. 1049, pp. 51-56, Washington, DC, 1985.
"Estimating Daily Vehicle Usage Distributions and the Implications for Limited-Range Vehicles,"
Transportation Research B, vol. 19B, no. 4, pp. 347-358, 1985.
and P.S. Hu, "Vehicle Usage in Multi-Vehicle Households and the Price of Gasoline," Transportation
Research Record, no. 988, pp. 19-24, Washington, DC, 1984.
and P.S. Hu and G.F. Roberts, "An Analysis of Geographical and Temporal Variation in Vehicles
Classification Count Statistics," Transportation Research Record, no. 987, pp. 21-28, Washington, DC,
1984.
David L. Greene 7 March 2013
-------
and G.F. Roberts, "A Comment on Fuel Consumption for Road Transport in the U.S.A," Energy
Economics, vol. 6, no. 2, pp. 145-147, April 1984.
"A Derived Demand Model of Regional Highway Diesel Fuel Use," Transportation Research B, vol. 18B,
no.l, pp. 43-61, 1984.
P.O. Patterson, F.W. Westbrook, D.L. Greene and G.F. Roberts, "Reasons for Changes in MPG Estimates,
Model Year 1978 to the Present," SAE Technical Paper Series, no. 840500, Society of Automotive
Engineers, Warrendale, Pennsylvania, February/March, selected for inclusion in 1984 SAE Transactions,
1984.
J. Soderstrom, E. Hirst, D. Greene and J. Trimble, "Have Department of Energy Conservation Programs
Saved Energy?" Evaluation Review, vol. 8, no. 1, pp. 93-112, February 1984.
E. Hirst, R. Marlay, D. Greene and R. Barnes, "Recent Changes in U.S. Energy Consumption: What
Happened and Why?" Annual Review of Energy, vol. 8, pp. 193-243, Annual Reviews, Inc., Palo Alto,
California, 1983.
"Streamlining the Collection and Processing of Traffic Count Statistics, A Comment," Transportation
Research Record, no. 928, pp. 18-19, 1983.
"A Note on Implicit Consumer Discounting of Automobile Fuel Economy: Reviewing the Available
Evidence," Transportation Research, vol. 17B, no. 6, pp. 491-499, 1983.
and C.K. Chen, "A Time Series Analysis of State Gasoline Demand, 1975-1980," The Professional
Geographer, vol. 35, no. 1, pp 40-51, February 1983.
G.F. Roberts and D.L. Greene, "A Method for Assessing the Market Potential of New Energy-Saving
Technologies," Transactions on Systems, Man, and Cybernetics, Institute of Electrical and Electronics
Engineers, vol. SMC-13, no. 1, pp. 3-7, January/February 1983.
Eric Hirst et al., "Effects of Improved Energy Efficiency on U.S. Energy Use: 1973-1980," Energy, vol. 7,
no. 11, pp. 897-907, 1982.
and G. Kulp, "An Analysis of the 1979-1980 Decline in gasoline Consumption in the United States,"
Energy, vol. 7, no. 4, pp. 367-375, 1982.
and G. Walton, "Data and Methodological Problems in Establishing State Gasoline Conservation Targets,"
Transportation Research Record, no. 815, pp. 24-30, 1981.
and E. Chen, "Scrappage and Survival Rates of Passenger Cars and Trucks in the U.S., 1966-77,"
Transportation Research, vol. 15A, no. 5, pp. 383-389, 1981.
"City Size Distribution and Income Distribution in Space: A Comment," Regional Development Dialogue,
vol. II, no. 1, pp. 124-126, Spring 1981.
"A State Level Stock System Model of Gasoline Demand," Transportation Research Record, no. 801, pp.
44-50, 1981.
"Estimated Speed/Fuel Consumption Relationships for a Large Sample of Cars," Energy, vol. 6, pp. 441-
446, 1981.
"The Spatial Dimension of Gasoline Demand," Geographical Survey, vol. 9, no. 2, pp. 19-28, April 1980.
"Regional Demand for Gasoline: Comment," Journal of Regional Science, vol. 20, no. 1, pp. 103-109,
1980.
David L. Greene 8 March 2013
-------
"Urban Subcenters: Recent Trends in Urban Spatial Structure," Growth and Change, vol. 11, no. 1, pp.
103-109, January 1980.
R. Dubin, D.L. Greene and C. Begovich, "Multivariate Classification of Automobiles Using an
Automobile's Characteristics' Data Base," Transportation Research Record, no. 726, pp. 29-27, 1979.
"State Differences in the Demand for Gasoline: An Econometric Analysis," Energy Systems and Policy,
vol. 3, no. 2, pp. 191-212, 1978.
and Joern Barnbrock, "A Note on Problems in Estimating Urban Density Models," Journal of Urban
Economics, vol. 5, April 1978.
and Rolf R. Schmitt, "An Alternative Derivation of the Intervening Opportunities Model," Geographical
Analysis, vol. 10, no. 1, January 1978.
Joern Barnbrock and D.L. Greene, "A Comment on Population Density and Trend Surface Analysis," Land
Economics, vol. 53, no. 2, May 1977.
CONTRIBUTIONS TO NATIONAL RESEARCH COUNCIL REPORTS
Transitions to Alternative Vehicles and Fuels, Report of the Committee on Transitions to Alternative
Vehicles and Fuels, National Research Council, National Academies Press, Washington, D.C., 2013.
Assessment of Fuel Economy Technologies for Light-Duty Vehicles, Report of the Committee on the
assessment of Technologies for Improving Light-duty Vehicle Fuel Economy, National Academies Press,
Washington, D.C., 2010.
J. Zucchetto, Trends in Oil Supply and Demand, Potential for Peaking of Conventional Oil Production, and
Possible Mitigation Options, a summary report of the Modeling the Oil Transition workshop, Member,
Planning Group and Keynote Speaker, Washington, DC, April 2006.
State and Federal Standards for Mobile Source Emissions, Report of the Committee on State Practices in
Setting Mobile Source Emissions Standards, National Research Council, National Academies Press,
Washington, DC, March 2006.
"Integrating Sustainability into the Transportation Planning Process," Conference Proceedings 37,
Transportation Research Board of the National Academies, Washington, DC, 2005.
Effectiveness and Impact of Corporate Average Fuel Economy (CAFE) Standards, Report of the
Committee, National Research Council, National Academy Press, Washington, 2002.
"Ecological, Environmental and Energy-Related Issues, in The Future Highway Transportation System and
Society, Transportation Research Board, National Research Council, National Academy Press, Washington,
DC, 1997.
Expanding Metropolitan Highways: Implications for Air Quality and Energy Use, Special Report 245,
Transportation Research Board, National Research Council, Washington, DC, July 1995.
Automotive Fuel Economy: How Far Can We Go? Report of the Committee on Automobile and Light
Truck Fuel Economy, National Research Council, National Academy Press, Washington, DC, 1992.
In Pursuit of Speed: New Options for Intercity Passenger Transport, Special Report 233, Transportation
Research Board, National Research Council, Washington, DC, 1991.
and D. Sperling and B. McNutt, "Transportation Energy to the Year 2020," pp. 207-231, in^4 Look Ahead:
Year 2020, Special Report 220, Transportation Research Board, National Research Council, Washington,
DC, 1988.
David L. Greene 9 March 2013
-------
CONGRESSIONAL TESTIMONY
"Testimony to the United States Senate Committee on Energy and Natural Resources by David L. Greene,
July 24, 2012", hearing on Natural Gas and Transportation.
"Technology-Neutral Incentives for Energy Efficient Low Greenhouse Gas Emitting Vehicles," Testimony
to the Finance Committee, United States Senate, April 23, 2009.
"Near-Term Options to Increase Fuel Economy and Reduce Petroleum Demand," Testimony to the
Committee on Energy and Natural Resources, United States Senate, July 23, 2008.
"Facing the Challenges of Oil Dependence and Climate Change: What Will It Take?" Testimony to the
Subcommittee on Energy and Water Development, U.S. House of Representatives Committee on
Appropriations, February 14, 2008.
"Policies to Increase Passenger Car and Light Truck Fuel Economy," Testimony to the U.S. Senate
Committee on Energy and Natural Resources, January 30, 2007.
"Is Cap-and-Trade a Sufficient Carbon Policy for Transportation?" Testimony to the U.S. Senate
Committee on Environment and Public Works, November 13, 2007, Washington, DC.
"Energy Challenges for Transportation in the 21st Century," Testimony to the National Surface
Transportation Policy and Revenue Study Commission, March 19, 2007, Washington, DC.
"Corporate Average Fuel Economy (CAFE) Standards," Testimony to the U.S. Senate Commerce
Committee, March 6, 2007, Washington, DC.
"Policies to Increase Passenger Car and Light Truck Fuel Economy," Testimony to the U.S. Senate
Committee on Energy and Natural Resources, January 30, 2007, Washington, DC .
"Observations on the H-Prize Act of 2006 (H.R. 5143)," Testimony to the U.S. House of Representatives
Committee on Science, April 27, 2006, Washington, DC.
"Improving the Nation's Energy Security: Can Cars and Trucks be Made More Fuel Efficient?" Testimony
to the U.S. House of Representatives Committee on Science, February 9, 2005, Serial No. 109-3,
U.S.G.P.O, Washington, DC.
"The Outlook for Surface Transportation Growth," Testimony to the Subcommittee on Surface
Transportation of the Committee on Transportation Infrastructure of the United States House of
Representatives, March 28, 1996.
CONTRIBUTIONS TO BOOKS
and C. Liu and S. Park, "Transition from Petro-Mobility to Electro-Mobility", in Transition to Renewable
Energy Systems, D. Stoltenand V. Scherer, eds., WILEY-VCH, Weinheim, Germany, 2013.
"The Energy Efficiency Potential of Global Transport to 2050", in A. Zichichi and R. Ragaini, eds.,
International Seminar on Nuclear War and Planetary Emergencies, 44th Session, World Scientific,
London, 2011.
"Measuring Energy Sustainability: Making Progress," in Linkages of Sustainability, T.E. Graedel and E.
van der Voet, eds., Strungmann Forum Reports, MIT Press, Cambridge, MA, 2010.
Loschel, A., J. Johnston, M.A. Delucchi, T.N. Demayo, D.L. Gauthier, D.L. Greene, J. Ogden, S. Rayner
and E. Worrell, "Energy: Stocks, Flows, and Prospects," in Linkages of Sustainability, T.E. Graedel and E.
van der Voet, eds., Strungmann Forum Reports, MIT Press, Cambridge, MA, 2010.
David L. Greene 10 March 2013
-------
D.L. Greene, "What Will Replace Liquid Hydrocarbon Fuels?" in M.J. Benton, ed., The Seventy Great
Mysteries of the Natural World, Thames and Hudson, Inc., New York, 2008.
and J. German and M. A. Delucchi, "Fuel Economy: The Case for Market Failure," in Reducing Climate
Impacts of the Transportation Sector, D. Sperling and J.S. Cannon, eds., Springer Science+Business Media,
springer.com, 2008.
Ribeiro, S.K., S. Kobayashi, D.L. Greene et al., "Transport and its Infrastructure," Chapter 5 in Climate
Change 2007: Mitigation of Climate Change, Working Group III contribution to the Fourth Assessment
Report of the Intergovernmental Panel on Climate Change, Cambridge Press, Cambridge, United Kingdom.
A. Dumas, D.L. Greene and A. Bourbeau, "North American Feebate Analysis Model," in Driving Climate
Change, D. Sperling and J.S. Cannon, eds., Academic Press, San Francisco, CA, 2007.
"Transportation and Energy," in The Geography of Urban Transportation, S. Hanson and G. Giuliano, eds.,
The Guilford Press, New York, 2004.
"Transportation and Energy Overview," in Encyclopedia of Energy, Cutler J. Cleveland, ed., Elsevier,
Oxford, United Kingdom, 2004.
M. Wegener and D.L. Greene, "Sustainable Transport" in Social Change and Sustainable Transport, pp.
35-42, W.R. Black and P. Nijkamp, eds., Indiana University Press, Bloomington, IN, 2002.
"Sustainable Transportation," in International Encyclopedia of the Social and Behavioral Sciences, N.J.
Smelser and P.B. Baltes, eds., Elsevier, Oxford, pp. 15,335-15,339, 2002.
W.R. Moomaw, J.R. Moreira, K. Blok, D.L. Greene et al., "Technological and Economic Potential of
Greenhouse Gas Emissions Reduction," Chapter 3 in Climate Change 2001: Mitigation, contribution of
Working Group III to the Third Assessment Report of the Intergovernmental Panel on Climate Change,
Cambridge University Press, Cambridge, UK, 2001.
"Transportation's Oil Dependence and Energy Security in the 21st Century," in Environmental Change,
Adaptation, and Security, S.C. Lonergan, ed., NATO Advanced Science Institute Series, Klumer Academic
Publishers, Boston, 1999.
S.L. Baughcum, J.J. Begin, F. Franco, D.L. Greene et al., "Aircraft Emissions: Current Inventories and
Future Scenarios," Chapter 9 in Aviation and Global Atmosphere, Intergovernmental Panel on Climate
Change, Cambridge University Press, Cambridge, UK, 1999.
and Robert Gibson, "Transportation, Energy and the Environment," Chapter 4 in Transportation Statistics
Annual Report 1998, BTS98-S-01, Bureau of Transportation Statistics, U.S. Department of Transportation,
Washington, DC, September 1998.
"Commercial Air Transport Energy Use and Emissions: Is Technology Enough?" pp. 207-228 in DeCicco
and Delucchi, eds., Transportation, Energy, and the Environment: How Far Can Technology Take Us?
American Council for an Energy Efficient Economy, Washington, DC, 1997.
B. McNutt, L. Fulton and D.L. Greene, "Is Technology Enough? A Synthesis of Views Expressed at the
Conference," pp. 251-256 in DeCicco and Delucchi, eds., Transportation, Energy, and Environment: How
Far Can Technology Take Us? American Council for an Energy Efficient Economy, Washington, DC,
1997.
and Robert Gibson, "Transportation, Energy, and the Environment," Chapter 4 in Transportation Statistics
Annual Report 1997, BTS97-5-01, Bureau of Transportation Statistics, U.S. Department of Transportation,
Washington, DC, October 1997.
and Donald W. Jones, "The Full Costs and Benefits of Transportation: Conceptual and Theoretical Issues,"
Chapter 1 of The Full Costs and Benefits of Transportation, Springer-Verlag, Heidelberg, 1997.
David L. Greene 11 March 2013
-------
"The Cost of Transportation's Oil Dependence," in O. Hohmeyer, R.L. Ottinger and K. Rennings, eds.,
Social Costs andSustainability, Springer-Verlag, Berlin, 1996.
"Twenty Years of Energy Policy: What Should We Have Learned?" in David L. Feldman, ed., The Energy
Crisis, The Johns Hopkins University Press, Baltimore, 1996.
"Environmental Impacts of Transportation," Chapter 6 in Transportation Statistics Annual Report 1996,
Bureau of Transportation Statistics, U.S. Department of Transportation, Washington, DC, October 1996.
R. Gibson and D.L. Greene, "Environmental Trends and the U.S. Transportation System," Chapter 7 in
Transportation Statistics Annual Report 1996, Bureau of Transportation Statistics, U.S. Department of
Transportation, Washington, DC, October 1996.
"Energy and Transportation," Chapter 4 in Transportation Statistics Annual Report 1996, Bureau of
Transportation Statistics, U.S. Department of Transportation, Washington, DC, October 1996.
"Transportation," Chapter 5 in Policies and Measures for Reducing Energy Related Greenhouse Emissions,
DOE/PO-0047, U.S. Department of Energy, Office of Policy and International Affairs, Washington, DC,
July 1996.
Guest editor, "Energy and Global Climate Change," Transportation Research, Special Issue, vol. 27A,
no. 3, May 1993.
"Regional Demand Implication for Gasoline Supply Shortages," in T.R. Lakshmanan and P. Nijkamp, eds.,
Systems and Models for Energy and Environmental Analysis, pp. 206-233, Gower, UK, 1983.
G. Samuels, A.B. Rose, D.L. Greene and J.N. Hooker, "Energy Conservation in Transportation," a chapter
in Vol. Ill of Advances in Energy Systems and Technology, Peter Auer, ed., Academic Press, New York,
1982.
"A Regional Stock System Model of Highway Gasoline Demand," in Changing Energy Use Futures,
Volume 1, Proceedings of the Second International Conference on Energy Use and Management, Los
Angeles, California, October 22-26, 1979.
OTHER PUBLICATIONS
"Energy Policy: Where are the boundaries?", Editorial, Energy Policy, November, 2013.
C.Z. Liu and D.L. Greene, Modeling the Demand for ESS in the United States, ORNL/TM-2013/369, Oak
Ridge National Laboratory, Oak Ridge, Tennessee, October, 2013, available at:
http://info.ornl.gov/sites/publications/Files/Pub45616.pdf
and G. Duleep, Status and Prospects of the Global Automotive Fuel Cell Industry and Plans for
Deployment of Fuel Cell Vehicles and Hydrogen Refueling Infrastructure, ORNL/TM-2013/222, Oak
Ridge National Laboratory, Oak Ridge, Tennessee, June.
and S. Park and C. Liu, Analyzing the Transition to Electric Drive in California, White Paper 4-13, Howard
H. Baker, Jr. Center for Public Policy, The University of Tennessee, Knoxville, TN, June, 2013.
and R.S. Lee and J.L. Hopson, OPEC and the Costs to the U.S. Economy of Oil Dependence, White Paper
1-13, Howard H. Baker, Jr. Center for Public Policy, The University of Tennessee, Knoxville, TN,
February, 2013.
"Rebound Effects in Transportation", a contribution to The Need to Account for Consumer Behavior in
order to Develop Robust Energy Efficiency Policies, I.L. Axevedo, M. Sonnberger, B. Thomas, G. Morgan
and O. Renn, International Risk Governance Council, Geneva, Switzerland, November 13, 2012, available
at: http://cedm.epp.cmu.edu/ReboundMeeting201 l/David%20Greene.pdf
David L. Greene 12 March 2013
-------
and C.Z. Liu, "Consumer Vehicle Choice Model Documentation", EPA-420-B-12-052, Office of
Transportation and Air Quality, U.S. Environmental Protection Agency, available at
http://www.epa.gov/otaq/climate/documents/420bl2052.pdf. 2012.
and K.G. Duleep, and G. Upreti, "Status and Outlook for the U.S. Non-Automotive Fuel Cell Industry:
Impacts of Government Policies and Assessment of Future Opportunities", ORNL/TM-2011/101, Oak
Ridge National Laboratory, Oak Ridge, Tennessee.
Bunch, D.S. and D.L. Greene, "Potential Design, Implementation, and Benefits of a Feebate Program for
New Passenger Vehicles in California," State of California Air Resources Board and the California
Environmental Protection Agency, Sacramento, California, 2011, available at
http://76.12.4.249/artman2/uploads/l/Feebate_Program_for_New_Passenger_Vehicles_in_California.pdf.
"Oil Peak or Panic?" Book review, Science, vol. 328, no. 5980, p. 828, May 14, 2010
"What's Greener than a VMT Tax? Discussion Paper, Howard H. Baker, Jr. Center for Public Policy, The
University of Tennessee, Knoxville, TN, May 2010.
and S.E. Plotkin, Reducing Greenhouse Gas Emissions from U.S. Transportation, Pew Center on Global
Climate Change, Arlington, VA, 2011.
How Consumers Value Fuel Economy: A Literature Review, EPA-420-R-10-008, U.S. Environmental
Protection Agency, March 2010.
"Why the Market for New Passenger Cars Generally Undervalues Fuel Economy," Discussion Paper No.
2010-6, International Transport Forum, OECD, Paris, January 2010.
D. McCollum, G. Gould and D. Greene, Greenhouse Gas Emissions from Aviation and Marine
Transportation: Mitigation Potential and Policies, Pew Center on Global Climate Change, Arlington,
Virginia, available at http://www.pewclimate.org/technologv/report/aviation-an-marine. December 2009.
"Energy Assurance: Essential Energy Technologies for Climate Protection and Energy Security,"
ORNL/TM-2009/314, Oak Ridge National Laboratory, Oak Ridge, TN, December 2009.
S. Ploikmetal.,Multipath Transportation Futures Study: Vehicle Characterization and Scenario Analyses,
ANL/ESD/09-5, Argonne National Laboratory, Argonne, Illinois, July 2009.
and K.G. Duleep, "Bootstrapping a Sustainable North American PEM Fuel Cell Industry: Could a Federal
Acquisition Program Make a Difference?" ORNL/TM-2008/183, October, 2008.
"Future Prices and Availability of Transport Fuels," International Transport Forum, Research Round Table
on Oil Dependence: Is Transport Running Out of Affordable Fuel? November 14-15, 2007, OECD, Paris.
American Physical Society Study Group, Energy Future: Think Efficiency, American Physical Society,
College Park, Maryland, 2008.
D.L. Greene et al., Analysis of the Transition to Hydrogen Fuel Cell Vehicles and the Potential Hydrogen
Infrastructure Requirements, ORNL/TM-2008/30, Oak Ridge National Laboratory, Oak Ridge, Tennessee,
March, 2008.
"Transportation," in The First State of the Carbon Cycle Report (SOCCR): The North American Budget
and Implications for the Global Carbon Cycle. A Report by the U.S. Climate Change Science Program and
the Subcommittee on Global Change Research [King, A.W., L. Dilling, G.R. Zimmerman, D.M. Ciarman,
R.A. Houghton, G. Marland, A.Z. Rose, and T.J. Wilbanks (eds)]. National Oceanic and Atmospheric
Administration, National Climate Data Center, Asheville, North Carolina, USA, pp. 73-84.
David L. Greene 13 March 2013
-------
and P.N. Leiby and D. Bowman, Integrated Analysis of Market Transformation Scenarios with HyTrans,
ORNL/TM-2007/094, Oak Ridge National Laboratory, Oak Ridge, Tennessee, June 2007.
D. Gordon, D.L. Greene, M.H. Ross and T.P. Wenzel, "Sipping Fuel and Saving Lives: Increasing Fuel
Economy without Sacrificing Safety," International Council on Clean Transportation, Washington, DC,
June 2007.
and P.N. Leiby, "Oil Independence: Realistic Goal or Empty Slogan," published as "Expert Commentary"
on The Lugar Energy Initiative website,
http://lugar.senate.gov/energy/links/commentary/08_greene_summary.html. 2007.
Editor, Modeling the Oil Transition: A Summary of the Proceedings of the DOE/EPA Workshop on the
Economic and Environmental Implications of Global Energy Transitions, ORNL/TM-2007/014, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, February 2007.
and P.N. Leiby, The Oil Security Metrics Model, ORNL/TM-2006/505, Oak Ridge National Laboratory,
Oak Ridge, Tennessee, May 2006.
and S. Ahmad, Costs of U.S. Oil Dependence: 2005 Update, ORNL/TM-2005/45, Oak Ridge National
Laboratory, Oak Ridge, Tennessee, March 2005.
"Climate Change Policy for Transportation While Waiting for H2," in The 10-50 Solution: Technologies
and Policies for a Low-Carbon Future, Pew Center on Global Climate Change, National Commission on
Energy Policy, Washington, DC, March 2004.
and K.G. Duleep and W. McManus, Future Potential of Hybrid and Diesel Powertrains in the U.S. Light-
Duty Vehicle Market, ORNL/TM-2004/181, Oak Ridge National Laboratory, Oak Ridge, Tennessee,
August 2004.
and J.L. Hopson, Running Out of and Into Oil: Analyzing Global Depletion and Transition Through 2050,
ORNL/TM-2003/259, Oak Ridge National Laboratory, Oak Ridge, Tennessee, October 2003.
and R.C. Gibson and K.G. Duleep, Energy Star Concepts for Highway Vehicles, ORNL/TM-2003/37, Oak
Ridge National Laboratory, Oak Ridge, Tennessee, June 2003.
R. Nye, D.L. Greene and J.W. Saulsbury, Providing Consumers with Web-based Information on the
Environmental Effects of Automobiles, ORNL/TM-2003/166, Oak Ridge National Laboratory, Oak Ridge,
Tennessee, June 2003.
and A. Schafer, "Reducing Greenhouse Gas Emissions from U.S. Transportation," Pew Center on Global
Climate Change, Arlington, Virginia, May 2003.
and S. Plotkin and K.G. Duleep, Examining the Potential for Voluntary Fuel Economy Standards in the
United States and Canada, ANL/ESD/02-5, Center for Transportation Research, Argonne National
Laboratory, Argonne, Illinois, October 2002.
S.M. Chin, O. Franzese, D.L. Greene, H.L. Hwang and R.C. Gibson, Temporary Losses of Highway
Capacity and Impacts on Performance, ORNL/TM-2002/3, Oak Ridge National Laboratory, Oak Ridge,
Tennessee, May 2002.
TAFV Alternative Fuels and Vehicles Choice Model Documentation, ORNL/TM-2001/134, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, July 2001.
R.N. Schock, W. Fulkerson, M.L. Brown, R.L. San Martin, D.L. Greene and J.A. Edmonds, The Insurance
Value of Energy R&D, UCEL-ID-141815, CGSR-2001, 003, Center for Global Security Research,
Lawrence Livermore National Laboratory, Livermore, California, May 2001.
David L. Greene 14 March 2013
-------
"Energy," in Chapter 5 of The Changing Face of Transportation, U.S. Department of Transportation,
Bureau of Transportation Statistics, BTSOO-007, Washington, DC, 2000.
with S. Plotkin, "Transportation Sector," Chapter 6 in Scenarios for a Clean Energy Future, Interlaboratory
Working Group on Energy-Efficient and Clean Energy Technologies, Oak Ridge National Laboratory,
ORNL/CON-476, Lawrence Berkeley National Laboratory, LBNL-44029, November 2000.
R.N. Schock, W. Fulkerson, M.L. Brown, R.L. San Martin, D.L. Greene and J. Edmonds, "How Much Is
Energy R&D Worth?" Energy, vol. 25, no. 2, pp. 4-8, 2000.
Costs of Oil Dependence: A 2000 Update, ORNL/TM-2000/152, Oak Ridge National Laboratory, Oak
Ridge, Tennessee, May 2000.
and J.M. DeCicco, Engineering-Economic Analyses of Automotive Fuel Economy Potential in the United
States, ORNL/TM-2000/26, Oak Ridge National Laboratory, Oak Ridge, Tennessee, February 2000.
and J.M. DeCicco, "Energy and Transportation Beyond 2000," Millennium Paper Series, Transportation
Research Board, National Research Council, Washington, DC, January 2000.
and many others, National Transportation Statistics 1999, BTS99-04, U.S. Department of Transportation,
Bureau of Transportation Statistics, Washington, DC, 1999.
and many others, Transportation Statistics Annual Report 1999, BTS99-03, U.S. Department of
Transportation, Bureau of Transportation Statistics, Washington, DC, 1999.
An Assessment of Energy and Environmental Issues Related to the Use of Gas-to-Liquid Fuels in
Transportation, ORNL/TM-1999/258, Oak Ridge National Laboratory, Oak Ridge, Tennessee, November
1999.
and J. Kahn and R. Gibson, An Econometric Analysis of the Elasticity of Vehicle Travel with Respect to
Fuel Cost per Mile Using RTEC Survey Data, ORNL-6950, Oak Ridge National Laboratory, Oak Ridge,
Tennessee, March 1999.
"Why CAFE Worked," Transportation Research Circular No. 492, Transportation Research Board,
National Research Council, Washington, DC, 1999.
R.N. Schock, W. Fulkerson, M.L. Brown, R.L. San Martin, D.L. Greene and J. Edmonds, Haw Much is
Energy R&D Worth as Insurance? UCRL-JC-131205, Lawrence Livermore National Laboratory,
Livermore, California, October 1998.
"Transportation and Energy Policy," in Twenty-Five Years of Energy and Environmental Policy, pp. 125-
140, University of Illinois at Chicago, Energy Resources Center, Chicago, Illinois, 1998.
"Ecological, Environmental and Energy-Related Issues," pp. 127-149 in The Future Highway
Transportation System and Society, Transportation Research Board, National Research Council, 1997.
and S. Plotkin, "Transportation Sector," Chapter 5 in Scenarios of U.S. Carbon Reductions, Interlaboratory
Working Group on Energy-Efficient and Low-Carbon Technologies, ORNL/CON-444, report to the Office
of Energy Efficiency and Renewable Energy, U.S. Department of Energy, Washington, DC, September
1997.
and many others, Energy Technology R&D: What Could Make a Difference, ORNL-6921/V2-P1, Oak
Ridge National Laboratory, Oak Ridge, Tennessee, September 1997.
"Energy and Environmental Consequences of Transportation: Indicators of Sustainability," 51st Session of
the International Statistical Institute Proceedings, Istanbul, Turkey, August 18-26, 1997.
David L. Greene 15 March 2013
-------
"Oil Dependence: The Value of R&D," 32nd Intersociety Energy Conversion Engineering Conference
Proceedings, volume 3, pp. 2148-2153, Honolulu, Hawaii, July 28-30, 1997.
and Steve Plotkin and K.G. Duleep, "The Potential for Energy-Efficient Technologies to Reduce Carbon
Emissions in the United States: Transport Sector," 32nd Intersociety Energy Conversion Engineering
Conference Proceedings, volume 3, pp. 2114-2119, Honolulu, Hawaii, July 28-30, 1997.
"Grin and Obey," Energy, vol. XXII, p. 6, April 1997.
"Consequences of U.S. Oil Dependence," Energy, vol. XXI, no. 5, pp. 3-8, November 1996.
D. Streets and many others, "Inventory of Technologies, Methods, and Practices for Reducing Emissions of
Greenhouse Gases," technical appendix to Climate Change 1995: Impacts, Adaptations, and Mitigation of
Climate Change: Scientific-Technical Analyses, Argonne National Laboratory, Argonne, Illinois, May
1996.
one of Principal Lead Authors, "Industry, Energy and Transportation: Impacts and Adaptation," in Climate
Change 1995: Impacts, Adaptations and Mitigation of Climate Change: Scientific-Technical Analyses, R.T.
Watson, M.C. Zinyowera, R.H. Moss and D.J. Dokken, eds., Intergovernmental Panel on Climate Change,
Cambridge University Press, 1996.
one of Contributing Authors, "Mitigation Options for the Transportation Sector," in Climate Change 1995:
Impacts, Adaptations and Mitigation of Climate Change: Scientific-Technical Analyses, R.T. Watson, M.C.
Zinyowera, R.H. Moss and D.J. Dokken, eds., Intergovernmental Panel on Climate Change, Cambridge
University Press, 1996.
"Commercial Air Transport Energy Use and Emissions: Is Technology Enough?" in Sustainable
Transportation: Is Technology Enough? Proceedings of the 1995 Asilomar Conference on Transportation
and Energy, American Council for an Energy-Efficient Economy, Washington, DC, 1995.
"Transportation, Energy and Environment: Global Policy Issues for the Early 21st Century," in
International Workshop on Motor Vehicles and Global Environmental Problems, Japan Automotive
Research Institute, Tokyo, Japan, October-November 1995.
P.N. Leiby, D.L. Greene and Harry Vidas, Market Potential and Impacts of Alternative Fuel use in Light
Duty Vehicles: A 2000/2010 Analysis, DOE/PO-0042, Office of Policy Analysis, U.S. Department of
Energy, Washington, DC, January 1996.
and X. Han, "The Unintended Consequences of Transportation," Chapter 3 in Transportation Statistics
Annual Report 1995, pp. 43-86, Bureau of Transportation Statistics, U.S. Department of Transportation,
Washington, DC, 1995.
with D.W. Jones and P.N. Leiby, The Outlook of U.S. Oil Dependence, ORNL-6783, Oak Ridge National
Laboratory, Oak Ridge, Tennessee, May 1995.
with S.M. Chin and R. Gibson, Aggregate Vehicle Travel Forecasting Model, ORNL-6872, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, May 1995.
D. Belzer, J. Roop, R. Sands and D.L. Greene, Energy Conservation Trends, Energy Efficiency in the U.S.
Economy, Report Three, DOE/PO-0034, U.S. Department of Energy, Office of Energy Efficiency and
Alternative Fuels Policy, Washington, DC, April 1995.
Transportation Energy Efficiency Trends, 1972-1992, ORNL-6828, Oak Ridge National Laboratory, Oak
Ridge, Tennessee, December 1994.
Alternative Fuels and Vehicles Choice Model, ORNL/TM-12738, Oak Ridge National Laboratory, Oak
Ridge, Tennessee, October 1994.
David L. Greene 16 March 2013
-------
and E. Hillsman, L. Fulton, A.K. Wolfe and J. Niles, Energy Emissions and Social Consequences of
Telecommuting, DOE/PO-0026, Technical Report One, Energy Efficiency in the U.S. Economy Series,
U.S. Department of Energy, Office of Policy, Planning and Program Evaluation, Washington, DC, June
1994.
R.R. Schmitt, D.L. Greene et al., Transportation Statistics Annual Report, Bureau of Transportation
Statistics, U.S. Department of Transportation, Washington, DC, 1994.
and Paul N. Leiby, The Social Costs to the U.S. of Monopolization of the World Oil Market, 1972-1991,
ORNL-6744, Oak Ridge National Laboratory, Oak Ridge, Tennessee, March 1993.
Impacts on Home Heating Costs of Incentives for Alternative Fuel Vehicles, Technical Report Eight, U.S.
Department of Energy, Office of Domestic and International Energy Policy, Washington, DC, May 1992.
with K.G. Duleep, Costs and Benefits of Automotive Fuel Economy Improvement: A Partial Analysis,
ORNL-6704, Oak Ridge National Laboratory, Oak Ridge, Tennessee, March 1992.
M.A. Deluchi, Q. Wang and D.L. Greene, Motor Vehicle Fuel Economy: The Forgotten Hydrocarbon
Control Strategy? ORNL-6715, Oak Ridge National Laboratory, Oak Ridge, Tennessee, July 1992.
"Transportation Energy Policy: Back to the Past or Ahead to the Future?" in Twenty Years of Energy
Policy: Looking Toward the Twenty-First Century, Proceedings of the Twentieth Annual Illinois Energy
Conference, Energy Resources Center, University of Illinois at Chicago, Chicago, Illinois, November 23-
24, 1992.
with M. Singh, E. Ecklund, R. Bechtold and C. Saricks, "Second Interim Report of the Interagency
Commission on Alternative Motor Fuels," Office of Energy Demand, Office of Policy, Planning and
Analysis, Washington, DC, 1992.
Q. Wang, D.L. Greene and M.A. Deluchi, "Effects of Increasing Fuel Economy on Gasoline Vehicle HC
Emissions," Proceedings of the 84th Annual Meeting and Exhibition, Air and Waste Management
Association, Vancouver, BC, Canada, June 16-21, 1991.
"Coverage and Quality Problems with Existing Data Resources for Freight Transportation," Proceedings of
the Special Conference on Freight Transportation Data: The Changing Federal Role Since Deregulation,
Transportation Research Circular, No. 367, Transportation Research Board, Washington, DC, 1990.
and M. Singh, First Interim Report on the Interagency Commission on Alternative Motor Fuels, Office of
Energy Demand Policy, Office of Policy, Planning and Analysis, U.S. Department of Energy, Washington,
DC, September 30, 1990.
"Commercial Aircraft Fuel Efficiency Potential Through 2010," Proceedings 1990 Intersociety Energy
Conversion Engineering Conference, Reno, Nevada, published by the American Institute of Chemical
Engineers, August 12-17, 1990.
Energy Efficiency Improvement Potential of Commercial Aircraft to 2010, ORNL-6622, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, June 1990.
and Anju Rathi, Alternative Motor Fuel Use Model: Model Theory and Design, and User's Guide,
ORNL/TM-11448, Oak Ridge National Laboratory, Oak Ridge, Tennessee, April 1990.
with Donald Igo, "Energy Dependence," Chapter 3 in National Strategic Transportation Study, U.S.
Department of Transportation, Washington, DC, 1989.
"A Context for Estimating Economic and Energy Security Benefits," Technical Report Two, Assessment of
Costs and Benefits of Flexible and Alternative Fuel Use in the U.S. Transportation Sector, DOE/PE-0091,
Office of Policy, Planning and Analysis, U.S. Department of Energy, Washington, DC, September 1989.
David L. Greene 17 March 2013
-------
and J.C. Liang, Modeling Operating Weight and Axle Weight Distributions for Highway Vehicles, ORNL-
6437, Oak Ridge National Laboratory, Oak Ridge, Tennessee, July 1988.
et al., "Research Priorities in Transportation and Energy," Transportation Research Circular, no. 323,
Transportation Research Board, Washington, DC, September 1987.
Simulating the Market for Automotive Fuel Efficiency: The SHRSIM Model, ORNL/TM-10074, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, February 1987.
and M.C. Holcomb, Off-Highway Use of Gasoline in the United States, U.S. Department of Transportation,
Federal Highway Administration, Office of Highway Information Management, Washington, DC, July
1986.
RUMS, A PC-Based FORTRAN Program for Estimating Consumer Surplus Changes Using Multinomial
Logit and Hedonic Demand Models, ORNL/TM-10069, Oak Ridge National Laboratory, Oak Ridge,
Tennessee, August 1986.
Driver Energy Conservation Awareness Training: Review and Recommendations, ORNL/TM-9897, Oak
Ridge National Laboratory, Oak Ridge, Tennessee, May 1986.
R.L. Graves, D.L. Greene, E.W. Gregory, II, Application of the Adiabatic Diesel to Heavy Trucks: A
Technology Assessment, ORNL/TM-9554, Oak Ridge National Laboratory, Oak Ridge, Tennessee, March
1986.
and R. Kowalski and F. Southworth, The Transportation Sector in Costa Rica and Opportunities for
Energy Conservation, a report of the Energy Conservation Services Program, U.S. Agency for International
Development, Office of Energy, Washington, DC, May 1985.
P.S. Hu, D.L. Greene and L.E. Till, Motor Vehicle MPG and Market Shares Report: First Six Months of
Model Year 1984, ORNL/TM-9391, Oak Ridge National Laboratory, Oak Ridge, Tennessee, October 1984.
and P.S. Hu and A.B. Rose, Transportation Energy Use and Efficiency in Tunisia, ORNL-6066, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, August 1984.
"Highway Fuel Use: Trends and Factors," Proceedings of the Energy Information Administration
Symposium on Petroleum Information, DOE/EIA-0425, Energy Information Administration, Washington,
DC, September 1983.
G.F. Roberts and D.L. Greene, Trends in Heavy Truck Energy Use and Efficiency, ORNL/TM-8843, Oak
Ridge National Laboratory, Oak Ridge, Tennessee, October 1983.
et al., An Overview of the Wharton School Assessments of the Short-Term Integrated Forecasting System,
DOE/NBB-0046, Energy Information Administration, Office of Statistical Standards, Washington, DC,
July 1983.
"Regional Demand Implications for Gasoline Supply Shortages," in T.R. Lakshmanan and P. Nijkamp,
eds., Systems and Models for Energy and Environmental Analysis, Gower, UK, pp. 206-233, 1983.
and E. Hirst, J. Soderstrom and J. Trimble, Estimating the Total Impact on Energy Consumption of
Department of Energy Conservation Programs, ORNL-5925, Oak Ridge National Laboratory, Oak Ridge,
Tennessee, November 1982.
"Gasoline use in the United States," Petroleum Supply Monthly, DOE/EIA-0109 (82-05), U.S. Department
of Energy, Washington, DC, 1982.
and G. Kulp, G.H. Walton and D.B. Shonka, Transportation Energy Use 1973-80: Changes, Trends, and
Causes, ORNL/TM-7953, Oak Ridge National Laboratory, Oak Ridge, Tennessee, April 1982.
David L. Greene 18 March 2013
-------
G. Samuels, A.B. Rose, D.L. Greene and J.N. Hooker, "Energy Conservation in Transportation," a chapter
in Vol. Ill of Advances in Energy Systems and Technology, Peter Auer, ed., Academic Press, New York,
1982.
E. Hirst, D.L. Greene et al., Energy Use from 1973 to 1980: The Role of Improved Energy Efficiency,
ORNL/CON-79, Oak Ridge National Laboratory, Oak Ridge, Tennessee, December 1981.
The Aggregate Demand for Gasoline and Highway Passenger Vehicles in the United States: A Review of
the Literature, 1938-1978, ORNL-5728, Oak Ridge National Laboratory, Oak Ridge, Tennessee, July
1981.
J.N. Hooker, A.B. Rose and D.L. Greene, End Use Energy Consumption Data Base: Transportation
Sector, DOE/EIA/CR-7405-01, U.S. Department of Energy, Washington, DC, February 1980.
A Statistical Analysis of State VMT Estimates in VMT Statistics, Lifetime VMT, and Current State Methods
of Estimating VMT, ORNL/TM-6327, Oak Ridge National Laboratory, Oak Ridge, Tennessee, February
1979.
G. Kulp, D.L. Greene et al., Regional Analyses of Highway Energy Use, ORNL-5587, Oak Ridge National
Laboratory, Oak Ridge, Tennessee, December 1979.
"A Regional Stock System Model of Highway Gasoline Demand," in Changing Energy Use Futures,
Volume 1, Proceedings of the Second International Conference on Energy Use Management, Los Angeles,
California, October 22-26, 1979.
and T.P. O'Conner, P.O. Patterson, A.B. Rose and D.B. Shonka, Regional Transportation Energy
Conservation Data Book, ORNL-5435, Oak Ridge National Laboratory, Oak Ridge, Tennessee, September
1978.
Econometric Analysis of the Demand for Gasoline at the State Level, ORNL/TM-6326, Oak Ridge National
Laboratory, Oak Ridge, Tennessee, July 1978.
An Investigation of the Variability of Gasoline Consumption among States, ORNL-5391, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, April 1978.
R.R. Schmitt and D.L. Greene, "Evaluating Transportation Innovations with the Intervening Opportunities
Model," Proceedings of the Northeast American Institute for Decision Sciences, April 1977.
M.G. Wolman, D.L. Greene and N.L. Froomer, An Analysis of the Physical Characteristics of the
Shanandoah River Which Affects Its Navigability and An Analysis of the Physical Characteristics of the
Potomac River Which Affects Its Navigability, prepared for the Baltimore District, U.S. Army Corps of
Engineers, Chesapeake Research Consortium, pub. nos. 37S and 38S, resp., March 1975.
G. Power, M.G. Wolman et al., Report on the Shenandoah River: An Investigation to Determine
Navigability and Report on the Potomac River: An Investigation to Determine Navigability, prepared for
the Baltimore District, U.S. Army Corps of Engineers, Chesapeake Research Consortium, pub. nos. 37 and
38, resp., March 1975.
FORTHCOMING PUBLICATIONS
C. Liu, D.L. Greene and D.S. Bunch, "Fuel Economy and CO2 Emissions Standards, Manufacturer Pricing
Strategies and Feebates", The Energy Journal, accepted for publication.
David L. Greene 19 March 2013
-------
"Energy and Environment", Ch. 8 in Transportation Statistics Annual Report 2013, Bureau of
Transportation Statistics, U.S. Department of Transportation, Washington, D.C.
"Analyzing the Transition to Electric Drive in California: Phase II, White Paper, Howard H. Baker, Jr.
Center for Public Policy, University of Tennessee, forthcoming.
C. Liu and D.L. Greene, "Consumer Choice of E85: Price Sensitivity and Cost of Limited Fuel
Availability", accepted for presentation at 93rd Annual Meeting of the Transportation Research Board,
Washington, DC, January, 2014.
David L. Greene 20 March 2013
-------
ORGANIZATIONAL CONFLICT OF INTEREST CERTIFICATE
Customer: U.S. Environmental Protection Agency
Contractor: ICF Incorporated, LLC, 9300 Lee Highway, Fairfax, VA 22031
Prime Contract: EP-C-12-011
Subcontract/Peer Reviewer: David Greene
In accordance with EPAAR 1552.209-70 through 1552.209-73, Subcontractor/Consultant certifies to the
best of its knowledge and belief, that:
A No actual or potential conflict of interest exists.
An actual or potential conflict of interest exists. See attached full disclosure.
Subcontractor/Consultant certifies that its personnel, who perform work on this contract, have been
informed of their obligations to report personal and organizational conflict of interest to Contractor and
Subcontractor/Consultant recognizes its continuing obligation to identify and report any actual or
potential organizational conflicts of interest arising during performance under referenced contract.
Subcontractor/Consultant
Date
100 Cambridgepark Drive, Suite 500 • Cambridge MA 02140 617.250.4200 617.250.4261 fax • www.icfi.com
-------
-------
James M. Sallee
The Harris School of Public Policy Studies
University of Chicago
1155 East 60th Street
Chicago, IL 60637
Phone: 773-316-3480
Fax: 773-702-2286
sallee@uchicago.edu
http://home.uchicago.edu/~sallee
Last updated: 10/23/13
(July 2008 - present)
EMPLOYMENT AND AFFILIATIONS
Assistant Professor, Harris School of Public Policy Studies
University of Chicago
Faculty Research Fellow, National Bureau of Economic Research
Public Economics and Energy and Environmental Economics Programs (April 2010 - present)
Visiting Researcher, University of California Energy Institute (August 2010 - December 2010)
EDUCATION
University of Michigan, Ph.D. in Economics (2008)
Dissertation: Three Essays in Public Economics
Committee: Joel Slemrod (Chair), Rebecca Blank, James Hines, Jeffrey Smith
University of Michigan, M.A. in Economics (2005)
Macalester College, B.A. in Economics and Political Science, Summa Cum Laude, OBK (2001)
PUBLISHED JOURNAL ARTICLES
"What Do Consumers Believe About Future Gasoline Prices? (with Soren T. Anderson and Ryan
Kellogg) Journal of Environmental Economics and Management Forthcoming.
"The Value of Honesty: Empirical Estimates from the Case of the Missing Children" (with Sara
LaLumia) International Tax and Public Finance, 20(2), April 2013, pp. 192-224.
"Car Notches: Strategic Automaker Responses to Fuel Economy Policy" (with Joel Slemrod)
Journal of Public Economics, 96(11-12), December 2012, pp. 981-999.
"Financial Reporting, Tax, and Real Decisions: Toward a Unifying Framework" (with Douglas A.
Shackelford and Joel Slemrod), International Tax and Public Finance, 18(4), August 2011, pp.
461-494.
"Using Loopholes to Reveal the Marginal Cost of Regulation: The Case of Fuel-Economy
Standards" (with Soren T. Anderson) American Economic Review 101(4), June 2011, pp. 1375-
1409.
"The Surprising Incidence of Tax Credits for the Toyota Prius" American Economic Journal:
Economic Policy, 3(2), May 2011, pp. 189-219.
"Fuel Economy Standards: Impacts, Efficiency, and Alternatives" (with Soren Anderson, Carolyn
Fischer and Ian Parry), Review of Environmental Economics and Policy, 5(1), Winter 2011, pp.
89-108.
-------
"A Cautionary Tale About the Use of Administrative Data: Evidence from Age of Marriage Laws"
(with Rebecca M. Blank and Kerwin Kofi Charles), American Economic Journal: Applied
Microeconomics,^}, April 2009, pp. 128 - 149.
"On the Optimal Allocation of Students and Resources in a System of Higher Education" (with
Alexandra M. Resch and Paul N. Courant) The B.E. Journal of Economic Analysis & Policy
(Advances Tier), 8(1), Article 11.
NON-REFEREED PUBLICATIONS
"The Energy Paradox and the Future of Fuel Economy Regulation", working paper, Institute for
Policy Integrity at New York University School of Law, December 2011
"Forecasting Gasoline Prices Using Consumer Surveys" (with Soren T. Anderson, Ryan Kellogg
and Richard M. Curtin) American Economic Review Papers & Proceedings 101(3), May 2011,
pp. 110-114.
"The Taxation of Fuel Economy" Tax Policy and the Economy v. 25, Editor Jeffrey R. Brown,
NBER: University of Chicago Press, 2011, pp. 1-38.
"Consumer Valuation of Fuel Economy: A Microdata Approach" (with Sarah E. West and Wei Fan)
Proceedings of the National Tax Association's 102nd Annual Conference on Taxation, 2009, pp.
254-259.
WORKING PAPERS
"The Economics of Attribute-Based Regulation: Theory and Evidence from Fuel-Economy
Standards" (with Koichiro Ito)
"New Evidence on Taxes and the Timing of Birth" (with Sara LaLumia and Nicholas Turner)
Submitted
"Rational Inattention and Energy Efficiency" Submitted
"The Intergenerational Transmission of Automobile Brand Preferences: Empirical Evidence and
Implications for Firm Strategy" (with Soren T. Anderson, Ryan Kellogg and Ashley Langer)
Submitted
SELECTED WORK IN PROGRESS
"Do Consumers Recognize the Value of Fuel Economy? Evidence from Used Car Prices and
Gasoline Price Fluctuations" (with Sarah West and Wei Fan)
AWARDS AND HONORS
Best Teacher in a Core Course, The Harris School (2012, 2013)
W.E. Upjohn Institute Early Career Research Grant (with Reed Walker) (2012)
Certificate of Excellence in Reviewing, Journal of Public Economics (2012)
John V. Krutilla Research Award from Resources for the Future (2009 - 2010)
National Tax Association Dissertation Award (2008)
National Science Foundation Graduate Research Fellowship (awarded 2003)
Population Studies Center Trainee Fellowship, University of Michigan (2003-2008)
-------
TEACHING
All for Master of Public Policy Students at the Harris School
Policy Approaches to Mitigating Climate Change
Topics in U.S. Tax Policy
Empirical Methods in Policy Analysis II
Science, Technology and Policy
REFEREE
American Economic Review, Journal of Political Economy, Quarterly Journal of Economics,
Journal of Public Economics, American Economic Journal: Economic Policy, American Economic
Journal: Applied Economics, RAND Journal of Economics, Journal of Environmental Economics
and Management, National Tax Journal, Journal of Labor Economics, International Tax and Public
Finance, Journal of Law & Economics, Canadian Journal of Economics, Nature, B.E. Journal of
Economic Analysis & Policy, Economic Inquiry, Journal of Human Resources, Economic Letters,
Environmental and Resource Economics, Journal of Policy Analysis and Management; Grants:
National Science Foundation, European Science Foundation, Sloan Foundation, Time-Sharing
Experiments for the Social Sciences
SELECTED PRESENTATIONS
Invited 2013: University of Pennsylvania (Wharton), Georgetown (Economics), Illinois (Economics),
Wisconsin (Economics) 2012: Maryland (Economics), Northwestern (Law), Universidad de
Chile (Business School), Oxford (Business School); 2011: Columbia (Economics), Maryland
(AREC), Syracuse (Maxwell), Illinois (Finance), Ohio State (Economics), Illinois
(Sustainability Center), NYU (Law conference), University of Illinois at Chicago
(Sustainability workshop), Treasury, EPA, Resources for the Future (Conference); 2010: MIT
(Economics), Yale (Forestry), Berkeley (ARE), Berkeley (UCEI), NBER Tax Policy and the
Economy, University of Chile; 2009: Cornell (Economics), Minnesota (Applied Economics),
North Carolina State University (Economics), Berkeley (POWER Conference), University of
Illinois at Chicago (Economics), Macalester College (Economics); 2008: Resources for the
Future, University of Chicago (Harris), University of Pennsylvania (Wharton), University of
British Columbia (Economics), University of Kentucky (Martin/Economics), University of
Indiana (SPEA), University of California, Irvine (Economics), Treasury, Ford Motor Company
Conference 2012: NBER Public Economics, National Tax Association, Michigan Tax Invitational
2011: National Tax Association, ASSA, Association of Environmental and Resource
Economics, International Institute of Public Finance, University of California Energy Institute;
2010: NBER Public Economics, Iowa State Bioenergy Camp; 2009: ASSA, National Tax
Association, Heartland Environmental and Resource Economics; 2008: APPAM, National Tax
Association; 2007: NBER Summer Institute (EEE), National Tax Association, APPAM
-------
-------
ORGANIZATIONAL CONFLICT OF INTEREST CERTIFICATE
Customer: U.S. Environmental Protection Agency
Contractor: ICF Incorporated, LLC, 9300 Lee Highway, Fairfax, VA 22031
Prime Contract: EP-C-12-011
Subcontract/Peer Reviewer: James Sallee
In accordance with EPAAR 1552.209-70 through 1552.209-73, Subcontractor/Consultant certifies to the
best of its knowledge and belief, that:
A
No actual or potential conflict of interest exists.
An actual or potential conflict of interest exists. See attached full disclosure.
Subcontractor/Consultant certifies that its personnel, who perform work on this contract, have been
informed of their obligations to report personal and organizational conflict of interest to Contractor and
Subcontractor/Consultant recognizes its continuing obligation to identify and report any actual or
potential organizational conflicts of interest arising during performance under referenced contract.
>' - IB^-^Ji;r
Subcontractor/Consultant
Date
100 Cambridgepark Drive, Suite 500 Cambridge MA 02140 • 617.250.4200 • 617.250.4261 fax —— www.icfi.com
-------
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Appendix B. Charge Letter
ICF International B-l January 31, 2014
-------
-------
December 18, 2013
Dr. David L. Greene
Senior Fellow, Howard H. Baker, Jr. Center for Public Policy
1640 Cumberland Avenue
Knoxville, TN 37996-3340
Subject: Peer Review of Light-Duty Vehicle Rebound Effect Research
Dear Dr. Greene,
ICF International has been contracted by EPA to facilitate a peer review. In late November we
corresponded by email and you indicated your availability to participate as a paid reviewer to review Ken
Small and Kent Hymel's report "The Rebound Effect from Fuel Efficiency Standards: Measurements and
Projection to 2035". You have been selected to participate on this panel. ICF will compensate you
$3,000 for your services. This charge letter provides you with a list of directed questions for your review,
the review schedule, and the materials we would like you to send to us at the conclusion of the review. In
addition, attached to this letter is a copy of the report that we would like you to review.
Charge Questions
Listed below are the four directed questions we would like you to pay special attention to when
conducting your review:
Element 1:
What are the merits and limitations of the authors' approach for estimating the vehicle miles traveled
(VMT) rebound effect for light-duty vehicles? Are key assumptions underpinning the methodology
reasonable? The VMT rebound effect is defined here as the change in VMT resulting from an
improvement in light-duty vehicle efficiency.
Element 2:
Is the implementation of the authors' methodology appropriate for producing estimates of the VMT
rebound effect? Specifically, are the input data and the methodology used to prepare the data appropriate?
Are sound econometric procedures used? Does the model appropriately reflect underlying uncertainties
associated with the assumptions invoked and the parameters derived in the model?
Element3:
The methodology used in this report attempts to account for asymmetric responses to increases vs.
decreases in per mile fuel costs (and fuel prices). Does the report's finding of an asymmetric response
seem reasonable given the methodology that the authors employed? In particular, do the authors' preferred
model specifications (3 .21 b and 4.21 b) seem appropriate for capturing driver response to an increase in
fuel efficiency?
100 Cambridgepark Drive, Suite 500 — Cambridge MA 02140 — 617.250.4200 — 617.250.4261 fax — www.icfi.com
-------
Element 4:
The report describes a methodology for projecting the VMT rebound effect for light-duty vehicles
forward in time. The concept of dynamic rebound is introduced to quantify the rebound effect over the
period of a vehicle lifetime, during which time the variables that influence the rebound effect are
changing. Is this methodology reasonable and appropriate, given the inherent uncertainty in making
projections about how future drivers will respond to a change in the fuel efficiency of their vehicles?
Schedule
The schedule for this peer review is as follows:
December 18, 2013: Charge letter distributed to reviewers
Early January, 2014: Kick-off conference call with reviewers
January 17, 2014: Comment/review due via email to Larry.orourke@ICFI.com
The kick-off conference call will be an opportunity for you to speak with the other reviewers, ICF and
EPA staff to provide you with any clarification you may require.
Materials
Upon completion of your review, you should submit your report under a cover letter that states 1) your
name, 2) the name and address of your organization, and 3) a statement of any real or perceived
conflict(s) of interest.
Should you have any questions or concerns, feel free to contact me via phone at 617-250-4226 or by
email at Larry.orourke@icfi.com. In addition, the EPA project manager for this effort is Jeff Cherry and
he may be reached at 734-214-4371. We will send you a meeting request for the kick-off conference call
shortly. Thanks for your participation!
Sincerely,
Larry O'Rourke
Manager, ICF International
Attachment: The Rebound Effect from Fuel Efficiency Standards: Measurements and Projection to 2035
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
Kenneth Gillingham Review Comments
Appendix C. Kenneth Gillingham Review Comments
ICF International C-l January 31, 2014
-------
-------
Review of Small and Hymel (2013)
The Rebound Effect from Fuel Economy Standards: Measurement and Projection to 2035
By: Kenneth Gillingham, Yale University
January 2014
Overview
This review of the final report by Ken Small and Kurt Hymel "The Rebound Effect from
Fuel Efficiency Standards: Measurement and Projections to 2035" first provides a brief
overview and then quickly turns to the four charge questions.
The report follows the methodology of Small and Van Dender (2007) and Hymel
et al. (2010), with updated data and some minor additions. This is a thoughtful and
careful effort aiming to address a difficult question: the change in VMT resulting from
an increase in light-duty vehicle efficiency across the entire United States.
The primary methodology is to bring together aggregate state-level data on driving
per adult M, fuel prices, vehicle stocks , fuel intensity, urbanization, and congestion.
The authors then estimate a system of simultaneous equations to address endogeneity
in key regressors, such as the cost per mile of driving. The system of equations is clearly
summarized in Hymel et al. (2010) as follows:
vmat = K.mvmaift_i + K.mvveht + amccongt + pfpmt + ffticaPlt + Pfxt* + UT W
veht = ctvveht-i + ccvmvmat + ftpvt + pv2pmt + ^Xvt + uvt (2)
fintt = ccffint^ + K.fmvmat + p{pft + p{cafet + ^Xft + u{ (3)
congt = Kcmvmat + cap2t + p€3Xct + ect. (4)
Here vmat is natural log of the vehicle miles travelled per adult M, veht is the natural
log of the number of vehicles per adult, fintt is the natural log of the fuel intensity (i.e.,
i/fuel economy), and congt is the log of the hours of travel delay per adult. In addition,
pmt is the log price per mile of driving, caplt is the log total length of roads divided by
state land area, pvt is the log of an index of the price of a new vehicle, cafet is a pre-
estimated measure of stringency of CAFE standards, cap2t is the log of urban lane miles
per adult, and the X's are additional variables such as the square of price, interactions
between pm and the other variables, time trends, and state fixed effects. All variables
are normalized for ease of interpretation.
The approach assumes first order autocorrelation in the error term for equations (i),
(2), and (3). Identification of the key parameter of interest (the price elasticity of VMT
demand ft™) relies primarily on within-state time series variation in M and the price of
-------
gasoline (conditional on the other covariates). The fuel cost per mile coefficient ft™ is
potentially endogenous because fuel economy itself is endogenous. This endogeneity is
addressed by including another equation for the fuel intensity (3). Equation (2) addresses
a potential endogeneity in veht and also allows for an interpretation of the effect of a
change in fuel economy on the size of the vehicle stock.
If I understand correctly, the model is estimated in the same way as Small and Van
Dender (2007), using a modified Cochrane-Orcutt transformation and nonlinear least
squares (to address autocorrelation in the context of a lagged dependent variable).
The results are presented with equation (4) (from Hymel et al. (2010)) both included
and not included. The results are largely in line with the results in the previous two
papers and other previous papers in the literature. With the updated dataset covering
1966-2009, there is a short-run rebound effect on the order of 5%, a long-run effect on
the order of 28-30%, evidence of the rebound effect declining with income, and evidence
of a greater response when gasoline prices are increasing than decreasing. There is also
some evidence of a structural break in 2003, with slightly larger rebound effects after
this year. The rebound effect is then projected forward linearly using forecasts of key
variables. When this leads to a negative rebound effect, it is replaced by zero.
Now I turn to each of the four charge questions. Since questions i and 2 are so closely
linked, I will address them together.
Elements i and 2
Element i: What are the merits and limitations of the authors' approach for estimating the
vehicle miles traveled (VMT) rebound effect for light-duty vehicles? Are key assumptions under-
pinning the methodology reasonable? The VMT rebound effect is defined here as the change in
VMT resulting from an improvement in light-duty vehicle efficiency.
Element 2: Is the implementation of the authors' methodology appropriate for producing esti-
mates of the VMT rebound effect? Specifically, are the input data and the methodology used to
prepare the data appropriate? Are sound econometric procedures used? Does the model appropri-
ately reflect underlying uncertainties associated with the assumptions invoked and the parameters
derived in the model?
There are many merits to the authors' approach for estimating the VMT rebound
effect. It tackles a difficult question using what is likely the best data publicly available
across all of the United States. It carefully considers many estimation issues and provides
estimates that appear to be reasonable. It provides a valiant (and reasonable) attempt at
forecasting the VMT rebound effect forward. There is no question that it was a major
effort and a thoughtful one at that. It would be difficult to do much better given the task
at hand.
As in any study, there are also limitations, most of which the authors recognize. All
-------
of these limitations relate to the difficulty of the question being asked. I will address
these limitations next, emphasizing unavoidable challenges of estimation and providing
a few suggestions.
i. To begin, the definition of the VMT rebound effect is vague. This is not the au-
thors fault, for they are clear about the question they intend to answer. But, the
definition, "the change in VMT resulting from an improvement in light-duty ve-
hicle efficiency," provides much room for different interpretations. It provides no
guidance on whether the improvement is costly, leading to higher vehicle prices or
costless, leading to lower vehicle prices. Similarly, it does not specify whether other
attributes of vehicles change along with vehicle efficiency. On one (unlikely) ex-
treme, one could imagine expensive improvements in light-duty vehicle efficiency
that also involve a trade-off leading to less desirable characteristics of the vehicles.
At this extreme, the number of vehicles in the fleet would decline (vehicles are
more expensive and less exciting) and at the same time driving is less exciting, so
people drive less. This would suggest a very small rebound effect. Consider an-
other (also unlikely) extreme, where improvements in light-duty vehicle efficiency
are free and lead to no change in the attributes of the fleet. This would suggest a
larger rebound effect. This extreme is the assumption made in the report. If we are
discussing a tightened greenhouse gas (GHG) standard for light-duty vehicles, the
truth could be expected to be somewhere in the middle. Put in terms of the no-
tation in the report, the methodology estimates e^ , where M includes both the
driving response and the "fleet size" response. In the report, the fleet size response
is positive, for vehicles are more efficient and no more expensive. This is entirely
consistent with what the authors state they intend to do, but not likely to be the
case in the real world. If the vehicle fleet shrinks (or stays constant), we would
expect fewer additional miles driven than in the results. Thus, for this reason the
results are likely a slight over-estimate of the rebound effect from a GHG standard.
2. A second limitation, heterogeneity, is entirely a data limitation. The authors clearly
recognize this. The only way data can be assembled on all states in the U.S. over
time is to use aggregate data at the state level. Despite improvements in data
availability in some states, this is the best we can do for all states. Using aggregate
data masks known heterogeneity in the rebound effect, which may be important
for projecting the rebound effect forward. This is recognized clearly by the authors
on page 3: "In particular, the model assumes that changes in fleet average fuel
economy will have the same impact on behavior whether those changes are caused
entirely by new vehicles entering the fleet, or partly by new vehicles and partly by
the retirement of older ones. It should be adequate insofar as the pattern of mileage
driven by vehicle age is reasonably stable; if it is not, a more fine-tuned analysis
tracking elasticities by vehicle age would reveal additional effects not captured
here." I believe this is an important caveat, given that elasticities do vary by vehicle
-------
age (I can see this in my own work). However, there is not much that can be done
about this using aggregate data. Is this a major bias? It's hard to say. It is not even
clear what the direction of the bias would be, since it could go either way. I see
this as an assumption worth noting, as the authors clearly do, and an area worth
researching further in the future. But I don't see any way around this given the
current U.S.-wide question being asked.
3. Another limitation is the reliance on within-state time series variation in the study.
Relying on time series variation is not necessarily a problem, but using a time
series over many years typically lends itself to using time series approaches. For
example, testing for the order of autocorrelation and for unit roots are common
time series approaches. To its credit, the methodology does account for first-order
autocorrelation. But what if the data are second-order autocorrelated? In this case,
the coefficients could still be consistently estimated, but the standard errors would
be incorrect. This raises a possible issue of incorrect standard errors. It is not clear
what the direction of the bias in standard errors would be.
4. Similarly, since the time series econometric approaches are not used, one might
have expected the standard panel data approach that includes time fixed effects to
be employed. The dataset would make this possible. In this case, the identify-
ing variation would be gasoline price shocks off the mean. I am sure the authors
have considered and run such a specification before. I suspect one of two things
happened: either there was not enough variation and the estimates were all statis-
tically insignificant, or the results were crazy because the variation identifying the
coefficients was not reliable variation. So instead, the paper includes linear time
trends in each equation. These are helpful and much better than nothing. They do
not control for other changes as flexibly as fixed effects, but they do retain more
variation. Another possibility could be decade fixed effects or a quadratic or higher
order polynomial in time. Would inclusion of these further time controls make a
major difference? Perhaps not, but it could be worth discussing and exploring as
further robustness checks. The direction of the bias would again be unclear. One
way in which it might not make a difference is if the time-varying unobservables
was only correlated with fuel intensity, which is effectively instrumented for in the
third equation.
5. Another limitation is the difficulty in finding great instruments for the fuel cost per
mile and fuel intensity. The system of equations can be thought of in an instrumen-
tal variables context. So the system of equations must have exclusion restrictions
(i.e., variables that are not in the first equation, but are in the third equation) in
order to address the possible endogeniety of the pmt variable. In my read of the
report and previous papers, it looks to me like the only exclusion restrictions are
the CAFE stringency variable cafet and lagged fuel intensity fintt_^ (although it is
a little odd to me that vmat is in the third equation; usually one would expect to
-------
see vmat-i so that the lagged variable is an instrument for itself). So one way of
looking at the results is that we are instrumenting for pmt with the CAFE variable
and lagged fuel intensity. Are these good instruments? Perhaps one could argue
so, although they are not obviously so. The identification of the rebound effect
does in part rest of this assumption. There is a similar assumption for the vehicle
stock variable veht, where the price of vehicles and the lagged vehicle stock are the
exclusion restrictions that help identify the vehicle stock variable veht in equation
(i). I am not going to say that these exclusion restrictions are flat-out wrong, for I
imagine you could argue for them and I personally would have a very tough time
finding much better ones in this context. The bottom line is that ($™ is a difficult
coefficient to reliably identify with aggregate data, so there is reason to be at least
somewhat cautious.
6. As the CAFE stringency variable cafet is a key exclusion restriction, it is impor-
tant to understand how it was derived. It was cleverly constructed, as a predicted
variable using vehicle efficiency data prior to the implementation of CAFE stan-
dards in 1977. In this sense, I like the variable and think it is useful. However,
given that it is a predicted variable, we know that using a predicted variable in
an estimation means that we really have a two-step estimation approach, which
requires adjusting the standard errors for the standard error in the first stage. One
could easily get around this (and address any possible autocorrelation without the
modified Cochrane-Orcutt approach) using bootstrapped standard errors. This is
what I would suggest as another robustness check. Typically, bootstrapped stan-
dard errors lead to larger standard errors, but given how statistically significant the
coefficients are in the current estimation, I would still expect statistical significance
for the key coefficients of interest. Note that the coefficients themselves would not
change.
7. A final limitation relates to the assumption of no measurement error in the vari-
ables, which may be important given the sources of the data (which to my knowl-
edge are the best available for data of this ilk). Hymel et al. (2010) provide a
very clear caveat on this point on page 1227: "Perhaps the greatest danger is that
persistent measurement error in a given state (across years) could cause an overes-
timate of the coefficient in a given equation on the lagged value of the dependent
variable. This coefficient is crucial in estimating the relationship between short-run
and long-run elasticities. Thus the rather large difference we find between these
elasticities (roughly a factor of five in the VMT equation) might be partly caused by
measurement error." I think this is a fair caveat that applies equally to this report.
If we have classical measurement error in the regressors, we would expect attenu-
ation bias of the coefficients, so ft™ could be biased downwards; thus it would be
an under-estimate of the true value. The two things that can be done for this are to
use instruments (which is done for some of the variables) and be very careful with
the data collection process, which I believe they have been.
-------
8. An assumption (not necessarily limitation) worth highlighting is the choice of a
partial-adjustment model with a lagged dependent variable. There is a long his-
tory in energy economics using partial-adjustment models. They rely on a few
assumptions. First, for consistency, there cannot be autocorrelation in the errors,
otherwise there is an endogeneity issue. I believe that the methodology in the re-
port addresses this concern. Second, for the interpretation of long-run elasticities,
one must believe that we are in a dynamic system converging to an equilibrium
response and that the structure we have put on this dynamic system is correct.
Many, if not most applied econometricians today harbor some doubts about this
approach, but we cannot rule it out. It relies on variation in the previous year's
dependent variable to provide guidance on how quickly we are moving to a hypo-
thetical equilibrium. Is this variation free of confounds? Hard to say. In any event,
it is a major assumption that may be reasonable, even if many economists feel more
comfortable with research designs where the identification is cleaner and there is
no lagged dependent variable. The robustness check that many economists would
want to see is the coefficient on pmt when the first equation is estimated separately
and without the lagged dependent variable. From Small and Van Dender (2007),
we can see that estimating the first equation separately does not change the coeffi-
cient on pmt much (an increase to -0.085). It would be nice to know what the result
would be without the lagged dependent variable as well. At the end of the day
though, these assumptions may be defensible.
To summarize, while there are many merits to this study, there are also some limita-
tions. Some are data limitations and some should best be thought of as possible concerns
that perhaps warrant further robustness checks and thought. I should emphasize that
all applied econometric work has possible concerns and it is impossible to address them
all. My overall take is that given the state of the literature, the coefficient estimates in
this report provide a reasonable sense of what the VMT rebound effect is in the U.S. on
average over the period 1966-2009.
Element 3
Element y. The methodology used in this report attempts to account for asymmetric responses
to increases vs. decreases in per mile fuel costs (and fuel prices). Does the report's finding of
an asymmetric response seem reasonable given the methodology that the authors employed? In
particular, do the authors' preferred model specifications (3.21 b and 4.21 b) seem appropriate for
capturing driver response to an increase in fuel efficiency?
This report uses a well-established approach to account for asymmetric responses to
increases and decreases in per mile fuel costs based on variation in fuel prices. There are
many energy economics papers that indicate a greater response to price increases than
-------
price decreases, and the authors find results that corroborate this literature. I believe the
sign and relative magnitudes of these results, with the caveats above applying of course.
That said, I agree with the authors in questioning whether the driver response to an
increase in fuel efficiency would be different than the response to gasoline prices. The
asymmetries could come about for two primary reasons. First, gasoline price increases
could be more salient than price decreases. Second, investments could be made when
gasoline prices are high, limiting a short-run downward response when gasoline prices
drop. Both factors probably play a role, and Figures 4.2 and 4.3 may be consistent with
both.
But if asymmetries come about because of the differing salience of increases and de-
creases in gasoline prices, should we expect the same effects to apply for changes in
vehicle fuel efficiency? My first inclination is that the answer is "not necessarily." Per-
haps the downward price movement would be the better indicator of what the response
would be to an increase in fuel efficiency, which is effectively what the asymmetric re-
sponse results do. But given that saliency of the gasoline price may be different than
saliency of the fuel price per mile, I see this as a relatively strong assumption.
The authors clearly recognize this, but must use the variation in the data that they
have. Given the strong assumption, I would be more more comfortable using the results
assuming the symmetric response. This seems to me to be a more neutral assumption,
for it is effectively the mean effect. Fortunately, it does not make a huge difference.
Element 4
Element 4: The report describes a methodology for projecting the VMT rebound effect for light-
duty vehicles forward in time. The concept of dynamic rebound is introduced to quantify the
rebound effect over the period of a vehicle lifetime, during which time the variables that influence
the rebound effect are changing. Is this methodology reasonable and appropriate, given the inher-
ent uncertainty in making projections about how future drivers will respond to a change in the
fuel efficiency of their vehicles?
Truly projecting the VMT rebound effect for light-duty vehicles forward in time re-
quires a detailed model of the vehicle stock, along with elasticity estimates for each part
of the age profile of the vehicle stock. It would involve allowing new vehicles to enter
into the stock, which would lead to several dynamics. These new vehicles are more
efficient, so they are driven more. Households also switch a bit to these vehicles from
others, likely less-efficient vehicles, reducing emissions, but perhaps leading to a slightly
more miles driven. Similarly, older vehicles are driven a bit less. As well, different types
of people may switch to the new vehicles (e.g., people who have long commutes).
The authors face real data limitations that prevent this ideal modeling of the fleet.
Instead they cleverly develop a "dynamic" rebound effect. The dynamic rebound effect
7
-------
attempts to take into account a variety of factors: the transition from the short-run to
long-run rebound effect, the change in income, urbanization/congestion over time, and
the decrease in driving from vehicles along the vehicle age profile. From my perspective,
given the caveat that a true vehicle stock model is unavailable, this approach is sound
for estimating the VMT rebound effect going forward in the next several years.
I am less comfortable linearly extrapolating as far out as 2030. It is very likely that
the relationship between the rebound effect and income is relatively linear within the
observed range of the variables, but moving forward, I believe it is less likely that the
relationship would continue. The issue is quite clear in the need to truncate the rebound
effect for any given state and year at zero. It seems more likely that there would be a
smooth decline in the rebound effect that asymptotes to a level above zero. Congestion
would reach saturation. Consumers would be wealthier so perhaps would be driving so
much more that the utility of driving on the margin is very low (which could imply a
larger rebound effect). These are just two possibilities. Perhaps with some exploration
the authors could estimate a non-linear specification a nonlinear effect that asymptotes.
If we must extrapolate out to 2030, I would feel more comfortable with this approach
than allowing the rebound effect for some states to approach zero and then be zeroed
out.
Would such an approach change the results much? I suspect not, but it is worth
considering.
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
David Greene Review Comments
Appendix D. David Greene Review Comments
ICF International D-l January 31, 2014
-------
-------
Review of "The Rebound Effect from Fuel Efficiency Standards: Measurement and Projection to 2035"
by Kenneth A. Small and Kent Hymel, December 24, 2013.
David L Greene
January 16, 2014
I have carefully read the paper, "The Rebound Effect from Fuel Efficiency Standards: Measurement and
Projection to 2035" by Small and Hymel. This review is based on the December 24, 2013 corrected
version.
Element 1:
What are the merits and limitations of Small's approach for estimating the vehicle miles traveled (VMT)
rebound effect for light-duty vehicles? Are key assumptions underpinning the methodology reasonable?
The VMT rebound effect is defined here as the change in VMT resulting from an improvement in light-
duty vehicle efficiency.
Response
The Small & Hymel (S&H) approach for estimating the direct rebound effect is theoretically and
methodologically rigorous and has been executed by the researchers without errors, to the best of this
reviewer's knowledge. It has both merits and limitations, as do all existing studies of this phenomenon.
Merits
The authors demonstrate an accurate understanding of the direct rebound effect as distinguished from
other definitions of the rebound effect. The model they have formulated and the data they use are
appropriate for measuring the direct rebound effect.
The system of equations used to estimate the rebound effect allows for fuel intensity (the inverse of
miles per gallon)1 to affect vehicle travel via, 1) the effect of a change in fuel cost per mile on miles
traveled per adult person, 2) the effects of fuel cost per mile on automobile ownership and 3) the effect
of increased travel on traffic congestion (4-equation model). This formulation allows for quantification
of the importance of these potential pathways by which fuel intensity might affect vehicle travel. The
general similarity in results between S&H's 4-equation system and their 3-equation system (omitting
congestion) adds to the evidence that the estimates are robust.
The lagged adjustment formulation used in the S&H model allows for estimation of both short-run and
long-run rebound effects. The authors have used appropriate econometric methods for estimating this
1 The terms "fuel economy" and "fuel intensity" are used throughout this paper. Fuel economy is defined as miles
per gallon of motor fuel. Fuel intensity is the inverse of fuel economy.
-------
type of model in a system of equations, taking into account the possibility that error terms within each
equation may be correlated over time, a potentially serious issue for such lagged adjustment models.
The approach makes use of a large volume of data covering the fifty states and the District of Columbia
over a period of 44 years. The source of data for vehicle travel is the U.S. Department of Transportation,
Federal Highway Administration (FHWA), which collects the data from the individual states. These data
have been scrutinized by the FHWA and checked against other data, such as permanent and periodic
traffic counts. The data are certainly not ideal (there is no ideal source for VMT data) but are very
unlikely to misrepresent year-to-year changes in vehicle travel due to the very large number of
permanent and temporary traffic counters in use across the United States. In their estimation methods,
the authors have used appropriate statistical procedures to account for any persistent state-specific
errors. Aggregate vehicle travel data, such as used in this study, are appropriate for estimating the
direct rebound effect since it is the effect of changes in fuel intensity on total vehicle travel that is of
greatest relevance to the Environmental Protection Agency's (EPA) and National Highway Traffic Safety
Administration's (NHTSA) rulemakings. Other sources of data, such as household travel surveys, cover a
large fraction of total vehicle travel but omit vehicle travel by businesses and governments and also by
heavier vehicles. Furthermore, models estimated on survey data generally do not insure that the
estimated individual household changes integrate to the total national change. Total national vehicle
travel as reported in the FHWA's table VM-1 is also a useful data source for estimating the rebound
effect but the quantity of data available is smaller by a factor of 50.
There have been many studies of the rebound effect and S&H include the most important research
papers in their review. In general, the studies based either on a national vehicle travel data time series,
time series cross-sectional state vehicle travel data or panel survey data (covering several years and
including significant fuel price changes) are very consistent with the empirical findings of S&H. S&H
demonstrate that when their estimation is restricted to the time periods covered by previous studies,
the rebound effects estimated by their method are very close to the central tendency of the studies.
Higher estimates of the direct rebound effect have come from studies in other countries and from U.S.
studies using only a single year of survey data. Statistical analysis based on a single year of survey data
is prone to spurious correlations. In general, models attempting to explain variations in vehicle travel
based on a single year of survey data have low explanatory power (in the statistical sense, i.e., low R2).
This makes controlling for factors that may influence both fuel economy and vehicle travel critical for
obtaining coefficient estimates that are not biased by correlations with omitted variables. More robust
estimates are likely to be obtained using time-series, cross-sectional data sources, such as used by S&H.
S&H have carefully investigated the possibility that the rebound effect is not constant over time. They
test this possibility first by estimating different rebound effects for different periods of time without
consideration of what might be causing any changes. They also test for a varying rebound effect by
means of a formulation the authors have used in previously published papers that estimates
correlations between the rebound effect and income and fuel price. The limitations of the latter
method for forecasting purposes are discussed below. However, the authors have shown significant
correlations and have proposed a plausible theoretical explanation for the results.
-------
S&H have also investigated the possibility that fuel price and fuel intensity may affect vehicle travel
differently, and that fuel price (or fuel cost per mile) rises and reductions may not have equal effects.
This is important because the rebound effect, strictly speaking, pertains only to fuel intensity and not to
fuel price yet many studies rely on estimation methods that constrain the elasticities of fuel price and
fuel intensity to be equal but opposite in sign. For the purposes of the EPA, it is the effects of fuel
intensity reductions that are of interest rather than the effects of fuel intensity increases. In theory, the
effects could be symmetrical but, as S&H note, there is a substantial literature that indicates that market
responses to fuel price rises and fuel price reductions are not symmetrical. The rigorous investigation of
this issue is a valuable contribution about which more will be said below. The results confirm that
responses to fuel price or fuel cost per mile reductions are smaller than the responses to increases.
They also find that it is not possible to estimate a statistically significant effect of fuel intensity alone
using their data and methods. This latter result is consistent with the small number of other studies that
have reported on this issue.
The inability to estimate the separate effects of fuel price and fuel efficiency on VMT is worthy of further
investigation. The authors' decision to proceed using fuel cost per mile is consistent with the
interpretation that this outcome is caused by a poor sample design for the fuel efficiency variable. That
is, the fuel efficiency of the on-road vehicle fleet changes very gradually and thus tends to follow a
smooth trend, making it difficult to distinguish the effects of fuel intensity from other smoothly trending
variables. In addition, state-level fuel economy is not directly measured but estimated by the states by
various methods (e.g., by dividing fuel use by vehicle travel). Fuel prices on the other hand, have
changed relatively quickly and by relatively large amounts. Fuel prices are also based to a large extent
on direct measurements. This makes it easier to accurately estimate at least the short-run price effect.
The authors' decision is therefore a prudent one given the information available. It is also appropriate
for them to note that, if anything, it is more likely to result in an overestimate of the rebound effect.
Given the above, the authors recommend using the rebound effect estimated using cost per mile, which
constrains the price and fuel intensity elasticities to be equal in magnitude and opposite in sign. This is
the most important assumption of their study, since without it the estimated rebound effect would not
be statistically significant from zero. They also note that this assumption, in all likelihood, leads to an
overstatement of the rebound effect. Their decision seems prudent although it is a subjective one and,
strictly speaking, not supported by the empirical data. The alternative would be to assign a value of zero
to the rebound effect. This, however, would imply that drivers do not behave rationally from an
economic perspective, since they would treat changes in cost caused by changes in the price of fuel
differently from changes in cost due to changes in fuel intensity. Economic theory suggests that such a
conclusion should itself be supported by more evidence than the lack of statistical significance of the
fuel intensity coefficient. S&H present their reasoning on this issue transparently, as they should.
Limitations
In their review of the literature, S&H should have included the important review of studies of the
rebound effect by Sorrell (2007) and companion reports by the UK Energy Research Center (Sorrell and
Dimitropoulos, 2007; Dimitropoulos and Sorrell, 2006). Since the UKERC study is a review of the
-------
literature, by itself it does not add much new material to the S&H review but it does cover more of the
literature and reaches conclusions that support S&H's interpretation of the literature.
The definition of the rebound effect on p. 6 is the definition appropriate when fuel economy
improvements come about due to pure technological change. That is, the improvement in fuel economy
does not involve trading off purchase cost or other vehicle attributes (e.g., size, acceleration) for fuel
economy. The rebound effects of fuel efficiency due to pure technological change versus fuel economy
standards are almost certainly different. Technological change shifts the trade-off between fuel
economy and cost (or other attributes) while standards generally cause manufacturers to move to a
different location within the same trade-off function. Of course, technological change is always
occurring and there is the likelihood that standards induce technological change but the basic point
remains valid since standards, in general, will induce a trade-off of fuel economy for other vehicle
attributes, especially manufacturing cost. For the purposes of evaluating the EPA/NHTSA rule makings,
trade-offs with vehicle cost are highly relevant.
Although the study does a good job of recognizing and describing a wide range of pathways for the
rebound effect, it omits part of the effect of increased vehicle prices on the long-run cost per mile of
travel. According to all studies of which I am aware, including the rule making itself, the 2025 fuel
economy and greenhouse gas (GHG) standards are expected to result in an increase in the long-run cost
of manufacturing vehicles. The increased cost will cause an increase in vehicle transaction prices,
assuming only that vehicles' selling prices increase with increasing long-run average cost. The S&H
model allows the increase in vehicle price to affect VMT through the effect of new vehicle prices on the
vehicle stock and the effect of vehicle stock on vehicle travel. But an increase in the capital cost of a
vehicle also affects the long-run cost of vehicle travel via usage-induced capital depreciation. This
mechanism is not included in either the 3-equation or 4-equation versions of the model and could be
important because capital costs are a large fraction of total vehicle ownership costs.
The potential for feedback effects to be generated via institutional processes is appropriately
acknowledged but a potentially important one is missing. That is the effect of major fuel economy
improvements on highway user fees. In the past, fuel economy improvements have been second only to
inflation as a threat to Highway Trust Fund revenues (e.g., Greene, 2011). Historically, motor fuel taxes
have been raised by federal and state governments in order to maintain adequate funding for highway
construction and maintenance. Whether this will continue to be the case in the future and what type of
tax may be used (possibly one that does not fall on motor fuel) are open questions but certainly relevant
ones. Raising motor fuel taxes would, ceteris paribus, increase the retail price of motor fuel, thereby
increasing the fuel cost per mile of travel and partially offsetting the rebound effect of fuel intensity. A
careful review and analysis of this subject would likely lead to the conclusion that raising fuel taxes in
order to maintain highway user fee revenues should be included in regulatory analyses of the rebound
effect. This is not something that S&H need to include in their econometric analysis but it should be
mentioned in the discussion of possible institutional effects.
-------
Summary for Element 1
In brief, S&H's study is a technically proficient assessment of the rebound effect of fuel economy on
vehicle travel using appropriate state-level vehicle travel and associated data. The conclusions drawn
are well supported by the empirical analyses in this paper and, in general, by the previous literature.
The authors have made several important contributions:
1. Re-estimating the rebound effect using more recent state-level data and demonstrating the
consistency of their historical estimates with the central tendency of the existing literature.
2. Estimating the effects of income and fuel price on the size of the rebound effect over time and
showing the ability of these factors to statistically explain a large portion of the apparent
changes.
3. Testing the potential asymmetry of response to increases and decreases in fuel cost per mile.
The analysis also shows that the asymmetric response to fuel price changes implies a smaller
rebound effect than that found assuming a symmetric response to fuel cost per mile.
4. The projections of future rebound effects are useful but may understate the rebound effect in
cases where many states' rebound effects approach zero. This is likely a consequence of the
linear functional form and truncation rule and could be an artifact of those assumptions.
Overall, this paper makes an important contribution to the literature and, like the authors' previous
work, represents the current state of knowledge about the rebound effect of motor vehicle fuel
economy on vehicle travel.
Element 2:
Is the implementation of the Small methodology appropriate for producing estimates of the VMT
rebound effect? Specifically, are the input data and the methodology used to prepare the data
appropriate? Are sound econometric procedures used? Does the model appropriately reflect underlying
uncertainties associated with the assumptions invoked and the parameters derived in the model?
Response
The S&H method represents best practice and is appropriate for producing estimates of the rebound
effect. As discussed above, the data used are well suited to the problem. The econometric methods are
also appropriate and consistent with the state of practice. Incorporating uncertainty, on the other hand,
poses a difficult challenge that has not yet been given much attention in the literature on the rebound
effect. There are uncertainties due to data shortcomings, issues with the experimental design available
in the historical record, uncertainties due to model formulation, uncertainties inherent in econometric
estimation and uncertainties about the future state of the world. S&H have addressed many of these
issues by constructing alternative projections based on different assumptions. These are useful.
However, adequately addressing uncertainty and incorporating it into a projection methodology
requires an identification of the nature of the uncertainties to be included, which should follow from the
purpose for representing uncertainty. It is not clear to this reviewer what the goal of including
-------
uncertainty is, and therefore it is not possible to give a definitive response concerning the S&H method's
handling of uncertainty.
Do to Definitions
It would be helpful to the reader for S&H to spend a little more time explaining the nature of the state
level data. According to this reviewer's understanding, state level data include VMT and fuel use by all
vehicle types, not only the light-duty vehicles affected by past fuel economy regulations. This introduces
substantial heterogeneity in the vehicle populations across states, from motorcycles to diesel-powered
18-wheelers, although light-duty vehicles still predominate. Fuel intensity is believed by this reviewer to
be total state highway use of motor fuel (not only gasoline) divided by total state highway vehicle travel.
It would be helpful to clarify these definitions in the report to alert the reader to the meaning of the
data and possibly help interpret the results. It is likely that state-specific constants will account for
much of the differences across states in the composition of traffic. Remaining effects of heterogeneity
are not likely to cause important problems for estimating the rebound effect.
Cost per Mile versus Fuel Intensity Rebound
The question of whether the data actually support the existence of a rebound effect for fuel economy
has been addressed above and is mentioned again here to emphasize its importance and the
uncertainty it creates. The estimates presented by S&H are based on the maintained hypothesis of
economically rational behavior, in the sense that consumers are assumed to respond to changes in fuel
cost per mile in the same way whether caused by a change in fuel price or a change in fuel economy.
However, the new research presented by S&H concerning the asymmetry of responses sheds new light
on this subject, as explained in greater detail in Element 3. The consequence of the analysis of
asymmetry is that there is now strong evidence that the market response to reductions in fuel intensity
(a goal of the fuel economy and GHG standards) is less than the response of the market to increases in
the price of fuel, and that it is also smaller than estimates of the rebound effect based on the
assumption of a symmetrical response to changes in fuel cost per mile. This finding of S&H is potentially
of major significance. It implies that the best estimate of the rebound effect for the purpose of
estimating the effects of fuel economy and GHG standards is the asymmetric elasticity of reductions in
fuel cost per mile. Since it is a relatively novel result with respect to the rebound effect, further research
is warranted, yet the results presented by S&H are strong and should now represent the current state of
knowledge.
Statistical Insignificance of Endogenous Variables in Some Equations
S&H do not provide an adequate discussion of the fact that some of the endogenous variables in either
the 3- or 4-equation models are not statistically significant. For example, in the 3-equation models, vma
does not appear to be statistically significant at the 0.05 level in any of the equations for vehicle stock,
and pf+vma is not statistically significant in the equation iorfint in models 3.3, 3.18, 3.21b, or 3.29
(Table Bl). In the 4-equation models, pf+vma is not statistically significant in the equation iorfint in
models 4.3, 4.13, 4.21, and possibly 4.23. This calls into question the necessity for the simultaneous
equation framework, at least as formulated, and requires explanation. The secondary, simultaneous
-------
equation effects are small relative to the direct effect of pm in the vma equation and so the empirical
significance of these pathways is not great but it would be interesting to see if the hypothesis of
simultaneity is rejected by the data or not.
In particular, the equation forfint in table 4.2 raises questions. Why it is preferable to interact fuel price
and VMT rather than test also for the main effects of the two variables? As explained on p. 30, the
interacted fuel cost variable turns out not to be statistically significant. This result increases the need
for an explanation of the choice of this formulation. Is this a parsimonious way of getting both variables
into thefint equation? Would they be less statistically significant individually? And if neither vehicle
travel nor the price of fuel is statistically significant in the equation for fuel intensity, doesn't this
undermine the rationale for including this equation in a system of equations? If this is the best
formulation and yet the log of fuel price times VMT is not statistically significant in the equation forfint
then it would seem that the data do not support including fint in a simultaneous equation formulation.
Again, this does not appear to be of great practical importance since the simultaneous equation effects
are relatively small.
A Caveat on Long-run and Short-run Effects and Lagged Adjustment Models
The lagged adjustment model used by S&H is a useful formulation and widely adopted for modeling
phenomena such as aggregate VMT. However, it implies two important maintained hypotheses. The
first is that the correlation between the dependent variable and its lagged value measures only the
adjustment process. If there are other causes of a strong positive correlation, the long run elasticities
will be overestimated. By using econometric methods that allow for error correlation in the lagged
adjustment equation S&H have taken a prudent step to deal with possible correlation between the
current and lagged values of the dependent variable from that source. Second, it implies the same
adjustment rate for all variables, which would seem to be a special case. These observations do not
diminish the value of this analysis or others using the lagged adjustment formulation but are something
to be borne in mind when interpreting results.
The Effect of Vehicle Cost on Vehicle Use
The S&H model allows changes in vehicle price to affect vehicle travel via its effect on the size of the
vehicle stock. However, this may not be adequate since increased vehicle cost also affects the cost per
mile of travel to the extent that use of a vehicle depreciates its value. There is no question that capital
depreciation is a component of the long-run cost per mile of travel. The question is how important it is
as a determinant of long-run travel demand.
Estimates of the elasticity of total vehicle travel with respect to car purchase cost were found in at least
one literature review to have a central tendency of -0.19 in the short run and -0.42 in the long run
(Goodwin et al., 2004, table 7). While the plurality of studies reviewed come from the United States, the
majority do not. In addition, it is not clear from the study cited how many studies combine the effect of
purchase cost via the size of the vehicle stock with the effect of purchase cost via long-run cost per mile.
Nonetheless, for illustrative purposes only, I will use the -0.4 elasticity. If a doubling of fuel economy
caused a 10-20% increase in VMT at a cost of $2,000 per vehicle for vehicles with a prior average cost of
-------
$25,000, the 8% increase in vehicle cost would reduce VMT by about 3%, offsetting 15-30% of the
estimated rebound effect. This reviewer is not arguing here that these numbers correctly represent the
magnitude of this possible effect for the United States but rather to illustrate the possibility that there
may be an important issue here that is worthy of formal investigation.
Element 3:
The methodology used in this report attempts to account for asymmetric responses to increases vs.
decreases in per mile fuel costs (and fuel prices). Does the report's finding of an asymmetric response
seem reasonable given the methodology that Small employed? In particular, do the authors' preferred
model specifications (3 .21 b and 4.21 b) seem appropriate for capturing driver response to an increase in
fuel efficiency?
On this subject, S&H have produced potentially important results. Their analysis supports the inference
that rebound estimates based on a symmetric response to fuel cost per mile overstate the rebound
effect of fuel intensity. The price asymmetry model has been found in other studies of the response of
gasoline demand to gasoline price and petroleum demand to petroleum price. Thus, it is very likely that
the difference between rises in fuel cost per mile and decreases in fuel cost per mile is attributable to
asymmetric market responses to rises in the price of fuel and not to asymmetric responses to changes in
fuel intensity. This would mean that the symmetric model, by estimating an average effect of rises and
reductions in fuel cost per mile, would overestimate the effect of reductions in fuel cost per mile. S&H's
results confirm this. This result is important because it implies that for purposes of estimating the
rebound effect of fuel economy regulations, the asymmetric elasticity of a reduction in fuel cost per mile
should be a more accurate estimate of the rebound effect than the fuel cost per mile elasticity
estimated assuming a symmetric relationship between fuel cost per mile and vehicle travel.
The analysis of the possibly asymmetric effects of fuel price rises and cuts appears to be separating price
effects (which are asymmetric) from fuel intensity effects (which are not asymmetric). As the authors
explain, in the asymmetric model the rebound effect is mathematically the sum of the asymmetric
effects. The partial effect of fuel efficiency (holding other variables constant) does not depend on
whether prices are rising or falling. Rather, it is the effect of the price of gasoline that depends on
whether prices rise or fall. Thus, this reviewer concurs with the authors' decision to adopt this result in
their preferred models (3.21 b and 4.21 b), especially since these empirical results are also consistent
with their earlier inference that by using fuel cost per mile alone (based on a symmetric model) one
would almost certainly overestimate the rebound effect.
Generally, two possible explanations are put forward for the asymmetrical response to fuel price rises
and cuts. The first is that consumers are more likely to extrapolate fuel price rises than cuts and thus
respond more strongly to fuel price rises when purchasing durable goods. The second explains the
persistence of asymmetry in the long run as a consequence of technological change or public policy (i.e.,
8
-------
efficiency standards) induced by fuel price rises. In either case, the asymmetry method used in this
section of the paper should be able to separate these irrelevant effects from the rebound effect.
Empirically, the effects of fuel price and fuel intensity changes are not the same (see above). The
asymmetric model offers a logical explanation of the conundrum.
Because the media and price volatility effects almost certainly apply to the effects of price but not fuel
efficiency, the authors are correct in abandoning the models including these variables. The anomalous
results in certain formulations also support this decision. As the authors note, the erratic behavior of
the Asymmetry model 3.23 suggests that it is not a plausible model. Because the asymmetric models
are also not able to separately estimate the fuel price and fuel efficiency effects, as the authors note,
their preference for the models of section 4.4.1 is well reasoned.
In this and previous work, S&H have found that the rebound effect varies with income. In this study,
they also found that it varies with the price of fuel. It would be interesting to test whether changes in
the distribution of income as well as average income have affected the rebound effect. There is some
evidence that the distribution of income has affected the growth rate of aggregate VMT. The result that
the rebound effect varies with income and fuel price is both important and useful for analyzing the
future costs and benefits of fuel economy and GHG regulations.
Element 4:
The report describes a methodology for projecting the VMT rebound effect for light-duty vehicles forward
in time. The concept of dynamic rebound is introduced to quantify the rebound effect over the period of a
vehicle lifetime, during which time the variables that influence the rebound effect are changing. Is this
methodology reasonable and appropriate, given the inherent uncertainty in making projections about
how future drivers will respond to a change in the fuel efficiency of their vehicles?
The dynamic rebound model provides a reasonable method of accounting for the fact that as fuel
economy improvements penetrate the vehicle stock, new vehicles have higher fuel economy than older
vehicles. What is not clear is how much of an improvement this method makes over basing the rebound
effect on the vehicle miles weighted average fuel intensity of the vehicle stock. If distributional effects
were important (if it were important to know how much the usage of different vehicles changed),
detailed modeling of changes in vehicle use in the vehicle stock by vintage would be necessary. It is not
clear that this is necessary for EPA's analysis of the costs and benefits of fuel economy regulations. That
said, there is no compelling reason not to use the dynamic method proposed by S&H.
In this and previous papers, the authors have presented strong evidence that the rebound effect has
changed over time and that the changes are correlated with changes income and fuel price. The income
result was also confirmed in a recent study using national time series data (Greene, 2012). There is also
theoretical justification for including these effects, since income affects the value of travelers' time and
-------
fuel prices affect the fuel cost share of the long-run cost per mile of travel. Thus, it is appropriate to
include these effects in the forecasting model. While there is uncertainty about future incomes and fuel
prices, basing the estimated rebound effect on price and income assumptions used elsewhere in the
estimation of costs and benefits of the standards will result in a more consistent assessment. That said,
the linear extrapolation of the income and price effects is problematic. Whatever the correct functional
form may be, it is not linear over the full range of possible future incomes and fuel prices. This leads to
the problem of rebound effects with theoretically implausible signs, which the authors have addressed
by truncation at zero. Truncation at zero is better than not truncating at zero. A better functional form
should be sought that approaches zero as income goes to infinity and fuel price goes to zero.
Final Comments
The S&H analysis is very well done, uses appropriate models, data and econometric methods and makes
several important contributions to knowledge of the rebound effect. The results are consistent with
both the central tendency of other estimates in the literature and with the best studies contained in the
peer-reviewed literature. The range of issues investigated and statistical tests performed is a particular
strength of the analysis. The projected rebound effects are useful and plausible. The results are useful
to EPA as they now stand. The issues raised in this review and those noted below suggest avenues of
additional research and model development that may or may not lead to improvements in the model as
currently recommended by S&H.
The chief limitations of the study are the possibly inadequate representation of the effect of vehicle
purchase costs on the long-run cost per mile of vehicle travel, the need for an interpretation of the lack
of statistical significance of key endogenous variables in many of model equations, and the truncation of
the rebound effect in the projecting model when the estimated rebound effect becomes negative.
It is appropriate to adopt the models that include the asymmetric response to reductions in fuel
intensity (models 3.21b and 4.21b) as the current best estimates of the rebound effect. The finding of
asymmetry in the elasticity of cost per mile should be incorporated in the projection methodology. It is
statistically significant and consistent with the peer-reviewed published literature. It also addresses the
inability to estimate a significant elasticity for fuel intensity alone and the conclusion that the rebound
effect is thereby overestimated. Use of the "price cut" elasticity of fuel cost per mile from the
asymmetric model has the advantage of at least removing the fuel price rise asymmetry from the
estimated rebound effect.
S&H's investigation of how the rebound effect may vary systematically with other factors is an
important contribution to the understanding of the rebound effect. Incorporating rebound effects that
vary with income (value of time) and fuel price (fuel cost share of operating costs) in forecasting the
rebound effect is supported both theoretically and empirically. The fact that the rebound effect varies
with both income (interpreted as representing the value of time) and fuel price (perhaps representing
the fuel cost share of the long-run costs of vehicle travel) suggests that an alternative model formulation
explicitly including all the important long-run costs of vehicle travel (and the elasticities of substitution
10
-------
among them) might produce an improved forecasting model. Such an approach might also permit
inclusion of use-related depreciation as a component of the cost per mile of travel. Fuel cost is not the
only component of the long-run cost of vehicle travel. The short-run cost of travel includes the traveler's
time and the long-run cost includes many factors, notably the capital cost of the vehicle.
The assumption of constant elasticity (as a function of income and fuel price) should be considered only
one possible functional form. In particular, it is recommended that forecasts of the rebound effect be
based on a more explicit representation of the total cost of vehicle travel, including fuel, maintenance,
capital and travelers' time costs. Because in the end S&H are left with only a partial explanation for the
apparent increase in the rebound effect after 2003, understanding the correct functional form of the
rebound effect should be given a higher priority.
It would also be appropriate to update the projected rebound effect estimates using the most recent
Annual Energy Outlook (e.g., 2014 Early Release). Undoubtedly this was not available at the time the
study was carried out.
11
-------
References
Dimitropoulos, J. and S. Sorrell, 2006. The Rebound Effect: Microeconomic Definitions, Extensions and
Limitations, Working Paper, UK Energy Research Centre, London, April.
Goodwin, P., J. Dargay and M. Hanly, 2004. "Elasticities of Road Traffic and Fuel Consumption with
Respect to Price and Income: A Review", Transport Reviews, vol. 24, no. 3, pp. 275-292.
Greene, D.L., 2012. "Rebound 2007: Analysis of National Light-Duty Vehicle Travel Statistics", Energy
Policy, vol. 41, pp. 14-28.
Greene, D.L., 2011. "What's Greener than a VMT Tax? The Case for an Indexed Energy User Fee to
Finance U.S. Surface Transportation," Transportation Research D-Environment, vol. 16, pp. 451-458.
Sorrell, S., 2007. The Rebound Effect: an assessment of the evidence for economy-wide energy savings
from improved energy efficiency, UK Energy Research Centre, London, October.
Sorrell, S. and J. Dimitropoulos, 2007. UKERC Review of Evidence for the Rebound Effect, Technical
Report 2: Econometric Studies, UKERC/WP/TPA/2007/010, UK Energy Research Centre, London, Octover.
12
-------
Peer Review of December 2013 LDV Rebound Report by Small and Hymel
James Sallee Review Comments
Appendix E. James Sallee Review Comments
ICF International E-l January 31, 2014
-------
-------
THE HARRIS SCHOOL
PUBLIC POLICY I THE UNIVERSITY OF CHICAGO James M. Sallee
Assistant Professor
773.316.3480
sallee@uchicago.edu
January 20, 2014
Larry O'Rourke
ICF International
9300 Lee Highway, Fairfax, VA 22031-1207
Dear Dr. O'Rourke:
Attached to this letter please find my peer review of "The Rebound Effect from Fuel Efficiency
Standards: Measurements and Projections to 2035" written by Kenneth Small, with contributions
by KentHymel.
I have no conflicts of interest relevant to the report or the contents of this review.
Sincerely,
James M. Sallee
Assistant Professor
University of Chicago
1155 EAST 60TH STRKKT I CHICAGO, II. 60637 I PHONE 773.702.8400 I I:AX 773.702.0926 I WWW.HARRISSCHOOL.UCHICAGO.EDU
-------
Summary statement:
"The Rebound Effect from Fuel Efficiency Standards: Measurement and Projection to 2035",
written by Kenneth Small (with contributions from Kent Hymel], uses an appropriate
methodology and defensible assumptions. It uses the best available data (given significant
constraints on what is available], and emphasizes modeling choices and specifications that
are sensible and consistent with both theory and data. As a reviewer, I agree with most of
the assumptions and emphases in the paper. Where I do disagree (detailed below], I
believe that the preference of one method or specification over the other involves an
element of subjective judgment about how to weigh the costs and benefits of different
approaches. I did not identify any issues that I believe are objectively incorrect. Thus, while
I might have made some different choices myself, I believe that the choices made in the
report are defensible.
My detailed comments are included below in a numbered list, categorized according to the
four charge questions that were given to me by ICF International. I did not restrict myself
to comments on how the immediate report ought to be changed given realistic constraints
on time and effort; many of my comments are intended to point to areas where future
reports could, in my opinion, make the biggest improvements. My comments should be
read in that light.
Before proceeding to those comments, two issues are worth highlighting. First is a bi
picture question regarding methodology and data. This report uses data aggregated to the
state-by-year level over five decades. Recent research (e.g., work by Kenneth Gillingham
and joint work by Chris Knittel and Ryan Sandier] has made use of microdata from vehicle
odometers, which is available for some cars in some recent years in some states. The
aggregate data used in the Small report analyzed here suffer from measurement problems
(detailed below, see item 7] and limit the available econometric identification strategies
(see items 1-3]. The odometer microdata suffer from limited coverage, both across states
and over time, and existing estimates are focused on a short-run elasticity that is
inconsistent with some of the measures emphasized in the report. In the end, which data
and methodology should be preferred likely depends on exactly what specification one
wishes to use. I think that a case can certainly be made for sticking with the aggregate data
used in the Small report, but I suspect that, in the near future when researchers have
gained access to data from a somewhat more representative set of states and have a few
more years worth of data, that the case for the microdata will become stronger. In any case,
it would be very valuable to know how projections based on the microdata estimates
compare to those used here, were it possible to construct such projections.
The second issue worth highlighting is how the report models the relationship between
income and the rebound effect for use in projections. In brief, the literature seems
consistent in finding evidence that the rebound effect varies over time and that, on a
decadal time scale, the effect is smaller in more recent years than in prior decades. The
paper posits that this may be due to rising income. This is theoretically sensible in that the
-------
total cost of driving involves a cost of time as well as a cost of fuel, and as income rises, so
does the wage and hence the time cost of driving, which eventually comes to dominate the
price per mile. In the report's projections, with income projected to rise, the rebound effect
is quickly driven to zero in many states, which greatly affects the final estimates. But, given
the nature of the identification, which relies on time series correlations between income
and the rebound effect (see items 1-3], it is difficult to have confidence that income is the
driving factor. Even if income is the driving force in the historical data, it is not certain that
it will continue to have the same relationship in the future. One must make a stand on the
relationship between income and this elasticity, and the one that the paper makes is
consistent with economic theory and with the data.
Thus, as with many modeling decisions, I think the paper's choice on how to handle this is
defensible, though alternative choices might be defensible as well (see item 11}. I highligh
this issue in particular because it appears to be pivotal to the results. Below, I include a few
thoughts on how the projections might be refined (item 13} and how this issue might affect
which results are most useful to report (item 14}. Here, I want to make the point that an
additional analysis that could corroborate the relationship between income and the VMT
elasticity would be very valuable.
I would find it reassuring if the cross-sectional relationship between income and the
rebound effect was similar to the estimated aggregate time-series relationship. According
to the research cited in the report, the available microdata evidence suggests otherwise; it
finds that the rebound effect is U-shaped in income. A rationale for this is that wealthier
people have more travel options, which makes them more responsive. This factor competes
against the time cost factor, and at different levels of income different factors dominate,
resulting in a U-shape. The projections might change significantly if the relationship
between income and the elasticity is U-shaped in the time series. This depends on whether
or not future aggregate income is high enough to reach the upward sloping portion of the U.
Rather than using a cross-section of microdata, one could look at a cross-section of states
states (or countries} to see how estimated elasticities are correlated with income. For
example, one could estimate the VMT-elasticity separately for each state for some span of
years (say, a decade} not controlling for income and then see how that correlates with state
income. Are wealthier states less responsive? One might reasonably argue that the cross-
sectional relationship between income and the VMT elasticity is a fundamentally different
parameter than the over-time relationship, but they seem to me to be based on the same
theoretical arguments. As a result, I would like to see some sort of corroborating
evidence—either in this report or in a completely separate study—though I recognize that
the suggestions made here are themselves far from perfect.
-------
Element 1: What are the merits and limitations of the authors' approach for estimating the
vehicle miles traveled (VMT) rebound effect for light-duty vehicles? Are key assumptions
underpinning the methodology reasonable? The VMT rebound effect is defined here as the
change in VMT resulting from an improvement in light-duty vehicle efficiency.
1. The paper uses a panel regression, but it is best understood as deriving results form
time-series variation because the panel regressions do not include time period fixed
effects and the lion's share of variation in the key measures come from the time-
series. In most cases, the extra credibility that is often attributed to panel data
models comes from their ability to include both time and entity level fixed effects.
The report does not use time fixed effects, and generally has very sparse controls for
time. The most important variable in the analysis is the price of gasoline. This is
measured at the state-year level, but once state fixed effects are controlled for, a vast
majority of the variation in the data will be attributable to fluctuations in the global
oil price (or national gasoline price}.
I do not necessarily advocate that the paper add time period fixed effects; if year
fixed effects were added, the remaining variation in gasoline prices that would
identify the coefficients would be state-specific fluctuations in gasoline prices in
each time period, which often represent short run imbalances in local supply and
demand that should not be expected to persist (and therefore may have a limited
impact on behavior}. In that sense, the report uses the best available variation, but
this implies that the paper's results are largely driven by the national time series in
gasoline prices and VMT, which has implications discussed in the next two points.
2. The nature of the panel identification means that, in my judgment, the additional
benefit of having 51 states as opposed to 1 national time series may be somewhat
overstated. I do not see mention in the paper of any attempt to control for
correlation across states in error terms. The standard way of handling this is to
cluster standard errors on some larger level of observation, the rule of thumb being
"at the level of variation in the key independent variable". Given my argument above
that identification is driven primarily by the national price of gasoline, one might
interpret this as implying that standard errors should be clustered at the time
period level (year}, though technically most of the variables vary at the state-by-
year level. I suspect that if the standard errors were clustered on time period that
much of the added precision that results from moving from a national time series to
a panel regression would be lost. To be clear, none of this implies bias in any
coefficients, but the confidence one might have in distinguishing between certain
specifications might be reduced by attention to the standard errors. As with other
issues, I believe there is ambiguity here, and one could perhaps defend more
vigorously the decision not to cluster.
3. The nature of the panel identification also opens the possibility for standard omitted
variable bias problems. With sparse time controls and trending variables, anything
-------
that is correlated with gasoline prices as well as with VMT per adult could induce
bias. Some factors that might be relevant are the fraction of driving that is personal
as opposed to work-related,1 the quality of automobiles,2 commuting norms,
changes in the fraction of families with two wage-earners, the expansion of urban
sprawl, etc. This is especially important for an analysis that spans so great a time
frame. The report attempts to control for measures of the most important variables,
but it is a priori difficult to be confident that all such secular trends have been
accounted for by a limited number of demographic variables. What is usually done
in response is to (a] show precisely how sensitive the coefficients of interest are to
the inclusion of the available set of controls and (b] show the robustness of the
coefficient to many additional tweaks.
Along these lines, an appealing permutation would be to add state-specific time
trends, and to add differential time trends for different periods of time where we
have reason to believe that there might be structural breaks. (The appendix to the
2007 working paper indicates that three distinct time trends are used, but this
includes a single trend for all years after 1980, which may be inadequate. Moreover,
I did not see the set of time controls used spelled out clearly in the current report] I
suspect that the author has tried these permutations, and I recognize that the tests
for structural breaks in the data do not yield conclusive results upon which to base
these decisions. But, I would hope to see greater evidence of robustness of the
results to richer controls for time, and perhaps to a broader set of demographic and
vehicle market controls.
4. The report argues that a secondary pathway through which CAFE standards might
impact VMT is through the overall size of the car market. The idea is that fuel
economy standards will cause people to buy more cars because fuel efficiency
standards lower the cost of driving, which thus increases the value of owning a car,
holding prices constant. (This is the difference between M and M in the report] This
argument is present in much of the related literature.
I find this objectionable from a theoretical point of view. In a standard market
model, the imposition of fuel economy standards could not raise the value of cars
(net of price] on average. The market should be offering cars that have a bundle of
attributes that maximizes private value to consumers. The introduction of fuel
economy standards forces automakers to alter the mix of attributes they offer—
perhaps through changes in technology or through a shift from size and
1 The price sensitivity of miles driven for work is likely different than miles driven for
personal reasons because of the difference in who is paying for fuel and whether time is
uncompensated. The data used on VMT do not distinguish these types of driving.
2 The time cost of driving is a function of the opportunity cost of driving and of the flow
utility of being in the car. More comfortable cars with improved media, and cell phones,
may substantially lower the cost of driving in that dimension.
-------
performance to economy. If standards force this mix to be altered, it is counter to
theory to suggest that this will create attribute mixes that consumers prefer,
conditional on price (which is controlled for in the regressions}.
This reasoning could be wrong if another market failure exists, such as the idea that
consumers are myopic and thereby underappreciate the value of fuel economy. In
that case, consumers could conceivably have increased private utility from the
standard. But, even this scenario does not rationalize an increase in the size of the
vehicle market because, if consumers are myopic, then they won't recognize that the
new vehicle fleet is preferable—the market was providing the fleet that seemed to be
value maximizing. This suggests that the market should shrink. It seems to me that
the final effect on market size depends on whether the standards raise or lower
producer mark-ups over marginal cost in equilibrium, which is theoretically
ambiguous.
Importantly, the report de-emphasizes this channel, which is found to be quite
small. So, while I disagree at points with the report on this issue, I do not think it has
an important impact on the final projections.
5. This report introduces measures of media attention, which are new to the literature.
This is used in two ways, one is as an additional regressor, another is as an auxiliary
data series useful for aiding interpretation. I agree with the latter usage, but not the
former. Media mentions of gasoline prices is not well motivated as an independent
regressor from a theoretical standpoint. It is meant, I believe, as a measure of the
salience of gasoline prices. But, the media surely reflects public attention as much as
it dictates it. Thus, it is fundamentally endogeneous. As such, I prefer models that do
not include it as a regressor.
At other times in the report, the media mention series is looked at by itself as an
interesting time series that might help interpretation. I think it is appropriate to use
in this sense—if it is a proxy for an endogenous measure of salience or awareness,
then it may be useful to look at this series and see if it happens to line up with the
time pattern of coefficient estimates from the baseline model, as a way of perhaps
interpreting what is going on in the main estimates. In the end, the report does not
emphasize these results over others, which mitigates my concern.
6. One weakness of the aggregated data used in this report is that it provides no
immediate way of modeling the relationship between vehicle age and VMT. Given
the lack of data on this, it seems appropriate for the report to abstract from such
issues, but this points to another area where the odometer microdata could be
useful. Those data could be used to detail the age-VMT relationship and to see how it
changes over time and in response to fuel price shocks and regulation. Such
information might be especially useful in refining the dynamic rebound effects
emphasized in the report.
-------
Element 2: Is the implementation of the authors'methodology appropriate for producing
estimates of the VMT rebound effect? Specifically, are the input data and the methodology
used to prepare the data appropriate? Are sound econometric procedures used? Does the
model appropriately reflect underlying uncertainties associated with the assumptions invoked
and the parameters derived in the model?
7. The report suffers from crucial data limitations, of which the author and the
literature more broadly are well aware. The key problem is that most of the
dependent variables are not independently measured, but are instead imputed
based on possibly inconsistent procedures across states and over time and through
a methodology that is not well explained by the Federal Highway Administration. To
recap, states generally have good data on gallons of fuel sold, because they collect
taxes by the gallon. States themselves, or the FHWA, use some estimate of fuel
efficiency of the vehicles on the road to translate gallons sold into VMT (M], b
calculating that M = F / E-hat, where E-hat is their estimate and F is fuel consumed.
The fuel intensity is measured in the report as 1/E = F/M, where again VMT is
imputed based on E-hat. Then, Gas Price per mile is calculated as Gas Price / E = Gas
Price * M/F = Gas Price / E-hat. Thus, the measurement of all of the most important
variables depends on some estimate of efficiency that states are using, which may be
inconsistent across states and over time, or that the FHWA is using, which at best is
based on surveys 5-years apart and may be wiping out differences across states by
using national averages for imputation. Any errors in measuring E are being passed
through the system because it is an input into all of the relevant variables, which
may create mechanical correlations across all of the variables of interest.
The author is aware of these issues and articulates them (although much of the
discussion is found only in the working paper version of Small and Van Bender], so
raising the issue would be belaboring the point, but for three reasons. One is that
this fundamental concern about data is an argument for shifting regulatory impact
analysis from the type of methodology used here and towards a reliance on the new
odometer-based microdata sooner rather than later.
A second is that it raises some concerns about the CAFE variable used in the paper,
which is imputed based on the relationship between fuel economy and VMT in the
years before CAFE. What were states or the FHWA doing to impute fuel economy
before EPA ratings existed in 1978? This is especially important because the CAFE
variable used in the paper, which is theoretically very clever, is based entirely on a
projection forward from data on fuel economy demand for the period before CAFE
was in place, which is a period in which there were no government measures of fuel
economy. How could states have had meaningful estimates of the on-road fuel
economy of the vehicles in their state prior to those years? Why do we think that
consumer demand for fuel economy would be the same before and after labels were
introduced? How did they even know how efficient were the models in the earlier
7
-------
years?
A third is that it is worth pointing out that the relationship between gallons of
gasoline consumed (the only thing actually measured directly] and VMT depends on
average on-road fuel economy, not EPA ratings. As driving conditions vary, the
relationship between VMT and on-road economy will differ. In particular, in
observations with greater urbanization and greater congestion, the more miles will
be spent in settings that garner lower average mpg for a given vehicle. A recent
working paper by Ashley Langer and Shaun McRae suggests that there is huge
variation in on-road fuel economy for identical vehicles.
8. There are some important differences in the estimates depending on whether or not
the latest years of data are included. I think it is arguably preferable to omit the
financial crisis, which would include both 2008 and 2009 in annual data. The paper
does not report results that omit only those two years. One might make the case that
the baseline specification should include data only up to 2007.
Elements: The methodology used in this report attempts to account for asymmetric
responses to increases vs. decreases in per mile fuel costs (and fuel prices). Does the report's
finding of an asymmetric response seem reasonable given the methodology that the authors
employed? In particular, do the authors'preferred model specifications (3.21 b and 4.21 b)
seem appropriate for capturing driver response to an increase in fuel efficiency?
In brief, I agree with the choice of models 3.21b and 4.2 Ib as the preferred model.
9. There are two types of asymmetry discussed in the analysis. One is that drivers may
respond differently to changes in fuel economy than to changes in fuel prices, so that
price-per-mile is not a sufficient measure of the price to which consumers respond. I
am sympathetic to the idea that there could be a difference, primarily because of the
salience of the fuel price. However, I think that the appropriate null hypothesis,
based on theory, is that consumers make decisions based on price-per-mile. In the
absence of compelling evidence that consumers react differently to the two
components of price, I think that the report should focus on estimates that assume
symmetry in this dimension. This is what the report chooses to do, and it is reflected
in the preferred models of 3.21b and 4.21b.
10. The second type of asymmetry is in whether the rebound effect is different for price-
per-mile increases as compared to decreases. The report ultimately favors a model
in which fuel price increases yield larger responses than fuel price decreases, and it
is deemed preferable to use a model based on asymmetry of fuel price, not
asymmetry of price per mile.
Here, I think the preferred specification is more ambiguous than with regard to the
other symmetry question, but I am in agreement with the author on the preferred
-------
methodology. There does seem to be sufficiently strong evidence of an asymmetric
response, in this paper and throughout the literature, to use a model that allows for
this difference.
Theoretically, it is sensible to assume that the asymmetry lies in increases or
decreases in the cost per mile (e.g., model 3.29], but the added econometric
challenge of solving the additional endogeneity problem that is induced by this
specification leads me to conclude that models based on asymmetry in fuel prices
(not price per mile] are preferable, for practical reasons. Thus, I agree with the
report's choice of models 3.21b and 4.2 Ib as the baseline preferred model.
Element 4: The report describes a methodology for projecting the VMT rebound effect for
light-duty vehicles forward in time. The concept of dynamic rebound is introduced to quantify
the rebound effect over the period of a vehicle lifetime, during which time the variables that
influence the rebound effect are changing. Is this methodology reasonable and appropriate,
given the inherent uncertainty in making projections about how future drivers will respond to
a change in the fuel efficiency of their vehicles?
In summary, I think that the paper makes defensible projections. That is, all of the
assumptions used in the models that are projected out to 2035 are reasonable. I
agree with the report that the baseline statistic should be the dynamic rebound
effect, which is the most theoretically relevant statistic for most applications.
11.1 do think, however, that an appealing alternative is to simply take the best available
estimates of the rebound effect from recent years, say 2000 to 2007, and project this
forward as a constant rebound effect over all future years without conditioning on
changes in income and other interacted variables. This alternative is dubious in that
it assumes that whatever conditions are at work in the most recent decade of data
will continue to be true in the future. But, it avoids dangers of extrapolating out of
context. That is, in the face of the inherent uncertainty in making projections two
decades into the future, a conservative methodology is to simply take the best
available recent estimate and assume that it will be constant in the future. If I were
the author of the report, I would provide such an estimate alongside the dynamic
rebound effects that are reported. An additional benefit of this alternative is that it
would allow for direct comparison to the projections that would come from using
odometer microdata estimates of the rebound effect, which could be used for this
"straight line" projection, but may be harder to integrate into the dynamic estimates
emphasized in the report.
12.1 do have a question/concern about the way that fuel price volatility is represented
in the projections. My understanding is that the AEO projects a smooth gasoline
price into the future. This is fine for models that do not include asymmetry, but for
models that do include asymmetry, a smoothly evolving gasoline price series and an
alternative that has the same average trend but experiences movement up and
-------
down around the trend will not produce the same rebound effect.
If this correct, then it is important for the models using asymmetry of adjustment to
fuel price increases and decreases to be based on some reasonable projection of
volatility. (I have in mind using the AEO projection of gasoline prices and the annual
volatility around a trend from the last 20 or 30 years to draw random forecasted
paths of the gasoline price, and then averaging the rebound effect projections that
result over many such paths.} I suspect that this will increase the rebound effect for
the asymmetric models, but that the effect on the forecasts will be small.
13. With rising income, the rebound effect is driven to zero in the projections, but the
effect is truncated at zero so that it cannot become negative. Might it be preferable
to truncate at a value above zero? Even as average income rises in the next two
decades, many individuals will remain at lower income levels and would therefore
be expected to remain responsive to fuel costs. Thus, it is hard to see the logic in
expecting that the average rebound effect could go all the way to zero in the near
future, so that some baseline above zero may be a more appropriate point of
truncation. It would be ad hoc to choose some point, but 0 is actually an ad hoc point
itself, given that it is meant to represent an average.
14. There is a great deal of uncertainty surrounding the final projections, due to
uncertainty in the estimated coefficients, the possibility of model error, and the
uncertainty in the forecasted inputs (like the price of gasoline and future income}.
The report lists point estimates for forecasts and includes a few different
specifications and three forecasted futures that vary the path of the future price of
oil. Additional representations of uncertainty might be appropriate.
A first possibility is to include standard errors around the forecasted values that
reflect the sampling uncertainty in the model estimation (i.e., the standard errors on
the coefficients}. This should be conceptually straightforward, though it multiplies
the number of numbers that must be reported in a table by two (though it is just
shading in a figure}.
The price of oil makes a substantial difference to the bottom line estimates. Thus,
depending on what the EPA foresees as the final use of this report, it may be worth
providing additional detail about the oil price scenarios that the AEO is using (are
these meant to represent extremes of a spectrum of plausible paths? Or are they
likely scenarios?}. Or perhaps additional results should be presented. That would
depend on the intentions of the user of the report.
A fuller version way of representing forecast and coefficient uncertainty is to model
the uncertainty in the forecasted variables and provide a collection of different
model results based on random draws of these variables. I think this would be
useful in making clearer which parameters are really pivotal, so users know where
10
-------
to draw their attention. If, for example, all that really matters is income growth
relative to oil price growth, then I would like to see a focus on that relative
parameter and to have spelled out for me why the range of estimates actually span
the useful set of scenarios to study. I recognize that this is a tall order and would
perhaps require a substantial separate analysis.
In terms of model error, which is more difficult to represent, the report lists
projections for several different specifications, which is useful. The one thing that
could perhaps be useful is to provide some explicit comparison, along the lines
mentioned above, of how these projections differ from a projection that uses just the
VMT elasticity estimate taken from the most recent decade of data and projected
forward without reducing it based on income and other demographic trends (the
straight line projection}.
11
-------
Draft Report - Peer Reviewed Version,
"The Rebound Effect from Fuel Efficiency
Standards:
Measurement and Projection to 2035"
-------
The Rebound Effect from Fuel Efficiency Standards:
Measurement and Projection to 2035
Kenneth A. Small
Department of Economics
University of California at Irvine
with contributions by
Kent Hymel
Department of Economics
California State University at Northridge
fe
Final Report
December 24, 2013
This report discusses empirical values of the "rebound effect" for travel in passenger vehicles in the
United States. The rebound effect refers to effects on the amount of travel that arises from changes in the
fuel efficiency for light-duty motor vehicles (passenger cars and light trucks), caused in turn by
regulations or technological developments. We briefly discuss the literature, then summarize previous
empirical estimates done at University of California at Irvine in collaboration with Kurt Van Dender and
Kent Hymel. Finally we present updated empirical estimates, which take advantage of newer data through
the year 2009, and derive the implications of the updated estimates for the rebound effect in the time
frame 2010-203 5.
The literature review and empirical methodology are described more fully in two published articles
(Small and Van Dender 2007a; Hymel, Small, and Van Dender 2010), and even more fully in the working
papers from which the published articles ware derived (Small and Van Dender 2007b). The empirical
estimates have been updated subsequently, by adding five new years of data, namely 2005-2009. The
projections are our own, and use a new methodology developed for this project which improves on that
-------
used for earlier reports by K. Small to EPA and an older report to the California Air Resources Board
(Small and Van Dender 2005).
1. Background and definitions
1.1 Determinants of motor-vehicle travel
The rebound effect is simply a statement of the near-universal economic principle of downward-sloping
demand: when the price of a good or service decreases, people purchase more of it. In this case the
service is passenger transportation, and its price to the user includes the cost of fuel. If the amount of
service is measured as vehicle-miles traveled (VMT), then the component of price accounted for by fuel
cost, here called "fuel cost per mile" PM, is equal to the price of fuel P/(e.g. stated in $/gallon) divided by
fuel efficiency E (e.g. stated in vehicle-miles/gallon):
:Pf/E.
Thus if fuel efficiency E is increased, fuel cost per mile decreases, and since this is part of the price paid
by consumers to drive, they will increase their VMT. See Greening, Greene and Difiglio (2000) for a
more extended discussion.
The responsiveness of demand to price is often summarized as a ratio of the percent change in demand,
AM/M, to the percent change in price causing it, &PM/PM, where M designates VMT in mathematical
equations and A designates a change in a quantity. A ratio such as this is called an elasticity, usually
defined for the situation when A/Vis very small so that the ratio becomes a derivative. Therefore we
define the elasticity of vehicle-miles traveled with respect to cost per mile as follows:
_PM dM_
£M,PM ~~
r
M dPM
-------
where the derivative dM/dPMis simply the limit of AM/APMas A/V becomes very small. An equivalent
way to write this is in terms of the natural logarithms of M and PM, which we denote by lower-case letters
vma andpm, respectively. (The notation vma stands for vehicle-miles per adult member of the population,
which is how we define M in our empirical work.) Of course the equation for vma contains other variables
besides pm, and these are held constant when considering the effects ofpm; this makes the derivative in
(2) a partial derivative, denoted using the symbol d. The elasticity written in this form is:
d(vma)
~, r? (3)
d(pm)
which could be a single coefficient in the equation for vma or, if pm enters in more than one way, a
combination of several coefficients.
One of the confusing aspects of the literature is that few studies have accounted for the fact that fuel
efficiency E is not simply mandated, but chosen jointly by consumers and motor-vehicle manufacturers,
within certain constraints set by regulation. Therefore one might ask the meaning of considering a change
in E as though it could simply be set by fiat. In our empirical work, Van Bender and we meet this
challenge by defining a system of three simultaneously determined travel-related quantities, each
applying to a state. The first dependent variable is VMT, written mathematically as M; it is a function of
PM (as already described), the size of the vehicle fleet, V, and various socio-demographic characteristics
including income. The second dependent variable is V, which is a function of several things reflecting the
demand for owning vehicles: a price index/V of new vehicles, the amount of travel M (since new
vehicles are purchased in large part to supply desired travel), the price of travel PM, and other
characteristics. Note that we do not distinguish among vehicles of various ages: thus implicitly we ignore
possible effects of these variables on the age composition of the fleet. Finally, the third dependent
variable, fuel intensity \IE (the inverse of fuel efficiency), is presumed to be chosen based on a
combination of motives including the wish to conserve on the cost of traveling M miles, the need to meet
various regulations on fuel efficiency and/or emissions, and tradeoffs with vehicle performance; in our
empirical work E is assumed to be a function of M, price of fuel PF, a variable measuring the stringency
during any given year of the US federal Corporate Average Fuel Economy (CAFE) regulations, and other
characteristics. This system is summarized in the left panel of Table 1.1.
Table 1.1. Simultaneous Equation Systems
Three-equation system
Four-equation system
-------
(without congestion)
Equation (dependent
variable)
VMT per adult
Vehicle stock per adult
Fuel intensity of vehicles
Symbol
Level
M
V
HE
Logarithm
vma
vehstock
fmt
(with congestion)
Equation (dependent
variable)
VMT per adult
Vehicle stock per adult
Fuel intensity of vehicles
Congestion delay per adult
Symbol
Level
M
V
HE
C
Logarithm
vma
vehstock
fmt
cong
An implicit assumption in the use of aggregate data is that that the response to aggregate changes in fuel
efficiency (or other variables) does not depend significantly on how those changes are distributed among
segments of the population. This could occur, for example, if drivers are sufficiently homogeneous. In
particular, the model assumes that changes in fleet average fuel economy will have the same impact on
behavior whether those changes are caused entirely by new vehicles entering the fleet, or partly by new
vehicles and partly by the retirement of older ones. This assumption enables us to apply the results of the
model to regulations that specifically impact new vehicles only. It should be adequate insofar as the
pattern of mileage driven by vehicle age is reasonably stable; if it is not, a more fine-tuned analysis
tracking elasticities by vehicle age would reveal additional effects not captured here.
It is worth noting that our system accounts for the effects of a change in regulations through two potential
pathways. We illustrate for an increase in fuel-efficiency standards, with no change in vehicle price. First,
the regulations increase the average fuel economy of the fleet, and that in turn reduces the cost per mile of
travel, PM, through equation (1); this may directly reduce the amount of travel because of downward-
sloping demand as just discussed. Second, the size of the vehicle fleet may increase because vehicles are
now more useful, in the sense that they can be driven more cheaply; this change in vehicle fleet size may
further affectM since, as already noted, M is expected to be a function of Fas well as other things. We
estimate a simultaneous-equations model ofM, V, and E that fully accounts for these effects. Empirically,
we find that the first path is by far the dominant one, so that one could ignore the second path as an
approximation; this may simply indicate that vehicle purchases are governed mainly by factors other than
the cost of driving.
Our model, through the influence of fuel cost on fuel efficiency, implicitly incorporates some changes in
the relative prices of vehicles of different sizes and types. (For example, vehicle manufacturers may
respond to a fuel efficiency regulation by offering discounts on their fuel-efficient vehicle types.)
However, the description just given of the effects of regulations assumes that the average price of new
-------
vehicles, Pv, is held fixed. Of course, the full effect of a regulation would also include any change in this
average price on new-vehicle sales. In many cases this would work in the opposite direction to that arising
from a change in fuel cost: if fuel cost declines due to regulations that force manufacturers to raise vehicle
prices, those higher prices would tend to reduce vehicle sales and thus, ultimately, travel, thereby
offsetting some of the rebound effect. Furthermore, changes in new-vehicle sales would also change
scrappage rates and the price structure of used vehicles of different ages. These effects are not usually
considered part of the "rebound effect", although that is just a matter of definition. Hence they are not
discussed here;1 but they are important to consider as part of the full effects of a regulatory change.
In order to distinguish the ultimate effect of both pathways on VMT, we use the symbol M to designate
the combined effect, and designate its elasticity with respect to cost per mile as s^ pM, reserving the
symbol SM PM for the changes operating through the first pathway only. Small and Van Bender (2007a)
show that these quantities are related by:
_ SM,PM + SM,VSV,PM ,,-.
SM,PM ~ i _ ~ *> '
1 GM,VGV,M
where SM,V denotes the direct elasticity of travel with respect to vehicle fleet, SV,M denotes the direct
elasticity of vehicle fleet with respect to amount of travel, and SV,PM denotes the elasticity of vehicle fleet
with respect to cost per mile of travel. All the quantities on the right-hand side of (4) are measured
directly as coefficients, or combinations of coefficients, of the three equations in our model.
In later published work in collaboration with Kent Hymel, the model described above was extended to
account for the interrelationship between travel and congestion, denoted by C and measured empirically
by estimated annual hours of delay due to congestion per adult. To accomplish this, a fourth equation is
added to the model predicting the amount of congestion in a state, averaged over both its urban and non-
urban areas. At the same time, the equation for vehicle-miles traveled is modified to include an influence
from congestion. The expectation is that more VMT causes congestion to rise, but that rise in congestion
also inhibits VMT. The result of these simultaneous influences is captured by the simultaneous estimation
and application of the VMT and congestion equations.
1 In principle the effect of any specified changes in average new-vehicle price due to regulations could be analyzed
using the results of the vehicle-fleet equation in our model, since that equation includes the variable Pv, which is an
index of nationwide new-car prices. However, the model does not estimate the coefficient of new-vehicle price very
precisely, because there is little variation in that variable (none across states); so we would have less confidence in
using it for that purpose. Probably a better approach for analyzing effects on vehicle purchases would be to consider
the entire range of vehicle sizes and models and how consumers shift between them.
-------
The result is that in the four-equation model, which includes congestion, equation (4) is modified by
adding an additional term in the denominator:
where SM,C is the direct elasticity of VMT with respect to congestion (presumably negative), and
conversely SC,M is the direct elasticity measuring how congestion is created by VMT (presumably
positive). The combined additional term, -SM,C-SC,/W, is expected to be positive (because the minus sign
cancels the negative sign of SM,C); therefore its presence reduces the magnitude of the rebound effect.
However, Hymel, Small, and Van Dender (2010) find this reduction to be numerically small, and more
than offset by the effects of other changes in the specification of the model and of including three
additional years (2002-2004) in the data used to estimate it.
1.2 Definition of the rebound effect: short-run and long-run
While terminology differs among authors, E^ m is conceptually what most writers have meant when
discussing the rebound effect. To summarize: it measures the ratio of the responsiveness of travelers to
the change in fuel efficiency resulting from regulations (with both expressed in percentage terms), while
recognizing that the change in fuel efficiency is not directly set by regulations but rather results from a
complex interactive process. This responsiveness accounts for both the direct effect of fuel efficiency on
the cost of using a given vehicle, and the indirect effect on travel through changes in the number of
vehicles purchased, but all the while holding average new-vehicle prices constant.
Our analysis, like nearly all in the literature, assumes that this responsiveness to fuel efficiency arises only
through the effect of fuel efficiency on fuel cost per mile. However, this assumption is debatable and is
not inherent in the definition of the rebound effect. Thus, one could posit that VMT responds to fuel price
PF and the exogenous components of fuel efficiency E separately and not just as a function of their ratio
PM=PF/E. We explore this question at several points in this report, but basically are unable to resolve it
conclusively.
-------
Because the elasticity £^ m is expected to be negative, it is convenient to express the rebound effect b
as a number that is normally positive:
.s
(5)
It is also common to express the rebound effect as a percentage rather than a fraction. Thus, if s^ pM
=-0.2, we say the rebound effect is 20%.
The empirical equation systems just discussed also account for the slowness with which changes can
occur, especially changes in the vehicle fleet size and average efficiency, which require purchases and
retirements of vehicles. They are able to do this because we observed a location (a state or District of
Columbia) every year - making the data set a cross-sectional time series, sometimes also called apanel
data set. Slow adjustment is accounted for by assuming that each of the three behavioral variables
explained by the models (M, V, and E) depends not only on the factors already mentioned, but also on the
previous year's value of that same quantity (called a lagged value of that variable). This is equivalent to
assuming that there is a desired level of M, V, or Fint=\IE, and that any deviation between this desired
level and the level attained in the previous year is diminished in one year by a fraction (l-cu), where a is
the coefficient of the lagged value of the variable. We allow a to differ across the three equations and
denote its corresponding values by of", av, and of. Note that congestion formation is an engineering rather
than a behavioral relationship, so no lag is postulated for that equation.
This slow adjustment process means that the short-run response (that occurring in the same year) is
smaller than the long-run response. Continuing to use the notation s^ pM for the elasticity determined
within this system, it is now a short-run elasticity because the long-run response is accounted for
elsewhere in the equation (through the lagged variables). We represent the corresponding short-run and
long-run rebound effects as bs and bL, respectively. They are approximately related by:
Li — f
^_ = ^f
\-am \-a
-------
where of" is the coefficient of the lagged dependent variable in the equation explaining vma. A more
precise relationship accounts for the fact that in the full three-equation and four-equation systems, the
lagged values in more than one equation can affect the long-run response; specifically, the long-run
rebound effect for the three- and four-equation models are:2
7 L _
~
(l-am}-amvavm 1(1- av}
- am - amcacm) - amvavm 1(1 - av)
where:
• av is the coefficient of the lagged dependent variable in the equation explaining the logarithm of
vehicle stock;
• cf" is the coefficient of vehicle stock in the equation explaining vma;
• avm is the coefficient of vma in the equation explaining vehicle stock;
• amc is the coefficient of congestion in the equation explaining vma;
• acm is the coefficient of vma in the equation explaining congestion; and
is the coefficient of pm in the equation explaining vehicle stock.
In addition to accounting for lagged values within the system determining our dependent variables, our
empirical system accounts for the possibility that the error terms in each equation are correlated over
time. That is, for any given state, the unknown random factors affecting a dependent variable may have
some elements that are the same year after year. Most of these common factors are accounted for by a
"fixed effects" specification, in which a distinct constant term is estimated for every state instead of just
one for the entire system.3 Empirically, the effects of lagged dependent variables are difficult to
distinguish from those of autocorrelation, a problem plaguing earlier studies investigating changes over
2 See Small and Van Dender (2007a), equation (7); and Hymel, Small, and Van Dender (2010), equation (14a).
3 This is one of two common specifications for panel data, the other being "random effects." A hypothesis test
known as a Hausman test soundly rejects random effects in favor of fixed effects for this data set.
-------
time; we are able to distinguish them because of the long time period covered by our panel data set: 36
years in the 2007 published paper, 39 years in the 2010 published paper, and 44 years in this report.
There are many ways besides those considered here that regulations on fuel efficiency or related
quantities might affect travel. As already noted, such regulations may raise vehicle prices, which would
affect the vehicle fleet size and thus, indirectly, the amount of travel. Regulations may affect fuel prices
through the impact of aggregate demand for fuel on petroleum markets. They may influence technological
developments, thereby affecting the costs and performance of future vehicles. A broader analysis of the
effects of fuel efficiency on travel might account for such factors, but they are outside the realm of the
"rebound effect" as we define it here and as most researchers have used the term.4 An advantage of our
more restricted definition is that it is a purely behavioral measure, not depending on supply factors (e.g.
the cost to manufacturers of meeting efficiency standards) or macroeconomic conditions (e.g. the
responsiveness of world oil prices to a particular policy in the US), and thereby more likely to be a stable
number applicable to many situations. However, it is important to be aware that if regulations raise the
price of new vehicles, then the response to that price rise would tend to offset somewhat the rebound
effect, as defined here, by curtailing the number of vehicles available to travelers. Similarly if regulations
curtail U.S. oil demand enough to lower world oil prices and this translates into a lower domestic gasoline
price, some additional travel will be stimulated as a result.
1.3 Dynamic rebound effect
A vehicle owner responds to a change in fuel efficiency not just in the first year or some hypothetical year
in the distant future, but continuously over that lifetime. Thus, the partial adjustment mechanism
postulated here, which is the basis for the distinction between short-run and long-run responses, implies a
continuing gradual change in VMT each year over the vehicle's life. But at the same time, the driving
force itself, i.e. the short-run rebound effect (5), is changing because the interaction variables that help
determine it (income, fuel cost per mile, urbanization, and possibly congestion) are changing. Thus, the
vehicle owner adjusts dynamically to both sources of change simultaneously. The results of tracking this
process can be expressed as the percentage increase in the vehicle's lifetime VMT divided by the
percentage decrease in fuel cost per mile that caused it. That ratio is here called the dynamic rebound
effect.
4 Greene (1992) and Gillingham (2011) refer to our definition, combined with any effect due to higher vehicle
prices, as the "direct" rebound effect. This constrast with the "indirect" rebound effect caused by income effects
(people having more money to spend after fuel purchases on other goods that use energy) and the "macroeconomic"
rebound effect (changes in energy use arising from effects of an energy policy on economy-wide prices and growth
rates). See Gillingham (2011, pp. 25-26).
-------
Calculating the dynamic rebound effect requires disaggregating the vehicle fleet by age, even though that
was not done in estimation. Thus, it involves an interpretation of what is happening within the aggregates
in the observed data. Specifically, the calculation relies on the assumption mentioned earlier that drivers
react the same way to a hypothetical difference in fuel cost per mile whether it occurs at time of purchase
or later. It works as follows. Consider the owner of a vehicle purchased in year t deciding how much to
drive in year (H-r). This owner is postulated to have a target amount of travel based on the average annual
mileage for vehicles of age r, adjusted for the short-run rebound effect as calculated by (5) using values
of interacting variables for year (H-r). Most of these interacting variables (income, urbanization, and
congestion) are simply as projected for that year. The other, fuel cost per mile, is projected based on fuel
prices for year (H-r) but holding fuel efficiency constant at the value that prevailed when the car was
purchased (year t).5
But this target mileage is not achieved immediately, because of the adjustment lags measured by the
coefficient am of the lagged dependent variable in the VMT model. The partial adjustment mechanism
implies that the actual mileage Mt in year H-r will be the weighted average of the previous year's
mileage, MT.\, adjusted for the natural evolution due to the age-mileage profile p T j, and the target
mileage, with weights am and (l-am), respectively:
where b Lt+r is the long-run rebound effect in year M-rfor a vehicle purchased in year t, and Mr is the
normal mileage for a car of this age: thus (\ — b t+T~)MT is the target mileage. The dynamic rebound
effect ut is then the fractional increase in mileage over the car's entire life that results from a fractional
increase S\r\ fuel efficiency:
5 The underlying hypothesis here is that it is new vehicle owners whose travel changes, and this calculation tracks
how it changes over that and subsequent years. Since the model itself does not distinguish new vehicle owners, the
change in fuel efficiency they experience is diluted by the fuel efficiency of existing used vehicles (assumed
unchanged by the regulations, as discussed earlier). But the resulting change in VMT of new vehicle owners is also
diluted by VMT of existing vehicle owners, so that the ratio which defines the rebound effect still applies to the
aggregates.
10
-------
The full calculation is described in somewhat greater detail in Appendix C.
Thus, for example, suppose a regulation in year 2020 results in a fractional increase Sin fuel efficiency of
new vehicles purchased that year. Income is rising and fuel price is falling, starting in year 2020 and
lasting over those vehicles' lifetimes. (Roughly this is what is projected in the "Low oil price" scenario
presented later.) Then the "target" response of VMT to a change in fuel efficiency for a new vehicle
purchased in year 2020 is getting smaller in magnitude as the vehicle ages, due to the effects of
interacting variables. But at the same time the driver is gradually adjusting to the change that began in
that year, meaning the response is shifting gradually from the short-run response to the long-run response.
These two forces work in opposite directions so the net result could be to either raise or lower the rebound
effect; in practice it usually implies a dynamic rebound effect between the short-run and long-run values.
In effect, this calculation takes account of both the gradual transition from short run to long run behavior
over the life of the vehicle, and the changing values of the rebound effects indicating changing
responsiveness to fuel cost. Iteration of (8) over additional values of r shows that all the terms in the
numerator of (9) are proportional to S, so the value chosen for c> does not affect the result.
2. Prior Literature
The first part of this section of the report is adapted from the review by Hymel, Small, and Van Bender
(2010), covering literature mostly before 2000—but with the addition of a recent meta-analysis covering
that same literature. The second part updates the review with a discussion of more recent studies.
2.1 Earlier Literature
Prior research has measured the rebound effect for passenger transport using a variety of data sources and
statistical techniques. Most but not all estimates lie within a range of 10 to 30 percent (expressing the
elasticity as an absolute value and as a percentage instead of a fraction). Greening, Greene, and Difiglio
11
-------
(2000) and Small and Van Bender (2007a) contain more complete reviews of the earlier literature. A few
key contributions are highlighted here.
The great majority of estimates are based on one of three types of data. The first and probably least
satisfactory is a single time series, either of an entire nation or of a single state within the U.S. Examples
are Greene (1992) and Jones (1993). These studies have difficulty distinguishing between autocorrelation
and lagged effects, and of course suffer from a small number of data points.
Second, some studies have instead used state-level panel data, most often from the US Federal Highway
Administration (FHWA). Haughton and Sarkar (1996), using such data from 1970-1991, estimate the
rebound effect to be 16% in the short run and 22% in the long run. They account for endogenous
regressors, autocorrelation, and lagged effects. Their study is comparable in many ways to that of Small
and Van Bender (2007), although the latter uses a longer time period, 1966-2001, and estimates three
equations simultaneously explaining VMT, vehicle stock, and fuel efficiency. Small and Van Bender
estimate the rebound effect to be 4.5% in the short run and 22.2% in the long-run on average, and also
find evidence that it has declined substantially over time due mainly to rising per-capita incomes. Barla et
al. (2009), applying the Small and Van Bender methodology to Canadian data, obtain short- and long-run
rebound effects of around 8% and 20%, respectively. Bue to their shorter time series (1990 to 2004) and
more limited cross section (15 provinces), they are not able to investigate changes in these elasticities
overtime.
A third type of data is from individual households. Mannering (1986), using a US household survey, finds
that how one controls for endogenous variables in a vehicle utilization equation strongly influences the
estimated rebound effect. He estimates the short- and long-run rebound effects (constrained to be
identical) to be 13-26%. Goldberg (1998) estimates a system of equations using data from the Consumer
Expenditure Survey for years 1984-1990. In a specification accounting for the simultaneity of the two
equations, she cannot reject the hypothesis of a rebound effect of zero. Greene, Kahn and Gibson (1999)
estimate the rebound effect to be 23% on average using a simultaneous-equation model of individual
household decisions. West (2004), using the Consumer Expenditure Survey for 1997, obtains a somewhat
larger VMT elasticity higher than these other studies, although her focus is mainly on how behavior
differs across income deciles.6
6 West reports an elasticity of VMT with respect to total operating cost (not just fuel cost) of -0.87 in the most fully
controlled specification. Presumably this is a long-run elasticity. If fuel accounted for 50 percent of operating cost,
roughly consistent with Small and Verhoef (2007, p. 97), this would imply an elasticity with respect to fuel cost per
mile of -0.435. As West notes, there are other reasons why this elasticity is not strictly comparable to others in the
literature, one being that it represents a behavior for the entire household with fuel efficiencies (hence fuel cost per
mile) averaged across its vehicle holdings.
12
-------
The studies based on individual households in a single cross-section suffer from a limited range for fuel
prices, a key variable for understanding the rebound effect. This disadvantage is partly overcome by
Dargay (2007), who observes repeated cross sections of different individuals in the UK. She estimates
short- and long-run rebound effects of 10% and 14%, respectively, but suggests that this long-run value
may be an underestimate.
Three reviews—Goodwin et al. (2004), Graham and Glaister (2004), and Brons et al. (2008)—provide
systematic statistical analyses of various studies. In the first two, estimated short- and long-run rebound
effects (based on fuel-price elasticities) average about 12 percent and 30 percent, respectively. In the
third, which is a meta-analysis of 43 studies containing 176 distinct elasticity estimates, the implied
rebound effects are larger: 17 percent short run and 42 percent long run for the United States, Canada, and
Australia.7 Brons et al. also find that studies using lagged values have a slightly smaller rebound effect
(by about 3 percentage points) than these values.8 Although the study by Brons et al. separately identifies
elasticities of driving per car and of car ownership, just as we do, they have only three observations of the
former and fifteen of the latter; so in fact their coefficients are mostly identified by variations among
studies of total price elasticity of gasoline consumption, and thus are only an indirect measure of the
responsiveness of driving.
Most of the studies just reviewed agree on long-run elasticities between -0.15 and -0.30 during the time
period of roughly the last third of the twentieth century. In addition, the differences among the studies
point out the importance of model specification. How one deals with dynamics — by including lagged
effects, autoregressive errors, both, or neither — can have a major impact on results. In particular,
omitting such dynamic effects appears to result in over-estimates of the magnitude of the elasticities in
question. In addition, results of US studies are sensitive to how they account for the influence of the US
Corporate Average Fuel Efficiency (CAFE) standards, which went into effect in 1978.
7 To calculate these numbers we begin with the sums of estimated "baseline" elasticities for kilometers per car and
for car ownership, i.e. columns (3) and (4), as shown in the last two rows of their Table 6, p. 2117. These baseline
estimates are defined as the values predicted by their meta-analysis model with all dummy variables taking their
most common value. This results in is short-and long-run driving elasticities of-0.331 and -0.581 percent,
respectively. The model includes a dummy variable "UCA" for studies in the US, Canada, or Australia, whose most
common value is zero; so we add the sum of columns (3) and (4) for the coefficient of UCA, which is +0.165,
resulting in elasticities of -0.166 and -0.416, respectively. There is considerable uncertainty around these values, as
the standard error of the coefficient of UCA in the equation predicting kilometers per car is very large (0.480).
8 This statement is based on the sum of coefficients of the dummy variable "Dynamic" in columns (3) and (4) of
their Table 6; that sum is 0.027.
13
-------
2.2 Recent Literature
More recent literature has extended this work in several directions, especially paying close attention to the
means of identification and controls for bias due to omitted variables. Particularly relevant to this report
are studies seeking to determine whether the determinants of the rebound effect or of the price-elasticity
of gasoline have changed in the decade starting in 2000. (We refer to such changes as structural change,
meaning changes in the manner in which underlying factors influence the elasticities, as opposed to
simply changes in those factors themselves.) Because that decade is characterized by more closely spaced
price fluctuations than has been typical, observers have sometimes noted substantial changes in behavior.
Brand (2009) summarizes some simple calculations of the VMT- and price-elasticities with respect to fuel
price, based on observations before and after a sharp increase in fuel prices: specifically, by comparing
the first ten months of 2007 and the first ten months of 2008. A calculation based on U.S. national
statistics yields a short-run VMT-elasticity of-0.12. This involves no controls, and Brand points out that
VMT was trending upward at 2.9% per year over a prior 21-year period of relatively stable prices, which
to us suggests a correction to this elasticity of-0.029, bring it to approximately -0.15.9
Hughes et al. (2008) undertake a more detailed analysis, using models with some control variables, to
compare the price-elasticity of gasoline in the years 1975-80 with that in the years 2001-06. They find a
large decline in magnitude, from -0.21 to -0.08 in what appear to be their favored specification. In the
case of the later period, that specification treats fuel price as endogenous, estimating it with instrumental
variables in a standard manner that accounts for price being determined simultaneously by demand and
supply relationships. This finding suggests that the VMT elasticity declined by a similar amount, since it
is a component of the fuel-price elasticity and no one has suggested that the other main component (the
elasticity of fuel efficiency) has been demonstrated to change significantly.
Hughes et al. also test whether the price-elasticity declines in magnitude with income, as found by Small
and Van Bender (2007) and Hymel et al. (2010). They find instead an effect in the opposite direction.
Thus, they explain the decline in price elasticity as likely due to factors other than those we suggest here.
Specifically, they cite suburbanization and declining public transit service, both of which lock travelers
more firmly into automobile use, and increased fuel efficiency, which is also consistent with one of the
findings of Small and Van Bender (2007) and Hymel et al. (2010). Interestingly, Litman (2010) cites
these same factors in a heuristic argument for an opposite argument: Litman suggests these factors were
9 Brand asserts without explanation a different number, -0.21, for the VMT elasticity accounting for the trend.
Litman (2010, abstract) cites Brand and an unpublished study by Charles Komanoff as supporting an elasticity of
-0.15.
14
-------
strong during the 1970-2000 period but likely less important during the 2000's. We have not seen any
formal argument, either theoretical or empirical, for why these factors should have a major effect in either
direction.
There are some limitations to the Hughes et al. results which make them less than decisive. The limitation
to a single five-year period for each estimation reduces the precision of their estimates compared to ones
that use longer time series. Also, they do not account for a full range of dynamic effects, as we think is
especially necessary to fully capture behavior in the rapidly changing 2000-2006 period.10
Greene (2012) carries out a number of analyses similar to those of Small and Van Bender (2007), using
national rather than state data but extending the sample to year 2007. Greene confirms several results of
Small and Van Bender: in particular, he finds a similar value for the price-elasticity of VMT, finds that it
has declined over time, and finds that it declines with income.
Two recent studies make use of odometer readings from California's smog test—arguably the most
accurate available measure of VMT—to provide estimates of the elasticity of VMT with respect to either
fuel price or fuel cost per mile, both using very large samples of individual vehicles. The first, by Knittel
and Sandier (2012), takes advantage of the existence of regions in which older vehicles must take a smog
test every two years. They use test data from 1998 through 2010 and a simple log-log specification, with
control variables for demographics and whether the vehicle is a light truck, and with fixed effects
representing year, vintage, and make. Knittel and Sandier interpret the resulting elasticities as covering a
time period of two years, since that is the time interval over which VMT is measured. The estimates of
VMT elasticity with respect to fuel cost per mile vary between -0.14 and -0.26, depending on whether or
not the make is subdivided further in defining fixed effects.11
The second study using California smog test data is by Gillingham (2013). Gillingham combines the test
data for years 2005-2009 with micro observations of new-vehicle registrations in 2001-2003, in order to
observe VMT over a several-year period, typically six or seven years due to the requirement that vehicles
are tested at those ages. (There are also some observations over four to six years for vehicles that are sold
10 To be more precise, they do not include lagged endogenous variables or autocorrelation in any of what we would
consider their preferred model results, namely those using instrumental variables to control for simultaneity between
supply and demand factors.
11 These numbers are the range of coefficients of log (dollars per mile) in Table 18.3 for Models 2, 4, and 5. In other
models, the authors find heterogeneity with respect to the size of the dollars per mile variable. They explore
heterogeneity further in a more recent working paper, in which they find the VMT elasticity to vary between -0.11
and -0.18 across quartiles of fuel efficiency (Knittel and Sandier 2013, Table A.2, next to last column).
15
-------
before six years have passed.) He finds an elasticity of VMT with respect to gasoline price of -0.25, a
finding quite robust to various specification checks. Gillingham interprets this as roughly a two-year
elasticity, because it is identified mainly by a price spike between 2007 and 2009. This means of
identification is also a weakness of the study: during this same time interval the economy entered the
most significant recession since the 1930s, accompanied by drastic turmoil in housing markets including
foreclosures requiring many people to move. Despite controlling for macroeconomic conditions through a
measure of unemployment and a consumer confidence index, one must worry that gasoline prices are
correlated with unobserved factors related to tumultuous economic conditions that also influence the
amount of driving.
The two studies just described have the advantage of very large samples of individuals, permitting greater
precision in estimation as well as accounting for heterogeneity across individuals. Both studies also
assume that VMT responds to contemporaneous gasoline prices, without explicit lags. Yet the suggestive
evidence shown by Knittel and Sandier, comparing graphs of gasoline prices and VMT over time, appears
to show a one to two year lag. As already noted, our analysis of earlier studies suggests that omitting such
dynamic effects may cause the estimated elasticities to be somewhat larger in magnitude than the true
short-run (or even two-year) elasticities, especially when the observations are averaged over periods of
more than a year as is the case in both of these studies.
Molloy and Shan (2010) provide an intriguing look at one possible source of VMT response to fuel price:
changes in household location. They analyze how housing construction within small areas responded to
fuel prices over the period 1981 to 2008.12 Their model includes lags up to four years, which they found
sufficient to account for virtually all the observed responses. Their results imply that a one percent
increase in gasoline price reduces construction over the next four years by one percent, which is 0.03
percent of the total housing stock (Table 2). This result suggests one possible explanation for why Small
and Van Bender (2007) and Hymel et al. (2010) find substantial lags in the response of VMT to changes
in fuel cost.
Our conclusion from the more recent literature is that it raises the strong possibility that the rebound
effect has become larger during the 2000s. But not enough time has passed to allow definitive tests,
especially because other factors were changing so drastically during that same time period. Our response
to this situation in our own study is twofold. First, we investigate explicitly whether there is a structural
break in the determinants of VMT during the decade 2000-2009. Second, we consider some other
explanations for changes in behavior over this time: specifically, asymmetries between response to rising
and falling gasoline prices, and possible behavioral responses to intense media attention to fuel prices.
12 The areas are "permit-issuing places, which are usually small municipalities" (Molloy and Shan 2010, p. 5).
16
-------
2.3 Is the rebound effect the same as the responsiveness to price of fuel?
As noted in Section 1.2, one can challenge the assumption that people respond with the same elasticity to
fuel price and to the inverse of fuel efficiency. This assumption is prevalent both because it is
theoretically attractive, based on full consumer rationality, and because it is difficult to separate the two
effects empirically. Nevertheless, only a few studies have tested the assumption and the evidence for it is
not very solid.
Small and Van Bender (2007) and Hymel et al. (2010) both report attempts to estimate models where fuel
price PF and efficiency E are entered as separate variables. They find that the measurement of a separate
coefficient for E is very small but too imprecise to use with confidence for policy analysis. They interpret
their findings as ambiguous, but acknowledge that they are unable to prove that the rebound effect,
defined as the elasticity with respect to E, is not zero.
Greene (2012, Tables 4-5), using a long time series (1967-2007) of aggregate US data, is similarly unable
to estimate the two elasticities separately with much precision, obtaining a small, statistically
insignificant, and wrong-signed coefficient for fuel consumption per mile (the inverse of fuel efficiency).
Nevertheless, in contrast to the two papers just described, he is able to statistically reject the hypothesis
that the coefficients are equal.
Gillingham (2011, table 3.1) similarly tests whether the two coefficients can be separately estimated,
using his very large disaggregate data set. When model-specific fixed effects are not included, he is able
to separately measure the two elasticities, finding them equal to -0.19 for fuel price and -0.05 for the
inverse of fuel efficiency, both statistically significant. This again suggests they are not equal, and that the
elasticity with respect to inverse fuel efficiency may actually be considerably smaller in magnitude than
the that with respect to fuel price. In some other specifications, the elasticity with respect to fuel
efficiency is small and statistically insignificant, as in the studies just discussed.13
13 In other work, Gillingham also measures a rebound effect using a much more elaborate model which includes
both vehicle purchase and utilization. He obtains a very small value, equal to 0.06 (i.e. 6 percent) multiplied by the
fraction of people who choose a different vehicle when faced with a hypothetical new set of vehicles offered
following a feebate policy (Gillingham 2011, Section 4.4.3).
17
-------
While these studies are too few and statistically imprecise to resolve the question definitively, together
they strongly suggest that the effect of fuel efficiency is smaller than that of fuel price, and possibly very
small indeed. Therefore, by adopting the conventional assumption that their effects are equal and
opposite, this study reports rebound effects that may well be larger in magnitude than those that actually
occur when policies are implemented.
3. Data and specification for this report
The data set used here is a cross-sectional time series, with each variable measured for 50 US states, plus
District of Columbia, annually for years 1966-2009. Variables are constructed from public sources,
mainly the US Federal Highway Administration, US Census Bureau, and US Energy Information
Administration. Data sources and a fuller description, including some weaknesses of the data, are given in
Small and Van Dender (2007a,b) and Hymel, Small, and Van Dender (2010).14 In addition, we have
collected variables on media attention to gasoline prices and on volatility of gasoline prices, as described
in Section 3.4.
In the following we list the primary variables used in the statistical estimation. All the dependent
variables, and many others as well, are measured as natural logarithms. Variables starting with lower case
letters are logarithms of the variable described. All monetary variables are real (i.e. inflation-adjusted).
Dependent Variables
M: Vehicle miles traveled (VMT) divided by adult population, by state and year (logarithm: vma, for
"vehicle-miles per adult").
V: Vehicle stock divided by adult population (logarithm: vehstock).
HE: Fuel intensity, FIM, where Fis highway use of gasoline15 (logarithm\flnf).
C: Total hours of congestion delay in the state divided by adult population (logarithm: cong). See
Section 3.1 for further details
14 Greene (2012, p. 18) provides an excellent discussion of the VMT data and their weaknesses. He concludes that
the errors that may occur in the FHWA data on VMT and fuel efficiency are unlikely to cause large errors in year-
to-year changes, which are what are used in both this and Greene's study.
15 This term is used by FHWA to mean use by vehicles traveling on public roadways of all types. It excludes use by
not licensed for roadways, such as construction equipment and farm vehicles.
18
-------
Independent Variables other than CAFE
PM. Fuel cost per mile, PplE. Its logarithm is denotedpm = \n(Pp)-\n(E) = pf+fint. For convenience in
interpreting interaction variables based onpm, we have normalized it by subtracting its mean over
the sample.
Pv. Index of real new vehicle prices (1987=100) (logarithm: pv).
16
PP. Price of gasoline, deflated by consumer price index (1987=1.00) (cents per gallon). Variable pfis
its logarithm normalized by subtracting the sample mean.
Other: See Small and Van Bender (2007b), Appendix A; and Small, Hymel, and Van Bender (2010),
Appendices A and B. The first three equations include time trends to proxy for unmeasured trends
such as residential dispersion, other driving costs, lifestyle changes, and technology. As described
below, in equation (8), the set of variables denoted XM includes the variable (pm)2 and interactions
between normalized pm and other normalized variables: log real per capita income (inc), and
fraction urbanized (Urban - used only in the three-equation model) and normalized cong (used only
in the four-equation model).
Each of these variables is updated to 2009 using the same or similar source as before. However, in several
cases, the responsible agency has revised the numbers for earlier years. We have taken advantage of these
revisions in the updated data series. In order to facilitate comparisons with earlier years, we also use two
other data series in this report, making three in all:
"Original" data: those used for the earlier published reports, along with 2005-2009 values that
employ as closely as possible to the same methodology as used earlier. (Only values through
2001 or 2004 are used for estimation; the purpose of the 2005-2009 values in this data series is
only for projection.)
• "Revised" data: those incorporating the data revisions just mentioned, including two described in
Sections 3.1 and 3.2 below, viz.: (a) smoothing of 2000-2010 population, and (b) substitution of
improved congestion data. The term "revised" implies that only values through 2001 or 2004 are
used for estimation.
"Updated" data: like "Revised," but including data through 2009.
16 We include new-car prices in the second equation as indicators of the capital cost of owning a car. We exclude
used-car prices because they are likely to be endogenous; also reliable data by state are unavailable.
19
-------
Appendix A shows summary statistics for the data used in our main specification. The next three sections
explain special features of certain important variables.
3.1 Congestion variables (four-equation model)
This description is adapted from Hymel, Small, and Van Bender (2010). The measure of travel delay uses
data from the annual report on traffic congestion constructed by the Texas Transportation Institute (TTI)
— see e.g. Schrank, Lomax, and Turner (2010). TTI has estimated congestion annually for 85 large
urbanized areas, starting in 1982, using data from the Highway Performance Monitoring System database
of the US Federal Highway Administration.
The TTI measure of congestion used here is annual travel delay, which is simply the aggregate amount of
time lost due to congested driving conditions. TTI has sometimes been criticized for using this measure as
an index of the nation's congestion problem because it includes congestion that would remain in an
optimized system. Irrespective of the validity of this criticism, for our purposes the TTI measure is
appropriate because it describes the experience of the typical driver. The measure is constructed largely
from assumed speed-flow relationships, but supplemented with speed observations on specific roads. As
with other data in this study, it is probably more reliable in the more recent years.
One criticism of the TTI measures, however, has been addressed in TTFs 2010 edition of its report. The
earlier measure, used in the cited papers by Small and Van Bender and by Hymel, Small, and Van
Bender, estimated speed from observed traffic volumes using volume-delay relationships. This inevitably
introduced some error into the speeds, hence into the estimated total hours of delay. Recently, however,
TTI has collaborated with Inrix®, Inc., to make use of speed data collected via a nationwide network of
mobile devices in vehicles. These measures are available for a few most recent years, but TTI has back-
casted them to 1982 in order to permit comparisons with its earlier measure. They are also available for
an additional 26 urban areas. All these changes increase the accuracy of the data on congestion, and so are
adopted here except in the "original" data series.
For the collaborative work described earlier and for this report, congestion delays in all covered urbanized
areas are aggregated to the level of a state, then divided by the state's adult population to create a per-
adult delay measure. This procedure implicitly assumes that congestion outside these 85 urban areas is
negligible, a reasonable assumption because congestion in the US is far more costly to drivers in large
than in small urban areas. Furthermore, since data are measured at the state level, it is appropriate that the
20
-------
congestion in the larger urbanized areas is, for most states, diluted by the lack of congestion elsewhere in
our equations predicting statewide travel response. A further advantages of the use of total delay, rather
than some measure of average congestion, is that it is relatively unaffected by possible differences in how
boundaries are drawn for urban areas in different states.
3.2 State population data
Several variables specification, including all but one of the endogenous variables, make use of data on
adult or total state population as a divisor. Such data are published by the U.S. Census Bureau as midyear
population estimates; they use demographic information at the state level to update the most recent census
count, taken in years ending with zero. However, these estimates do not always match the subsequent
census count, and the Census Bureau does not update them to create a consistent series. As a result, the
published series contains many instances of implausible jumps in the years of the census count. In both of
the published papers discussed above, we applied a correction assuming that the actual census counts
taken every ten years are accurate, and that the error in estimating population between them grows
linearly over that ten-year time interval. This approach is better than using the published estimates
because it makes use of Census year data that were not available at the time the published estimates were
constructed (namely, data from the subsequent census count). See Small and Van Bender (2007b) for
details.
For this report, the same procedure was applied to the 2000-2009 data because the needed Census counts
for 2010 were available in time. This adjustment appears in the "revised" and "updated" data series, but
not in the "original" data series.
3.3 Variable to measure CAFE regulation (RE)
As in the earlier collaborative work, we define here a variable measuring the tightness of CAFE
regulation, starting in 1978, based on the difference between the mandated efficiency of new passenger
vehicles and the efficiency that would be chosen in the absence of regulation. The variable becomes zero
when CAFE is not binding or when it is not in effect. In our system, this variable helps explain the
efficiency of new passenger vehicles, while the lagged dependent variable in the fuel-intensity equation
captures the inertia due to slow turnover of the vehicle fleet. Because the CAFE standard is a national
one, this variable does not vary by state.
21
-------
The calculation proceeds in four steps, described more fully in Small and Van Bender (2007a), Appendix
B. First, we estimated a reduced-form equation explaining log fuel intensity from 1966-1977, prior to
CAFE regulations.17 Next, this equation is interpreted as a partial adjustment model, so that the
coefficient of lagged fuel intensity enables us to form a predicted desired fuel intensity for each state in
each year, including years after 1977. Third, for a given year, we averaged desired fuel intensity (in
levels, weighted by vehicle-miles traveled) across states to get a national desired average fuel intensity.
Finally, we compared the reciprocal of this desired nationwide fuel intensity to the minimum efficiency
mandated under CAFE in a given year (averaged between cars and light trucks using VMT weights, and
corrected for the difference between factory tests and real-world driving). The variable cafe is defined as
the logarithm of the ratio between the mandated and desired fuel efficiency, with that ratio truncated
below at one. Thus a value of zero for cafe means the constraint is not binding, since desired fuel
efficiency is as high as or higher than the mandated level.
The resulting variable suggests that the CAFE standard was strongly binding for the first decade of the
CAFE standards; its tightness rose dramatically until 1984 and then gradually diminished until it was
stopped being binding at all, either in 1995 (according to the 4-equation model) or 2005 (according to the
3-equation model).18 This pattern is obviously quite different from a trend starting at 1978 and from the
CAFE standard itself, both of which have been used as a variable in VMT equations by other researchers.
Implicit in the definition of the regulatory variable is a view of the CAFE regulations as exerting a force
on every state toward greater fuel efficiency of its fleet, regardless of the desired fuel efficiency in that
particular state. Our reason for adopting this view is that the CAFE standard applies to the nationwide
fleet average for each manufacturer; the manufacturer therefore has an incentive to use pricing or other
means to improve fuel efficiency everywhere, not just where it is low.
3.4 Variables on media coverage and volatility of gasoline prices
Variables measuring media coverage of gasoline price changes are based upon gas-price related articles
appearing in the New York Times newspaper. We queried the Proquest historical database for years 1960
to 2009, and tallied the annual number of article titles containing the words gasoline (or gas) and price (or
cost). This count was the basis for the variable used in the econometric analysis: it is formed from the
annual number of gas-price-related articles divided by the annual total number of articles, both in the New
York Times. This ratio ranged from roughly 1 in 4000 during the 1960s to a high of 1 in 500 in 1974. An
17 This step differs slightly between the three- and four-equation models because they contain slightly different sets
of exogenous variables. Thus, the actual values of the variable cafe differ slightly between the two models.
18 See Small and Van Dender (2007a), Fig. 1, for a graphical depiction.
22
-------
analogous count of front-page articles yielded a similar pattern of coverage. Its logarithm, after
normalization by subtracting its mean, is shown in 3.1. In our specifications, we use either the logarithm
of the ratio just defined (called Media in the statistical models) or a dummy variable (called
Media_dummy) defined as one in years where the ratio was greater than the 1996-2009 median value and
zero otherwise.19
Figure 3.1. Media coverage of gas prices
-2.0
A superior measure of media coverage would include broadcast news, other newspapers, radio, and the
Internet. But such measures are not readily available for the entire the time series from 1960-2009. So the
validity of the two variables as a measure of overall coverage of gasoline prices relies in part on the New
York Times' influence on other media outlets. Evidence of so-called "inter-media agenda setting" suggests
that other media outlets follow the New York Times when choosing their news topics. One study by Golon
(2006) found that the topics covered by the New York Times in the morning were correlated with evening
broadcast news coverage topics, with correlation coefficients between 0.14 and 0.26. In addition, it is
reasonable to assume that national topics such gas-price changes would be similar across news outlets
19This dummy variable was equal to one in years 1973-1981, 1983, 1990-1992, 1994-1997, 2000, 2004-2006, and
2008.
23
-------
even in the absence of direct influence of the New York Times.
To measure uncertainty in fuel prices, we constructed a variable whose value in year t is the logarithm of
the variance of fuel prices over the years t-4 through t. (We chose this five-year interval as the most likely
time over which new vehicle purchasers would be aware of volatility.) This measure varies across States.
For both the media and uncertainty variables, we interact the variable in question with either the fuel price
or the per-mile cost of driving.
4. Results of the Empirical Analysis
A major limitation of the previous literature is its inability to determine whether or not the rebound effect
has changed over time. Theoretical arguments, especially by Greene (1992), suggest that it should.
Basically, the argument is that the responsiveness to the fuel cost of driving will be larger if that fuel cost
is a larger proportion of the total cost of driving. If initial fuel cost is high, that increases the proportion;
but if the perceived value of time spent in the vehicle is high, either because of congestion (closely related
to urbanization) or because of a high value of time (closely related to income), that decreases the
proportion. Thus we expect the rebound effect to increase with increasing initial fuel cost, and decrease
with increasing income and urbanization. On the few occasions when such factors are even discussed,
most analysts have presumed that income is the dominant one and therefore have hypothesized a decline
in the rebound effect overtime, due to rising real incomes. Previously used data sets, however, have
covered too short a time span to test any of these arguments satisfactorily.20
With the longer time span of the data sets compiled for the earlier collaborative papers, and the even
longer data set used here (44 years), there is a much better opportunity to see such changes. We explore
them in three distinct ways. First (Section 4.1), we see whether the basic model, estimated over different
time periods but each with a constant rebound effect, yields different results. We find a substantial
20 A recent exception is two studies by Wadud, Graham and Noland (2007a, 2007b) using time-series cross sections
of individual households from the US Consumer Expenditure Survey. Cross-sectionally, they find a U-shaped
pattern of the absolute value of the price elasticity of fuel consumption, taking values of 0.35 for the lowest income
quintile, falling to 0.20 for the middle, and rising again to 0.29 for the highest (2007b, Table 2). But when they hold
other variables constant while allowing income to vary both cross-sectionally and over time (1997-2002), they
obtain a nearly steady, though small, decline of the absolute value of elasticity with income, from 0.51 in the lowest
two income quintiles to 0.40 in the highest.
24
-------
diminution in the rebound effect in the period since 1995; it's harder to say whether it has risen again
since 2000.
Second (Section 4.2), we explore income, fuel costs, urbanization, and congestion as the causes of these
changes. Each of these factors is entered in the model in such a way that the rebound effect can vary with
it rather than varying over time in an unexplained manner, and we do indeed find substantial variation in
exactly the manner predicted by theory: the rebound effect (measured as a positive number) declines with
increasing income (as well as with either urbanization or congestion), and it increases with increasing fuel
cost. By far the most important of these sources of variation is income, which has a profound effect on
projections for the rebound effect in future years. In Section 4.3, we consider explicitly how the newer
data now available (2002-2009) affect the results from the earlier published studies.
Third (Section 4.4), we consider asymmetry in the response to increases and decreases in fuel prices,
finding a much larger response to increases. We also consider the possible role of media coverage and
price volatility in explaining this asymmetry.
4.1. Variation by Time Period
This section presents the results of estimating a relatively simple version of the three-equation system
described earlier. In this version, the variable pm (the logarithm of fuel cost per mile) is simply included
in the equation explaining vma (the logarithm of vehicle-miles traveled per adult). Its coefficient, the
"structural elasticity," is the elasticity of VMT with respect to fuel cost per mile, holding vehicle fleet
constant. Accounting for how the vehicle fleet also varies with fuel cost, and how lagged adjustment
creates differences between short-run and long-run responses, we get the short- and long-run rebound
effects from equations (4), (5), and (7).
In order to see whether the rebound effect changes over time, we carry out this estimation on two
subsamples: 1966-1995 and 1996-2009. Table 4.1 shows the estimated structural elasticity SM,PM • As
described earlier, these are nearly identical (except for the minus sign) to the short-run rebound effects,
and their values come immediately from the estimated results. The table shows that the short-run rebound
effect falls by 46 percent and 72 percent, without and with consideration of congestion respectively,
between these two time periods.
25
-------
Table 4.1. Short-run structural elasticity of VMT with respect
to fuel cost per mile, estimated on subsamples
Coefficient of pm
(standard error in
parentheses)
Three-equation model
Four-equation model
1966-1995
-0.0458
(0.0037)
-0.0469
(0.0058)
1996-2009
-0.0246
-0.0071
-0.0131
(0.0075)
This result of a falling rebound effect is consistent with results noted earlier by Hughes et al. (2008) and
Greene (2012).
4.2. Variation of rebound effect with income, fuel cost, and other variables
4.2.1 Motivation
Before proceeding with the formal estimation, we motivate the approach taken here by considering what
goes into the costs of automobile travel from the traveler's point of view. Figure 1 shows three categories
of the short-run costs of driving and how they are likely to progress over coming decades, based on
compilations of Small and Verhoef (2007) for an urban commuting trip by automobile.21 The values
placed by travelers on travel time and unreliability 22are taken from statistical literature examining how
people are willing to trade off those factors against money. We have then projected fuel costs per mile
into the future, using the Energy Information Administration's projections for fuel prices and fuel
efficiency in their 2011 reference scenario (US EIA 2011). We have projected the values of travel time
and unreliability into the future by assuming that the amounts of time and unreliability are unchanged (a
conservative assumption given trends toward increased congestion) while the values of time and
21 The initial values are for 2005, taken from Small and Verhoef (2007, Table 3.3) and restated at 2007 prices.
22 In this context, unreliability refers to day-to-day variability in the travel time faced for a given type of trip. It is
typically measured by the standard deviation of travel time across days, although sometimes other measures of
dispersion (such as the difference between the 80th and 50th percentiles) are used instead. Its presence means that
people cannot accurately predict when they will arrive at their destination. There is a substantial literature, reviewed
by Small and Verhoef (2007), showing that travelers are averse to unreliability independently of their aversion to
travel time.
26
-------
unreliability increase with rising per capita real income according to an elasticity of 0.8, a
recommendation of Mackie et al. (2003) based on many studies of how value of time depends on income
(Small and Verhoef 2007, Section 2.6.5).
Figure 4.1.
Costs of Driving
100
0)
Other Costs
I Travel Time and
Unreliability
I Fuel Cost
2005
2015
2025
Thus, it appears that despite the general prognosis for rising fuel prices, the actual fuel costs are likely to
decline, due mainly to increases in fuel efficiency of automobiles; and the prominence of fuel costs in
drivers' decisions is likely to decline even more, due to increases in the value of time (and, to a lesser
extent, to amount of time spent in heavy congestion). Our econometric model can capture these
possibilities by simply specifying it in a way that allows the rebound effect to vary with income, fuel cost
per mile, and other variables that may impinge on travel time: namely, urbanization and congestion.
4.2.2 Implementation
To see how this can be done, recall from Section 1.1 that the rebound effect is a combination of
elasticities of either three or four distinct equations (known as "structural equations"). Because of the
relative sizes of these elasticities, the rebound effect is approximated by just one of them: namely SM.PM,
giving the effect of fuel cost per mile in the structural equation for vehicle-miles traveled per adult. In the
27
-------
notation used here, which uses lower-case names for variables that are expressed in natural logarithms,
that elasticity is given by equation (3), i.e. SM.PM = d(vma)ld(pm).
In the previous subsection, fuel cost per mile was described as a single variable (pm in logarithmic terms)
included in the equation for vehicle-miles traveled per adult (vma in logarithmic terms). The elasticity
was just its coefficient, which we may call fipm for convenience.23 But it is easy to specify the equation for
vma so thatpm appears not only as a single variable, but also interacted with other variables including
itself. We define four such variables: pm-inc,pm-pm=pm2,pm-Urban, andpm-cong, where inc is the
logarithm of per capita real income, Urban is the fraction of state population that is urbanized, and cong is
congestion as measured by the logarithm of total congestion delay per adult. We denote the coefficients of
these four "interacted variables" by fi\, fh., /ft, and /?4. In practice, /?4 is set to zero in the three-equation
system (since cong is not measured there), and /% is set to zero in the four-equation system (since its
estimates were small and statistically insignificant).
Then the derivative in (3) consists of four terms:
SMPM = ^ . = Ppm +A -inc + 2j32 -pm + fc • Urban + J34 • cong . (8)
d(pm)
The factor 2 in this equation is a consequence of properties of the derivative of the quadratic function
(pm)2. Inserting (8) into equations (4) and (7) for the short- and long-run rebound effects, we see that that
those rebound effects also depend on inc, pm, Urban, and cong.
In order to facilitate interpretation of coefficients, we "normalize" the values of inc, pm, Urban, and cong
by subtracting from each variable its mean value over our entire data set. This has no effect on the
coefficients except to change the constant terms in the equations; but it means that the coefficient f}pm of
the variable pm still gives the estimated elasticity s^fMatthe point where each of the interacting variables
is equal to its mean value in our data set - as can be seen by setting the three normalized variables in (8)
to zero. This is especially convenient because the short-run and long-run rebound effects are
approximately -SM.PM and -SM.PM l(\-of"), respectively, where of" is coefficient of lagged vma in the vma
equation. Thus, one can see the approximate value of the estimated short- and long-run rebound effects,
under average conditions over the sample period, just by looking at -flpm and oT.
23 This coefficient is named jBl m in Small and Van Dender (2007), eqn. (4) and Hymel et al. (2010), eqn. (9a).
28
-------
4.2.3 Estimation results: interaction variables
The models are estimated using the maximum-likelihood simultaneous-equations estimator in Eviews 5
(Quantitative Micro Software 2004). Technical details are provided in Small and Van Bender (2007a) and
Hymel, Small, and Van Bender (2010).24 The full results of estimating the three- and four-equation
models on updated data from 1966 through 2009 are presented in Appendix A; some of the most
important coefficients are summarized here in Table 4.2.25
24 For this report, however, we have replaced the multiple imputations for the missing data by a single imputation;
that is, we predict the values of the missing data only once, rather than multiple times using random draws from the
equation estimating them. For this reason, our estimates of standard errors probably understate the true standard
errors.
25 For reasons that will be explained in the next section, these models are named "Model 3.3" and "Model 4.3"
respectively. For simplicity, coefficient estimates and standard errors are shown to three decimal places in these
tables. In some later tables, they are shown to four decimal places.
29
-------
Table
4.2. Selected results of main model
with updated
Three-equation model
SH/Inrlv.1 ^ T»
Equation and
Variable
Equation for
vma:
pm
pm*inc
pm2
pm* Urban
pm*cong
me
lagged vma
Equation for
flnf.
pf+vma
cafe
laggedfint
Coefficient Coefficient
Symbol Estimate
J3pm -0.047
/?! 0.053
y& -0.012
j& 0.012
A
0.078
oT 0.835
-0.005
-0.035
of 0.904
Standard
Error
0.003
0.011
0.006
0.009
0.012
0.010
^^
0.004
0.011
0.010
data, 1966-2009
Four-equation model
SH/Inrlv.1 A T»
Coefficient Standard
Estimate Error
-0.046
0.056
-0.022
-0.003
0.083
0.825
-0.007
-0.061
0.889
0.003
0.011
0.006
0.002
0.012
0.010
0.004
0.010
0.010
Notes to Table 4.2:
vma = logarithm of vehicle-miles traveled per adult
pm = logarithm of fuel cost per mile (normalized)
inc = logarithm of income per capita
Urban = fraction of population living in urban areas
cong = logarithm of annual total congestion delay per adult
fmt = logarithm of fuel intensity, i.e. log (1/E) where E = fuel efficiency
30
-------
pf= logarithm of fuel price
cafe = variable reflecting how far the CAFE standard is above the desired fuel
efficiency based on other variables (Small and Van Dender 2007a,
Section 3.3.3)
pf+vma is log (price of fuel * vehicle-miles traveled), representing the natural
logarithm of the incremental annual fuel cost of a unit change in fuel
intensity; thus it may be interpreted as the logarithm of the "price" the
user must pay in annual operating costs, per unit of fuel intensity, for
choosing a vehicle with higher fuel intensity.
-------
Most coefficients shown in Table 4.2 easily pass the conventional test of statistical significance, having
estimates more than twice the standard deviation of those estimates. Exceptions are /?4, which indicates
how the rebound effect varies with congestion, and the coefficient of annual fuel cost (pf+vma in
logarithms) in the equation explaining fuel efficiency. The coefficients of" of lagged vma show that the
long-run effect of any variable on VMT is about l/(l-om) or roughly six times as large as the
corresponding short-run effect. Average fleet fuel efficiency responds to changes with an even longer lag,
causing the long-run effects of these variables to be \l(\-of) or roughly 9-10 times as large as the
corresponding short-run effects.
The coefficient of inc confirms the conventional expectation that vehicle-miles traveled rises with rising
income: the income-elasticity is approximately 0.1 in the short run and 0.5 in the long run. CAFE
standards are shown to be important determinants of average fleet fuel efficiency. Another way to
interpret this is that each year, fleet turnover and/or changes in driving patterns are able to close (!-
-------
Table 4.3. Estimated Rebound Effects: Model 3.3
Average values (real 2009 $) 1966-2009 2000-2009
Per capita income ($/year) $28,452 $36,805
Fuel price ($/gal) 2.06 2.18
Fuel cost per mile (cents/mi) 11.75 9.77
Calculated rebound effect: Short run Long run Short run Long run
Three-equation model (w/ congestion) 4.7% 29.5% 2.8% 17.8%
Four-equation model (w/o congestion) 4.6% 28.4% 2.5% 15.0%
The decline in the rebound effect portrayed in Table 4.3 is consistent with the overall findings of Section
4.1. But now we have an explanation for why the rebound effect is lower today than in the last decades of
the previous century. Furthermore, the measured dependence on income, fuel cost, and other variables
permits a calculation of both short-run and long-run rebound effects at any level of those variables. In
Section 5 we take advantage of this to forecast rebound effects through 2035, based on outside projections
of the relevant variables, especially incomes and fuel costs.
To our disappointment, the additional years of data do not change the fact that, as discussed in Small and
Van Bender (2007), we cannot definitively isolate the separate effect of fuel efficiency from that of fuel
price. In fact, as described there, when we look at fuel efficiency as a separate variable, it exerts no
statistically significant influence on VMT. This could be taken as evidence that the rebound effect is in
fact zero, but we adopt the more conservative approach of taking it to be the VMT elasticity with respect
to fuel price. This is especially conservative (in the sense of perhaps leading us to overstate the rebound
effect) in light of Greene's (2012) finding of similar magnitudes as we find, but in his case confirming
statistically that the effect of fuel efficiency is in fact smaller than that of fuel price.
4.2.4 Combined interaction variables and structural breaks
The fact that the rebound effect varies with income, fuel cost, and other variables explains some of the
variation in time observed earlier. But does it explain all of it? To find out, we added to Models 3.3 and
4.3 additional structural breaks at times likely to produce changes in behavior due to other factors. We
considered breaks starting at years 1982, 1995, 2003, or 2005.
33
-------
Generally, we are unable to find consistent and statistically significant structural breaks at years starting
in 1982, 1995, or 2005. However, we do find evidence of an increase in the rebound effect, even
controlling for the effects of interacting variables, starting in 2003. This is seen by simply adding a
dummy variable for years 2003-2009 to Models 3.3 and 4.3 which is done in the models labeled 3.18 and
4.13. These estimation results are shown in Table 4.4, along with the calculation of rebound effect for the
most recent five-year period (2005-2009), which falls entirely within the time after the structural break.
Table 4.4. Models with interacted coefficients and
structural break starting in 2003
Coefficients (standard
errors in parentheses)
pm
pm *Dummy_2003_ 09
pm*inc
pm2
pm*Urban
pm*cong
vma lagged
Calculated rebound
effects:
1966-2009
Short run
Long run
2005-2009
Short run
Long run
Model 3.3
-0.0466
(0.0029)
0.0528
(0.0108)
-0.0124
(0.0059)
0.0119
(0.0094)
0.8346
(0.0102)
4.7%
29.5%
3.1%
19.4%
Model 3. 18
-0.0464
(0.0029)
-0.0251
(0.0076)
0.0699
(0.0121)
-0.0113
(0.0060)
0.0078
(0.0096)
0.8279
(0.0105)
5.0%
30.9%
5.1%
31.1%
Model 4.3
-0.0461
(0.0030)
0.0561
(0.0111)
-0.0224
(0.0060)
-0.0031
(0.0022)
0.8249
(0.0105)
4.6%
28.4%
3.1%
18.6%
Model 4. 13
-0.0460
(0.0030)
-0.0237
(0.0071)
0.0721
(0.0121)
-0.0186
(0.0061)
-0.0032
(0.0022)
0.8189
(0.0107)
5.0%
29.9%
5.0%
29.8%
34
-------
The estimates show that the elasticity increases sharply in magnitude starting in 2003. In the models that
take this increase into account, the short-run rebound effect computed at average values of variables over
the entire time period is slightly larger, 5.0% instead of 4.6-4.7%. The long-run effect at this sample
average also is slightly higher, though not by much because the estimated lag parameter (coefficient of
vma lagged) is now smaller. Most important, the effect of income (coefficient of pm *inc) is measured to
be notably larger, and that of fuel cost (coefficient ofpm2) becomes slightly smaller in magnitude. These
latter changes cause the rebound effect to decline more rapidly over time. This essentially cancels the
effect of the dummy variable in calculating the rebound effect over the last five years of the sample, so
the rebound effect is virtually the same as in the entire sample. However,, the models containing a break
at 2003 will still lead to a sharp decline in the projected rebound effect for years well into the future, as
the effect of income is stronger in these models. This is true even if the conditions causing this structural
break are assumed to continue to hold; if instead they are reversed, the future rebound effect becomes
smaller still.26
Probably the best lesson to take from the measured structural break in 2003 is that the evolution of the
rebound effect is more irregular than is portrayed in the simpler models such 3.3 and 4.3, but the overall
magnitudes those models measure are not affected much by this irregularity. One can speculate that the
irregularity occurs because gasoline price started increasing rather sharply in 2003, and this was
accompanied by a great deal of publicity. Both events may have caused consumers to become more aware
of the significance of fuel prices, and perhaps also to revise their expectations about what future fuel costs
would be. These responses may in turn have caused them to begin to adjust their living patterns in ways
that involve less driving—a process that can continue gradually as they adapt family structure, household
car sharing, and residential and workplace locations. We explore these potential explanations in Sections
4.4 and 4.5.
26 Projections with Model 4.13, shown in Appendix, show the dynamic rebound effect declining from
approximately 20% in 2010 to 15% in 2020 and 10% in 2030, mainly due to trends in income, all on the assumption
that whatever factors caused the upward shift in 2003 remain in place indefinitely. If instead those factors disappear,
the projected dynamic rebound effect is about 10% in 2010, declining to 5% in 2020 to 1% in 2030.
35
-------
4.3 Effects of newer data
The results in Section 4.2 portray somewhat larger rebound effects than the studies Small and Van Bender
(2007) and Hymel, Small, and Van Bender (2010), which used these same two systems of models (the
three-equation system without congestion, and the four-equation system with congestion). As described at
the beginning of Section 3, there are two main differences between those studies and the present study:
the data have now been revised, especially data on congestion, and the data have been extended to 2009.
This subsection shows that it is mainly the latter change, the extension to 2009, which accounts for the
differences.
In Table 4.5, we present the primary coefficients of interest and the implied rebound effects in 2000-2009
for three closely related estimates, all using the model without congestion. The first (Model 3.1) is the
original estimate from the published paper, which uses data through 2001. The second (Model 3.2) is the
identical estimate, using identical years, but with the data revised as described. The third (Model 3.3) is
the same as the second except now the sample for estimation runs through 2009.
36
-------
Table 4.5. Selected results of model estimated on different versions of data:
three-equation model
Estimation period
Model estimates:
pm
pm*inc
pm2
pm* Urban
vma lagged
Calculated rebound effects at
values for:
1966-2009: short run
1969-2009: long run
2000-2009: short run
2000-2009: long run
Original as
published
(Model 3.1)
1966-2001
Coeff. Std. Err.
-0.045 0.005
0.058 0.014
-0.010 0.007
0.026 0.011
0.791 0.013
4.2%
20.5%
2.2%
10.7%
Estimated with
revised data
(Model 3.2)
1966-2001
Coeff. Std. Err.
-0.046 0.005
0.057 0.015
-0.007 0.007
0.028 0.011
0.800 0.013
4.2%
21.5%
2.4%
12.3%
Estimated with
revised & updated
data (Model 3.3)
1966-2009
Coeff. Std. Err.
-0.047 0.003
0.053 0.011
-0.012 0.006
0.012 0.009
0.835 0.010
4.7%
29.5%
2.8%
17.8%
Although the coefficients of pm look almost identical across the three models, the coefficient in each case
has the meaning of the (approximate) short-run elasticity at the sample average21 In the first two models,
the sample average covers a restricted set of years, so when the rebound effect is calculated for the longer
period 1969-2009 it is somewhat lower than that coefficient (due mainly to the effect of increasing
income). Thus, as shown, Model 3.3 produces a higher short-run rebound effect than the other two. The
difference is even greater for the long-run rebound effect because the estimate of the coefficient for the
lagged dependent variable ("vma lagged") is substantially greater; this means the multiplier
which converts from short-run to long-run elasticity, is also greater: 6.1 instead of 4.8 or 5.0.
27 This is due to the way the variables pm, inc, and Urban are normalized: namely, they are created from the
unnormalized versions by subtracting the sample mean.
37
-------
Table 4.6 carries out the same exercise for the four-equation model. In contrast to the three-equation
model, in this case, adding additional years to the estimation sample reduces the short-run rebound effect
somewhat, for either time period shown. But as before, the multiplier to convert short-run to long-run
elasticities is larger when more recent years are included. In calculating long-run elasticities, the second
effect dominates the first and they are larger when the full data set is used for estimation.
Table 4.6. Selected results of model estimated on different versions of data:
four-equation model
Estimation period
Model estimates:
pm
pm*inc
pm2
pm*cong
vma lagged
OO
Calculated rebound effects at
values for:
1966-2009: short run
1969-2009: long run
2000-2009: short run
2000-2009: long run
Original as published
(Model 4.1)
1966-2004
Coeff. Std. Err.
-0.047 0.004
0.064 0.016
-0.025 0.007
-0.012 0.003
0.795 0.013
k\
-5.0%
-25.2%
-2.8%
-14.1%
Estimated with
revised data
(Model 4.2)
1966-2004
Coeff. Std. Err.
-0.051 0.005
0.067 0.015
-0.017 0.007
-0.012 0.003
0.789 0.013
-5.0%
-25.1%
-3.2%
-16.4%
Estimated with
revised & updated
data (Model 4.3)
1966-2009
Coeff. Std. Err.
-0.046 0.003
0.056 0.011
-0.022 0.006
-0.003 0.002
0.825 0.010
-4.6%
-28.4%
-2.5%
-15.0%
38
-------
Another feature that appears in this set of models is that the data revision alone makes some difference for
estimates for the period 2000-2009, as seen by comparing Models 4.1 and 4.2. Specifically, the influence
of fuel cost on the rebound effect, as given by the coefficient ofpm2, is smaller; this results in a larger
rebound effect in Model 4.2 than in Mode 4.1. The changes due to extending the sample (Model 4.3)
mostly compensate for this.
The finding that adding data for years up to 2009 modestly increases the estimated average rebound
effect, at least in the three-equation model, is consistent with the finding of Section 4.2 that the rebound
effect seems to have taken a sharp jump to a larger value starting in 2003. This observation leads to two
further lines of investigation. In Section 4.4, we explore the possibility that rising fuel prices elicit an
inherently larger response than falling prices. In Section 4.5, we explore specific mechanisms by which
that might occur, namely through media attention and/or changes in how consumer form expectations
about future prices.
4.4 Asymmetry in response to price changes
Several researchers have found evidence that for various types of energy purchases, demand is more
responsive in the short run to price rises than to price decreases. In this section, we investigate whether
such asymmetry applies to vehicle-miles traveled as a function of gasoline price.
4.4.1 Models based on rises versus falls of fuel price
Our preferred approach is to decompose fuel price into components, following the procedure used to
decompose demand for gasoline use in Dargay and Gately (1997).2S Based on experimentation, we have
simplified the three-way decomposition used by these authors into a two-way decomposition,
measured for each state in our sample.29 In this subsection, we consider a decomposition of p/, the
logarithm of fuel price, as follows:
28 Nearly identical types of decomposition are also used for other types of energy consumption by Gately and
Huntington (2002) and Dargay (2007).
29 We do this by not distinguishing between increases that occurred before and after the maximum price observed in
the data. In addition, we do not place special importance on the year 1973 as do Dargay and Gately (1997), in part
because we already have a dummy variable for 1977 in our specification to capture special influences on travel
behavior during that year.
39
-------
Pf= P/1966 + pf_rise + pf_cut
where pf_rise is the cumulative effects of all annual increases in fuel price since the start of the sample
(here 1966); and pf_cut is the cumulative effects of all annual falls in fuel price. In other words, the value
for state / in year t is defined as:
1967
pf _cutt>t =
1967
Because we include state fixed effects in our specification (i.e., there is a constant term for every state),
all coefficient estimates depend on state-specific annual changes in a relevant variable; so in this
specification, the coefficients of pf and variables constructed from it are replaced by two separate
coefficients, one depending on upward annual changes and the other on downward annual changes.
The two decomposed variables, when added together, fully describe annual changes in variable pf.
Therefore any two of the three variables pf, pf_rise, and pf_cut can be used in the specification, with
results that are fully equivalent except for the way a t-statistic is used to test a null hypothesis. The most
convenient choice proves to be the two variables, p/and pf_cut. In that case, the effect of price rises is
given by the coefficient of pf, while the effect of price falls is given by the sum of the two coefficients.
These variables are used to replace p/in both the equation explaining the logarithm of vehicle-miles
traveled (vma) and that explaining the logarithm of fuel intensity (fint). In both cases, fuel price is also
combined with other variables, as in the specifications shown earlier (as well as in the published
articles). Specifically, the main variable giving the rebound effect was previously the logarithm of fuel
cost per mile: pm=pf+fint, to which is now added an additional variable, either pfcut or (pf_cut+fint). The
variable giving the effect of fuel price was previously given as the logarithm of annual fuel cost savings
per unit change in fuel intensity, (pf+vma), to which is now added the additional variable (pf_cut+vma).
The results for these two alternative specifications, labeled 3.20b and 3.21b, respectively, are
summarized in Table 4.7, with the base model 3.3 (no asymmetry) shown for comparison. A more
complete listing of coefficients is given in the appendix.
40
-------
Table 4.7. Selected coefficient estimates: asymmetric specifications
(a) Three-equation models
Equation and variable:
vma equation:
pm= pf+ fint
pf_cut
pf_cut+ fint
pm*inc
pm2
pm* Urban
vma lagged
fint equation:
pf+ vma
pf_cut+ vma
Equation and variable:
Model 3.3
Coeff. Std.
Error
-0.0466 0.0029
0.0528 0.0108
-0.0124 0.0059
0.0119 0.0094
0.8346 0.0102
-0.0050 0.0041
(b) Four-equation
Model 4.3
Coeff. Std.
Error
Model 3.20b
Coeff. Std.
Error
-0.0520 0.0046
0.0124 0.0093
0.0569 0.0110
-0.0159 0.0061
0.0124 0.0094
0.8256 0.0110
-0.0185 0.0057
0.0316 0.0124
models
Model 4.20b
Coeff. Std.
Error
Mode! 3.21 b
Coeff. Std. Error
-0.0639 0.0049
0.0340 0.0078
0.0577 0.0108
-0.0207 0.0061
0.0131 0.0093
0.8334 0.0105
-0.0097 0.0060
0.0143 0.0123
Model 4.21b
Coeff. Std.
Error
vma equation:
pm= pf+ fint
pf_cut
pf_cut+ fint
pm*inc
pm2
pm*cong
vma lagged
fint equation:
pf+ vma
pf_cut+ vma
-0.0461 0.0030
0.0561 0.0111
-0.0224 0.0060
-0.0031 0.0022
0.8249 0.0105
-0.0074 0.0041
-0.0498 0.0046
0.0100 0.0093
0.0548 0.0111
-0.0225 0.0061
-0.0013 0.0021
0.8221 0.0107
-0.0125 0.0055
0.0085 0.0112
-0.0629 0.0049
0.0340 0.0079
0.0573 0.0110
-0.0275 0.0061
-0.0016 0.0021
0.8305 0.0107
-0.0041 0.0058
-0.0080 0.0112
These results suggest that the rebound VMT elasticity measured previously becomes modestly stronger
(i.e. larger in absolute value) when measured only for price rises. For example, comparing base model
3.3 to asymmetric model 3.21b, that elasticity rises in magnitude, from -0.0466 to -0.0639, when
41
-------
changing from the former to the latter. Note that in these models the rebound effect itself does not
depend on whether prices are rising or falling; rather, there is a direct effect of price on VMT which is
asymmetric. In all cases, price cuts have a smaller effect on driving than price rises, a difference that is
strongly statistically significant (t-statistic 4.3 or 4.4) in two of the four specifications (3.21b, 4.21b).
Greene (2012) measures similar differences between the effects of rising and falling prices, although in
his case he cannot rule out statistically that they are identical.
The implications of the two asymmetric specifications for rebound effects are different. In Models 3.21b
and 4.21b, because variable fint (representing the logarithm of inverse of fuel efficiency) is included with
both p/and pf_cut, the rebound effect is assumed equal to the price elasticity for price cuts. For
example, in Model 3.21b that elasticity is approximately -0.0299 (the sum of coefficients of the two
variables containing fint): i.e. a short-run rebound effect of approximately 3.0%. This is less than half the
rebound effect with respect to fuel price rises in the same model, which is 6.4% (short-run structural
elasticity of -0.064). As with other responses, the short-run response would be multiplied by
approximately six in the long run.
In the alternate specification of Models 3.20b and 4.20b, by contrast, the rebound effect is assumed the
same as the price elasticity for price rises. In that case there is no definitive difference between price
rises and cuts, because the coefficient of pf_cut is small and statistically insignificant.
In these models, a change in fuel efficiency, unlike one in fuel price, is the same regardless of whether
fuel efficiency is increased or decreased. In one pair of models (those numbered 20b) this effect is the
same as that of a fuel price rise; in the other (numbered 21b) it is the same as that of a fuel price cut.
The latter seems more likely theoretically because changes in fuel efficiency are noticed less
dramatically than changes in fuel price, and because most of the changes in fuel efficiency we are
interested in are improvements, i.e. they lower the fuel cost per mile as does a price cut. Furthermore,
the asymmetry in behavior is both more notable and more precisely measured in the second
specification, as already noted. For these reasons, we prefer the two models numbered 21b.
4.4.2. Models based on rises versus falls of fuel cost
We also estimated models that base the asymmetry on the variable measuring fuel cost per mile (pm),
instead of on fuel price (pf). These models assume that people respond differently depending on
42
-------
whether their fuel cost per mile is rising or falling, regardless of whether this is due to a change in fuel
price or in fuel efficiency.
The variables are formed analogously to the previous subsection. The fuel cost per mile, pm (the price of
mileage), is decomposed into pm_rise and pm_cut. This raises a new problem because pm_rise and
pm_cut are, like pm, endogenous; but not in a simple way because their values in a given year depend
on values of pm in previous years. In the case of pm, endogeneity is accounted for as part of the three-
or four-equation model.30 A full endogenous treatment would be impossible, so we have used an
approximation instead: the variables are replaced by predicted values, pm_rise_hat and pm_cut_hat,
each of which is the value predicted by a regression of the corresponding variable (pm_rise or pm_cut)
on all the exogenous variables in the system - that is, on the same set of variables as those used as
instruments in the 3SLS estimation routine. This is basically what instrumental variables does in the case
of a simpler endogenous variable, so the result of this approximation should be reasonably accurate
although the standard errors of these variables may be inaccurately measured.
Table 4.8 shows selected results of a specification, named Model 3.23, analogous to that of Model
3.21b. The latter is shown for comparison. Each model contains three interaction variables, whose
coefficients are shown just below the second dashed line.
30 Formally, this is accomplished by entering the variable pm as the sum of two variables, pf+fint, where fint is the
logarithm of fuel intensity (see Section 3, "Dependent variables", definition of HE). Since fint is the dependent
variable of the third equation of our model system, the simultaneous estimation performed by the three-stage least
squares procedure treats it as endogenous where it enters the first equation as part of pm.
43
-------
Table 4.8. Selected coefficient estimates: asymmetry in response to fuel cost
per mile
(a) Three-equation models
Equation and variable:
vma equation:
pm= pf+ fint
pm_rise_hat
pm_rise_hat(-1)
pm_rise_hat(-2)
pf_cut + fint
pm_cut_hat
pm_cut_hat(-1)
pm_cut_hat(-2)
pm_cut_hat(-3)
pm*inc
pm2
pm* Urban
vma lagged ^^p
fint equation:
pf+ vma
pfrise
pf_cut+ vma
pf_cut
vma
Model
Coeff.
-0.0639
0.0340
0.0577
-0.0207
0.0131
0.8334
-0.0097
0.0143
3.21b
Std.
Error
0.0049
0.0078
0.0107
0.0061
0.0093
0.0104
0.0060
0.0123
Model
Coeff.
-0.0623
0.0284
0.0535
-0.0180
0.0187
0.8084
-0.0133
0.0042
0.0107
3.23
Std.
Error
0.0055
0.0093
0.0112
0.0062
0.0099
0.0122
0.0062
0.0096
0.0166
Model 3.29
Coeff.
-0.1134
0.0490
0.0210
-0.0037
-0.0485
0.0171
0.0239
0.0281
-0.0276
0.0273
0.8802
-0.0108
-0.0154
-0.0533
Std. Error
0.0153
0.0216
0.0129
0.0105
0.0141
0.0150
0.0108
0.0120
0.0068
0.0103
0.0119
0.0064
0.0097
0.0179
44
-------
(b) Four-equation models
Equation and variable:
vma equation:
pm= pf+ fint
pm_rise_hat
pm_rise_hat(-1)
pm_rise_hat(-2)
pf_cut+ fint
pm_cut_hat
pm_cut_hat(-1)
pm_cut_hat(-2)
pm_cut_hat(-3)
pm*inc
pm2
pm*cong
vma lagged
fint equation:
pf+ vma
pfrise
pf_cut+ vma
pf_cut
vma
Model
Coeff.
-0.0629
0.0340
0.0573
-0.0275
-0.0016
0.8305
-0.0041
-0.0080
4.21b
Std.
Error
0.0049
0.0079
0.0110
0.0061
0.0021
0.0107
0.0058
0.0112
Model
Coeff.
-0.0615
0.0325
0.0534
-0.0245
-0.0042
0.8229
-0.0122
0.0024
0.0210
4.23
Std.
Error
0.0054
0.0091
0.0115
0.0063
0.0022
0.0112
^
0.0063
0.0086
0.0152
Model
Coeff.
-0.0629
-0.1068
0.0426
0.0343
-0.0051
-0.0540
0.0161
0.0233
0.0394
-0.0005
-0.0046
0.8656
-0.0144
0.0267
-0.0081
4.29
Std.
Error
0.0049
0.0159
0.0229
0.0137
0.0108
0.0149
0.0163
0.0117
0.0129
0.0002
0.0029
0.0125
0.0063
0.0118
0.0153
The variable pm_cut_hat, just like the previous variable pf_cut, is an increasing function of cost per
mile.31 Given its construction, we expect a negative sign on pm (which is the direct short-run rebound
elasticity if fuel costs are rising) and also on the sum of coefficients of pm and pm_cut_hat (which gives
the direct short-run rebound elasticity if fuel costs are falling). The coefficient on pm_cut_hat itself tells
us the degree of asymmetry: it is positive if the magnitude of the elasticity is smaller for price cuts than
for price rises. Equation (3.23) shows exactly this, very similarly to (3.21b). The short-run rebound effect
is given by elasticity -0.0623 when prices are rising, and -0.0339 (=-0.0623+0.0284) when prices are
falling. The rebound effect is influenced by pm, income, and Urban much as before. The fact that the
coefficient on pm_cut_hat is statistically significant (more than twice its standard error) indicates that
we can confidently reject the hypothesis that the magnitude of response to cost rises and cuts are the
same.
31 The actual values ofpm-cut are negative by construction, but become less so aspm increases.
45
-------
Model 3.29 deals with an alternative view of how asymmetry might work. Perhaps the difference in
response between cost rises or cuts is not so much in the magnitude, but in the speed with which the
response occurs. All the models considered in this report already have an "inertia" built into them, in the
form of a lagged dependent variable which governs the speed of response to all variable changes. But in
Model 3.29, we allow also for the possibility that the speed of the response differs between rises and
cuts in cost per mile.
Model 3.29 shows a very plausible and revealing pattern. Adjustment to price rises takes place quickly;
in fact it overshoots and then retreats to a small value after two years. But the adjustment to price cuts
occurs more slowly: it is essentially zero in the year of the price change (0.0037); takes a modest value
after one year (0.0523, from the sum of the first two coefficients below the first dashed line); remains
approximately the same for a second year (sum of three coefficients); and then retreats to a value of
0.0112 (sum of all four coefficients). These response patterns are shown in Figure 4.2.
46
-------
Figure 4.2. Short-run elasticity of VMT with respect to a sustained change in fuel cost
per mile (Model 3.29)
n nfi
n D4
n n?
^ n
Response elasticity
3 i 0000
->• P O O O O
J -^ 00 O> -^ M C
0
1
Year folk
2
3
>wing change
• Rise in cost per mile DFall in cost per mile
In these models, unlike those in the previous subsection, the response to a change in fuel efficiency
depends on what's happening to overall fuel costs. If fuel price is rising more rapidly than fuel efficiency,
then the variable remains constant; therefore, these models predict that people would still respond to a
small change in fuel efficiency according to the combination of coefficients of variable pm. In other
words, they respond to any change in fuel efficiency, including an improvement, as they would to a rise
in fuel price. Thus, the effect of a CAFE tightening could differ depending on whether overall fuel prices
are generally rising or not, and if they are on how fast. The behavioral rationale is as follows: if fuel costs
are rising due to increasing fuel prices and this has heightened people's awareness, then an
improvement in fuel efficiency would have a large effect on their driving decisions because it would help
offset that fuel price rise at a time when they are highly sensitive to it. This is a debatable assumption, as
it implies a degree of rationality in calculating fuel costs that people may not have in reality. Indeed, as
noted elsewhere, our results cannot definitively show that the rebound effect differs from zero if the
responses to fuel price and fuel efficiency are estimated separately. Thus it is possible that all the
rebound results are overstated, and actually are measuring the response to changes in price rather than
in fuel efficiency. For this reason, we prefer the models of Section 4.4.1.
47
-------
Four-equation results. The same kind of model development was done for four-equation models, with
similar results as shown in Table 4.8(b) and Figure 4.3.
Figure 4.3. Short-run elasticity of VMT with respect to a sustained change in fuel cost
per mile (Model 4.29)
Ons
.uo
Onfi
.04
Ono
f?
'o
'^ n nn
t/5 U.UU
2.
« -0.02
I/)
c
O n r\A
G. -U.U'I-
I/)
0)
Q£ n r\c
-U.ub
Ono
.Uo
0-1 n
. IU
019
0
1
Year folio
n Rise in cost per mil
2 3
wing change
e D Fall in cost per mile
48
-------
4.5 Effects of media attention and expectations
Two important findings of previous sections are that the responsiveness of vehicle travel to costs sharply
increased starting around 2003, and that this responsiveness is much larger when fuel prices or costs are
rising than when they are falling. These findings naturally invite the question: why? In this section, we
consider two factors that may help explain the variations in responsiveness.
The first is variations in media attention to fuel prices and costs. Motor vehicle fuel is a moderately
important part of many people's budgets, and the price of crude oil which tends to underlie fuel price has
even more pervasive effects on consumers. As a result, there is a tendency for turmoil in gasoline or oil
markets to gain much attention in public media. Could it be that this attention is the underlying cause of
some of the variations found in this report?
The second is the uncertainty in future fuel costs. There is evidence that at most times, consumers' best
guess at future prices, i.e. their expectation, is the current price.32 However, we hypothesize that if prices
are viewed as highly uncertain, a recent change in price is more likely to be viewed as temporary.
Therefore, the responsiveness to price changes may be muted during times when recent history suggests
that prices are volatile.
Results for three promising models are presented in Table 4.9. For comparison, we also show the most
comparable base model incorporating asymmetry but not media or uncertainty: namely, Models 3.21b and
4.21b. Variables Media, Media_dummy, and log (fuelprice variance) are as explained in Section 3, all
normalized by subtracting their mean values on the entire sample. (As with other interacting variables,
this normalization is done for convenience: as a result the coefficients ofpm remains equal to the
estimated short-run structural elasticity of VMT with respect to fuel cost when interacting variables take
their mean values in the sample.)
32 Supporting evidence comes from two separate surveys, reported by Anderson et al. (2011) and Allcott (2011),
both of which asked people directly about their price expectations. Technically, the stated result can arise from
consumers assuming a "random walk" in fuel prices: starting at the current level, they are equally likely to go up or
down at each new time period. Anderson et al. (2011) find that this assumption accurately explains their answers
except in late 2008, when they expected (correctly, as it turned out) that the recent fall in prices would prove to be
temporary.
49
-------
Table 4.9. Selected coefficient estimates: asymmetry with media coverage or
fuel-price uncertainty
(a) Three-equation models
Model 3.21b
Equation and variable:
Coeff. Std.
Error
vma equation:
pm=pf+fint -0.0639 0.0049
pf_cut+ fint 0.0340 0.0078
pm*dummy_Q3Q9
pf * ( Medi a_dummy )
pf_ri se* Medi a
pm* \og(fud price variance)
pm*inc 0.0577 0.0107
pm2 -0.0207 0.0061
pm* Urban 0.0131 0.0093
vma lagged 0.8334 0.0104
fint equation:
pf+ vma
pf cut+ vma
-0.0097 0.0060
0.0143 0.0123
Model 3.35
Coeff
Model 3.37
Model 3.42
Std. Coeff. Std. Coeff
Error Error
Std.
Error
-0.0587 0.0052 -0.0641 0.0057* -0.0699 0.0069
0.0286 0.0081 0.0332 0.0083 0.0529 0.0091
-0.0216 0.0079 -0.0265 0.0078
-0.0301 0.0101 -0.0319 0.0101 -0.0316 0.0101
0.0028 0.0007
0.0583 0.0109 0.0711 0.0126 0.0779 0.0124
-0.0053 0.0075 -0.0064 0.0075 -0.0126 0.0070
0.0118 0.0094 0.0100 0.0097 0.0091 0.0095
0.8325 0.0106 0.8276 0.0109 0.8321 0.0108
-0.0124 0.0059 -0.0104 0.0058 -0.0079 0.0058
0.0220 0.0120 0.0129 0.0118 0.0031 0.0115
Model 3.45
Coeff.
-0.0666
0.0210
-0.0347
-0.2680
0.0081
0.0807
-0.0302
0.0118
0.8247
-0.0033
-0.0225
Std.
Error
0.0053
0.0083
0.0084
0.0544
0.0024
0.0136
0.0081
0.0106
0.0117
0.0058
0.0114
(b) Four-equation models
Equation and variable:
vma equation:
pm= pf+ fint
pf_cut+ fint
pm* dummy _0309
pf * ( Media_dummy )
pf_rise*Media
Model
Coeff.
-0.0629
0.0340
4.21b
Std.
Error
0.0049
0.0079
Model
Coeff.
-0
0
0
.0638
.0352
.0061
4.35
Std.
Error
0.0050
0.0080
0.0058
Model
Coeff.
-0.0729
0.0420
-0.0359
0.0071
4.37
Std.
Error
0.0054
0.0081
0.0071
0.0058
pm* \og(fud price variance)
pm*inc
pm2
pm* Urban
vma lagged
fint equation:
pf+ vma
pf cut+ vma
0.0573
-0.0275
-0.0016
0.8305
-0.0041
-0.0080
0.0110
0.0061
0.0021
0.0107
0.0058
0.0112
0
-0
-0
0
-0
-0
.0575
.0296
.0025
.8314
.0060
.0031
0.0110
0.0065
0.0021
0.0106
0.0057
0.0110
0.0825
-0.0263
-0.0028
0.8314
-0.0059
-0.0022
0.0122
0.0066
0.0021
0.0106
0.0057
0.0110
Model
Coeff.
-0.0706
0.0506
-0.0308
-0.0080
-0.0100
0.0944
0.0037
-0.0044
0.8275
-0.0049
-0.0018
4.42
Std.
Error
0.0054
0.0083
0.0072
0.0063
0.0019
0.0124
0.0085
0.0021
0.0109
0.0057
0.0110
Model
Coeff.
-0.0719
0.0626
-0.0321
-0.3117
-0.0044
0.0905
-0.0114
-0.0057
0.8423
-0.0035
-0.0129
4.45
Std.
Error
0.0053
0.0085
0.0072
0.0490
0.0019
0.0124
0.0074
0.0021
0.0112
0.0057
0.0111
50
-------
The media variable is specified to influence the response to fuel price but not to fuel efficiency, because
the variable involves news about fuel price. Therefore, including this variable does not affect the rebound
effect except insofar as it changes coefficients ofpm and its interactions. The uncertainty variable, by
contrast, represents a consumer's own experience with variation in fuel costs, and therefore is specified so
as to influence both responses (i.e., it is interacted withpm rather than/?/).
Consider first the four-equation models. The last of these models (4.45) suggests that both media
coverage and fuel-price volatility, taken together, have significant effects in increasing the magnitude of
the elasticity of VMT with respect to fuel price, just as we hypothesized. The effect of Media is strongest
when it is entered as a continuous rather than a dummy variable and when it is interacted with price rises
(pf_rise). The effect of these additional variables on coefficients involving pm is minimal except for one:
the coefficient ofpm2 becomes smaller when fuel price volatility is included. This could mean that the
previously observed tendency of the price elasticity (and rebound effect) to increase with fuel price is
explained in part by correlation between high prices and media coverage. But the results are not
consistent enough to draw a firm conclusion on this point.
In the three-equation models, the media variables alone seem powerful (Models 3.35 and 3.37), but when
fuel price variability is included (Model 3.45), its coefficient has an unexpected sign. We do not have a
good explanation for this. Generally, the sensitivity shown in these models to the precise form in which
variables are entered into the equation is an undesirable property, and probably indicates that we have
reached the limits of our ability to discern these fine-grained effects using this data set.
Comparing Model 3.35 or 4.35 with the higher-numbered models, which all contain the variable "dummy
0309", we see there continues to be a structural break toward a larger rebound effect in years 2003-2009,
even with these other variables are accounted for. The amount of this break (an increase in the short-run
rebound effect of roughly 2.0 to 3.5 percentage points) is about the same size as found previously, in
Table 4.4 (Models 3.18 and 4.13). Therefore, it seems these new variables have not captured whatever
factors changed the responsiveness to price and fuel efficiency starting in 2003. Thus, further research is
needed if one wishes to understand the reason for this change, and in particular the likelihood that it will
persist into the future.
Taking into account explanatory power, consistency across three- and four-equation models, and
consistency with theory, our preferred models remain those that omit media and volatility variables:
namely, Models 3.21b and 4.21b. While the exploration of media and volatility elicit considerable
evidence that one or both of these factors helps explain.
51
-------
5. Implications of the Empirical Analysis: Projections to 2035
By distinguishing the causes of the observed decline in the rebound effect, we are in a position to consider
how the rebound effect is likely to change in the future. By inserting projected values for real per capita
income, real fuel costs of driving, urbanization, and congestion into our model, we obtain a projection for
the rebound effect. Of course, like any projection, the farther into the future we project, the uncertain are
the values of these variables. In addition, in both cases projections show one or both variables moving
outside the range in which they were observed in our sample; as a result, statistical uncertainty in the
estimated model can magnify the uncertainty in the projected values.
The models estimated here imply the rebound effect is a linear function of the logarithms of per capita
income and fuel cost per mile. This is probably a good approximation within limited ranges of those
variables, but for extreme values the linear function becomes less satisfactory. In particular, since rising
income lowers the rebound effect, linearity implies that the rebound effect could become negative at high
enough incomes. This is unrealistic and so to avoid it, we truncate the rebound effect for any given state
and year at zero. As a result, the aggregate rebound approaches zero only gradually as incomes rise,
because an increasing number of states hit this limit. In the base projections here, the number of states
with zero rebound effects rises from one in 2008 to either five or seven in 2035, depending on whether the
three- or four-equation model is used.
The first two of the variables needed for projections — per capita income and fuel cost per mile — are
projected in the 2011 Annual Energy Outlook published by the U.S. Energy Information Administration
(US EIA 2011). WWe refer to these input projections as AEO2011. The AEO's projections are national,
whereas the rebound effects calculated here vary by state. Thus for each state, we use the average of 2008
and 2009 as a starting value, and then change the two variables (per capita income and fuel cost per mile)
by the same proportion that the national projection changes from those same two starting years.
It is worth noting that these projected values do not take into account any change that might occur from
the regulation itself. Thus, for example, the rebound effect in 2025 is based on fuel efficiency projections
from AEO that do not incorporate the impact of tightened efficiency regulations in years 2017-2024.
Because the effect of fuel costs is to raise the rebound effect, this means the projections here slightly
overestimate the rebound effect compared to one that tracks the cumulative effects of the regulations on
average fuel economy in each year.
For urbanization, we extrapolate from the changes observed in national averages within the data set from
1999 to 2009. Specifically, the proportion of non-urban population and the number of hours of delay are
52
-------
each assumed to change at the same annual rate as observed over that decade. That annual rate is -0.4%,
resulting in average urbanization (fraction of population in urban areas) rising from 74.3% in 2010 to
76.7% in 2035.
For congestion, we use a projection by the U.S. Federal Highway Administration that under current
funding for infrastructure, congestion will increase at an average annual rate of 1.26 percent (US FFiWA
2011) between 2006 and 2026.33 Applying this same rate to the entire projection period implies that
annual hours of delay per person, averaged over states, rises over from 8.6 to 11.9. (Congestion affects the
projections only for the four-equation model.)
The projection methodology computes the short-run and long-run rebound effects, based on the formulas
already given using values of the "interaction variables" (per capita income, fleet-average fuel efficiency,
urbanization, and congestion) as just described for every state and every year from 2010-2035. The same
methodology is used to "back-cast" the values of rebound effect that our model implies occurred during
years 2000-2009, using the actual values of interacting variables.
For a given year, the short-run and long-run rebound effects refer to projected changes in VMT that
would occur from a permanent change in the cost per mile beginning in that year, relative to its baseline
projected value, if all the relevant interaction variables (income, fuel price, urbanization, and congestion)
were to remain constant in time following this change. The short-run rebound describes the change in
VMT during the year in question, whereas the long-run rebound describes the change in VMT in the
distant future caused by this same permanent change. The long-run rebound is larger in magnitude than
the short-run rebound because people adjust slowly to a change, as demonstrated by the coefficients on
the lagged dependent variables in the equations. (Especially, the coefficient of approximately 0.8 on
lagged vehicle-miles per adult indicates that about 80% of the choice about travel in a given year is
determined by "inertia," i.e. by travel the previous year, whereas only 20% is given by the new "target"
travel resulting from new conditions.) These projections provide the best comparison with other values
for the "rebound effect" estimated in the literature, which are based on the same hypothetical experiment.
For purposes of regulatory analysis, however, a more relevant measure is how much the path of VMT is
shifted by a permanent change in cost per mile in a given year. This measure takes the interacting
variables to be changing over time, as in fact they are projected to be, rather than being held constant. It
tracks how the VMT changes in the years following a regulatory change from two sources
simultaneously: (a) the transition from short to long run, as already described; and (b) the changes in
33 US FHWA (2008), Exhibit 7-9, column headed "Percent Change in Delay on Roads Modeled in HERS
Congestion Delay per VMT, Funding Mechanism: Fixed Rate User Charges."
53
-------
variables that influence the rebound effect. This is what was defined earlier as the dynamic rebound effect.
(See Section 1 and Appendix C for details of its calculation.)
5.1 Results: Projections using models without media or uncertainty
Tables 5.1 through 5.3 summarize the results of projecting Models 3.3 and 3.21b, our preferred symmetric
and asymmetric models and for the corresponding four-equation models. Year by year details of these
projections are given in the appendix. Table 5.1 compares the two models, both using the AEO 2011
"Reference Case," while Tables 5.2 and 5.3 give results for each model if input variables are instead taken
from the AEO 2011 "High Oil Price" and Low Oil Price" cases. Figures 5.1 through 5.3 present some of
the same information—specifically, for the dynamic rebound effect—graphically. Figure 5.1 also shows,
for comparison, the results of Models 3.23 and 4.23 with asymmetry based on fuel cost; this graph
illustrates one of the problems with using such a model to project rebound effects, which is that the effect
can fluctuate wildly from year to year due to the fact that projected cost per mile is relatively flat but with
small variations up or down in various years.
54
-------
Table 5.1
Projection Results: Rebound Effect (expressed as positive percentage), comparing symmetric and
asymmetric models
(a) Three-equation models: Model 3.3 (symmetric) and 3.21b (asymmetric)
\
2000-2009 2010
Model 3.3 (symmetric)
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Model 3.21 b (with asymmetry
based on fuel price)
Short Run Rebound
Dynamic Rebound
Long Run Rebound
(b) Four-equation
Model 4.3 (symmetric)
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Model 4.21 b (with asymmetry
based on fuel price)
Short Run Rebound
Dynamic Rebound
Long Run Rebound
2.8% 2
NA 11
17.8% 17
0.7% 1
NA 4
4.2% 5
models: Model
Historical
2000-2009
2.5%
NA
15.0%
0.5%
NA
2.4%
.8%
.4%
.6%
.0%
.2%
.8%
4.3
2017
2.4%
8.8%
15.4%
0.8%
2.3%
4.5%
(symmetric)
2010 2017
3.
13.
18.
1.
5.
6.
0% 2.9%
2« 10.7%
2% 17.2%
1% 1.0%
4% 3.3%
4% 5.9%
2025
1.6%
5.3%
10.2%
0.2%
0.2%
1.0%
and 4.21b
2025
2.0%
6.6%
11.6%
0.3%
0.3%
1.4%
2030
1 .2%
3.8%
7.2%
0.0%
0.0%
0.2%
2035
0.8%
3.2%
4.8%
0.0%
0.0%
0.0%
Regulated
average
2017-2025
2.0%
6.9%
12.9%
0.4%
1.0%
2.7%
(asymmetric)
2030
1.5%
4.7%
8.3%
0.1%
0.0%
0.2%
2035
1.0%
3.9%
5.6%
0.0%
0.0%
0.0%
Regulated
average
2017-2025
2.4%
8.6%
14.5%
0.6%
1.5%
3.5%
55
-------
Table 5.2
Projection Results: Rebound Effect (expressed as positive percentage) with symmetric models,
comparing different oil price cases
(a) Three-equation symmetric model (Model 3.3)
Reference Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Historical
2000-2009
2.8%
NA
17.8%
2.8%
NA
17.8%
2.8%
NA
17.8%
2010
2.8%
11.4%
17.6%
2.8%
14.4%
17.6%
2.8%
7.8%
17.6%
2017
2.9%
11.1%
18.1%
3.3%
14.5%
20.8%
2.4%
7.1%
14.8%
2025
2.8%
10.8%
17.7%
3.5%
14.4%
22.1%
2.2%
6.5%
13.8%
2030
2.8%
10.5%
17.9%
3.6%
14.1%
22.6%
2.1%
6.0%
12.9%
2035
2.8%
10.1%
17.4%
3.5%
13.7%
22.2%
1.9%
5.5%
11.8%
Regulated
average
2017-2025
2.0%
6.9%
12.9%
2.9%
10.6%
18.3%
0.9%
2.3%
5.8%
56
-------
(b) Four-equation symmetric model (Model 4.3)
Selected Projection Results: Rebound Effect (expressed as positive percentage)
Four-equation model estimated on 1966-2009 revised & updated data (Model 4.3)
Reference Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Historical
2000-2009
2.5%
NA
15.0%
2.5%
NA
15.0%
2.5%
NA
15.0%
2010
3.0%
13.2%
18.2%
3.0%
18.6%
18.1%
3.0%
6.9%
18.1%
Projei
2017
2.9%
10.7%
17.2%
4.4%
17.4%
26.5%
1 .0%
2.4%
5.8%
2025
2.0%
6.6%
1 1 .6%
3.5%
13.0%
21.1%
0.1%
0.1%
0.4%
2030
1 .5%
4.7%
8.3%
2.9%
1 1 .0%
17.5%
0.0%
0.0%
0.1%
2035
1 .0%
3.9%
5.6%
2.5%
9.9%
14.5%
0.0%
0.0%
0.0%
Regulated
average
2017-2025
2.4%
8.6%
14.5%
4.0%
15.1%
24.0%
0.5%
0.8%
2.8%
57
-------
Table 5.3
Projection Results: Rebound Effect (expressed as positive percentage) with asymmetric models,
comparing different oil price cases
(a) Three-equation asymmetric model (Model 3.21b)
Reference Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Reference Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Historical
2000-2009
0.7%
NA
4.2%
0.7%
NA
4.2%
0.7%
NA
4.2%
(b) Four-equation
Historical
2000-2009
0.5%
NA
2.4%
0.5%
NA
2.4%
0.5%
DrrMar^ari
2010
1.0%
4.2%
5.8%
0.9%
8.5%
2017
0.8%
2.3%
4.5%
2.1%
7.5%
5.7%^ 12.7%
1.0%
2.0%
5.7%
asymmetric
2010
1.1%
5.4%
6.4%
1.1%
11.8%
6.3%
1.1%
NA 2.5%
2.4%
6.3%
0.0%
0.0%
0.1%
model
Prn
2017
1.0%
3.3%
5.9%
2.8%
11.3%
17.4%
0.0%
0.0%
0.0%
2025
0.2%
0.2%
1.0%
1.2%
3.4%
7.2%
0.0%
0.0%
0.0%
(Model 4
ippfprl
2025
0.3%
0.3%
1.4%
1.9%
6.5%
11.6%
0.0%
0.0%
0.0%
2030
0.0%
0.0%
0.2%
0.7%
1 .7%
3.9%
0.0%
0.0%
0.0%
.21b)
2030
0.1%
0.0%
0.2%
1.3%
4.3%
7.7%
0.0%
0.0%
0.0%
2035
0.0%
0.0%
0.0%
0.3%
1.3%
1.9%
0.0%
0.0%
0.0%
2035
0.0%
0.0%
0.0%
0.8%
3.1%
4.5%
0.0%
0.0%
0.0%
Regulated
average
2017-2025
0.4%
1.0%
2.7%
1.6%
5.3%
10.0%
0.0%
0.0%
0.0%
Regulated
average
2017-2025
0.6%
1.5%
3.5%
2.4%
8.8%
14.7%
0.0%
0.0%
0.0%
58
-------
59
-------
25.0%
5.0%
0.0%
Figure 5.1
Selected projection results: Symmetric and two asymmetric models
(a) Three-equation models
Dynamic rebound effects: Comparison of three-
equation models (Reference oil price case)
2000
2010
2020
2030
-Base model (3.3)
•Asymmetry based on
price (Model 3.21 b)
-Asymmetry based on
cost (Model 3.23)
(b) Four-equation models
25.0%
Dynamic rebound effects: Four-equation models
(Reference oil price case)
-Base model (4.3)
-Asymmetry based on
price (Model 4.21b)
-Asymmetry based on
cost (Model 4.23)
2000 2005 2010 2015 2020 2025 2030 2035
60
-------
Figure 5.2
Selected Projection Results: Symmetric Models
(a) Three-equation model
Dynamic rebound effects: Three-equation base model (3.3)
15.0%
10.0%
5.0%
0.0% 4
• High oil price case
Reference case
•Lowoil price case
2000 2005 2010 2015 2020 2025 2030 2035
(b) Four-equation models
Dynamic rebound effects: Four-equation base model (4.3)
-High Oil Price Case
- Reference case
• Low Oil Price Case
0.0%
2000 2005 2010 2015 2020 2025 2030 2035
61
-------
Figure 5.3
Selected projection results: Preferred asymmetric models
(a) Three-equation model
Dynamic rebound effects: Three-equation model with
asymmetry based on price (3.21 b)
15.0% -i
10.0%
5.0%
0.0%
-High oil price case
- Reference case
-Lowoil price case
2000 2005 2010 2015 2020 2025 2030 2035
(b) Four-equation models
Dynamic rebound effects: Four-equation model with
asymmetry based on price (4.21 b)
15.0%
10.0%
5.0%
0.0% -^
-High Oil Price Case
•Reference case
•Low Oil Price Case
2000 2005 2010 2015 2020 2025 2030 2035
The projections from asymmetric models show more fluctuations than those from symmetric models,
because the sharp break between years of rising and falling fuel costs causes jumps in the short-run and
long-run rebound effects. This occurs each year when the change in fuel price switches sign, as happened
in 2009 (becoming negative) and 2010 (becoming positive again). In the "low oil price" projections, it
happens again in 2011 as the price spike in 2010 is projected to be reversed, and then again in 2017 when
62
-------
the 2011-2016 downward trend changes to a steady though very gradual increase. These fluctuations are
mainly seen in the short-run and long-run rebound effects, as illustrated in Figure 5.4.
Figure 5.4
Projection results for preferred models with asymmetry
(a) Three-equation model
15.0%
10.0%
5.0%
0.0%
Projections of Rebound Effect
Model 3.21b: Three-equation model estimated on 1966-2009 data
2000 2005 2010 2015 2020 2025 2030 2035
(2000-09 are estimates; 2010-35 are projections)
(b) Four-equation model
15.0%
10.0%
5.0%
0.0%
Projections of Rebound Effect
Model 4.21b: Four-equation model estimated on 1966-2009 data
2000 2005 2010 2015 2020 2025 2030 2035
(2000-09 are estimates; 2010-35 are projections)
63
-------
The dynamic rebound effect does not have such large jumps, because it effectively averages the responses
over the lifetime of a vehicle purchased during the year in question. Thus, if over the next 15 years the
impact on VMT is sometimes large and sometimes small, this is diluted first by the "inertia" in consumer
response, which is tracked in the dynamic rebound calculation, and also by the summation over years in
mileage driven. For this reason, it can be larger than the long-run rebound effect in years when fuel costs
have just fallen, because the long-run rebound effect assumes that all variables, including the indicator for
falling prices, will remain unchanged over the life of the vehicle.
64
-------
The projection results thus far are summarized in Table 5.4, focusing on the regulated average value of
the rebound effect (i.e., average over years 2017-2025). The first two panels present dynamic rebound
effects, the third presents long-run rebound effects.
Table 5.4
Selected summary measures
(a) Dynamic rebound effect: symmetric models
(Average over years 2017-2025)
High Oil Price Case
Reference Case
Low Oil Price Case
Three-equation
model (3.3)
10.6%
6.9%
2.3%
Four-equation
model (4.3)
15.1%
8.6%
0.8%
Average
12.8%
7.8%
1 .5%
Note: Rebound effect is defined as minus the elasticity of VMT with respect to fuel cost
per mile, expressed as positive percentage). Dynamic rebound effect refers to total miles
driven by a vehicle over its life. "Regulated average" over 2017-2015 is weighted by
projected sales of all light duty vehicles.
(b) Dynamic rebound effect: asymmetric models
(Average over years 2017-2025)
High Oil Price Case
Reference Case
Low Oil Price Case
Three-equation
model (3.21 b)
5.3%
1 .0%
0.0%
Four-equation
model (4.21 b)
8.8%
1 .5%
0.0%
Average
7.0%
1.3%
0.0%
(c) Long run rebound effect: asymmetric models
(Average over years 2017-2025)
Three-equation Four-equation
model (3.21 b) model (4.21 b) Average
High Oil Price Case 10.0% 14.7% 12.4%
Reference Case 2.7% 3.5% 3.1%
Low Oil Price Case 0.0% 0.0% 0.0%
Note: Unlike the dynamic rebound effect, which accounts for changes in fuel
prices after a car is purchased, the long-run rebound effect forecasts the result if
fuel prices remained the same throughout the life of the vehicle. This is why it can
sometimes be smaller than the dynamic rebound effect.
Recently, a Reference Case projection has become available using the 2012 version of the Annual Energy
Outlook (AEO2012). In order to see whether this substantially affects the projections of the rebound
effects, a comparison is presented in Figure 5.5. Using our base models (Models 3.3 and 4.3), the
projected dynamic rebound effects are about two percentage points larger using AEO2012, because of
65
-------
its higher energy prices. In the case of the asymmetric models, however, this differential disappears by
the end of the projection period because the rebound effect falls essentially to zero due to the strong
effect of variable pm_cut in reducing the rebound effect.
Figure 5.5
Comparisons of projections using AEO2011 and AEO2012
(a) Three-equation models
Dynamic rebound effects: Three-equation base
model (3.3)
15.0%
10.0%
5.0% -
0.0%
-Reference case:
AE02012
-Reference case:
AEO2011
2000
2010
2020
2030
66
-------
Dynamic rebound effects: Three-equation model
with
asymmetry based on price (3.21 b)
15.0%
10.0%
5.0%
0.0% -
2000
2010
2020
2030
• Reference case:
AEO2012
• Reference case:
AE02011
(b) Four-equation models
Dynamic rebound effects: Four-equation base model
(4.3)
on no/ _,
zU.Uvo
4 c no/
TO. 1)70
1
4 n no/
ID. 1)70
5 no/
.1)70
Ono/
^^^ T^A
"^£>~
^"•••••B
— *— Reference case:
AE02012
-•— Reference case:
AE02011
ii
.1)70 I I
2000 2010 2020 2030
67
-------
Dynamic rebound effects: Four-equation model with
asymmetry based on price (4.21 b)
20.0%
0.0%
2000
Reference case:
AE02012
Reference case:
AE02011
2010
2020
2030
5.2 Results: Projections using models with media variable
Table 5.5 and Figure 5.6 show the results of projecting Model 3.35. Because the media variable is
specified so that it affects the response of VMT to price but not to fuel efficiency, its only impact on the
projections is the way it changes other coefficients. As it happens, the only notable effect it has is to
lessen the impact of future changes in fuel cost per mile, whose effect on projections is not very large
anyhow except in the "high oil price" case. Thus, the projections for the AEO reference case are little
different from those with the corresponding model without media variable (Model 3.21b): they are
slightly lower during the early part of the regulatory period, leading to a "regulated average" dynamic
rebound effect of 0.7%.
Table 5.5
Projection results for model with media coverage variable:
Three-equation model
68
-------
Model 3.21 b
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Model 3.35
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Historical
2000-2009
0.7%
NA
4.2%
0.7%
NA
4.2%
2010
1 .0%
4.2%
5.8%
1.1%
3.3%
6.4%
2017
0.8%
2.3%
4.5%
0.6%
1 .4%
3.7%
2025
0.2%
0.2%
1 .0%
0.2%
0.2%
0.9%
2030
0.0%
0.0%
0.2%
0.0%
0.0%
0.2%
2035
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Regulated
average
2017-2025
0.4%
1.0%
2.7%
0.4%
0.7%
2.2%
Figure 5.6
Projection results for model with media coverage variable:
Three-equation model
20.0% -i
15.0%
10.0%
5.0% i
^-A^A-^A
****^ ^*v*±*-±^
*****^**«. **w
~^***"-«J ^^T^.^
—•—Short run
-•—Dynamic
—A— Long run
0.0% ' 1 1 • • i i» ii immmmmmmmmuu
2000 2005 2010 2015 2020 2025 2030 2035
(2000-09 are estimates; 2010-35 are projections)
69
-------
In the four-equation model, the media variable has virtually no effect on results, so the projections would
be essentially the same as in Model 4.21b.
We do not project the rebound effect using the models containing price volatility, because we do not have
an obvious way to forecast volatility. Nor is any significant volatility included in the AEO forecasts.
Nevertheless, one can expect the future to contain some periods of stability and some of volatility,
causing the rebound effect to fluctuate in some unknown manner around the trends we have projected.
6. Conclusions
The research reported here confirms the findings of previous studies that the long-run rebound effect,
measured over a period of several decades extending back to 1966, is 28-30% (Table 4.3). We also find a
short-run (one-year) rebound effect of 4.6-4.7%, which is harder to compare to previous studies because
previous work contains so much variation depending on the treatment of dynamics and of CAFE
regulations.
This research also provides strong evidence that the rebound effect became substantially lower in more
recent years, and that probably this was due to a combination of higher real incomes, lower real fuel costs,
and higher urbanization. Because time spent in travel rises with urbanization and its attendant congestion,
and the value of that time rises with incomes, all three of these differences tend to make fuel costs a
smaller portion of the total cost of traveling. Thus it is not surprising that people would become less
sensitive, on a percentage basis, to changes in those fuel costs. Our base model implies that the long-run
rebound effect was 15-18% on average over the years 2000-2009 (Table 4.3). Projections suggest that the
effect of income is very strong, reducing the long-run rebound effect from about 11-14% in 2010 to 3-5%
in 2035, according to the base model (Figure 5.1)
There is strong evidence of asymmetry in responsiveness to price increases and decreases. This makes
interpretation of the rebound effect somewhat more difficult, because it accentuates the unresolved
question as to whether travelers respond to a change in fuel efficiency in the same way as to a change in
fuel price. Different assumptions lead to quite different implications for detailed projections. Still, the
overall tendency of the results is to show that the rebound effect is likely to be moderate, and to decline
with income. Furthermore, accounting for asymmetry greatly reduces the rebound effect when it is
identified, as seems plausible, with the observed response to fuel price declines. For example, using the
AEO 2011 reference case, the projected dynamic rebound effect averaged over the years 2017-2025 and
70
-------
averaged between the three-equation and four-equation models is 7.8% using a symmetric model, but
only 1.3 percent using the preferred asymmetric model (Table 5.4).
There is weaker evidence that media coverage, and perhaps recent fuel-price volatility, also affect
travelers' responsiveness to changes in fuel cost. This evidence tends to confirm expectations that such
variables are important, but it is not conclusive at this point. Furthermore, it does not undermine the most
important finding of this and earlier work, which is that the rebound effect will decline over time as
incomes rise.
71
-------
References
Allcott, Hunt (2011), "Consumers' Perceptions and Misperceptions of Energy Costs," American
Economics Review: Papers and Proceedings, 101(3): 98-104.
Anderson, Soren T., Ryan Kellogg, James M. Sallee, and Richard T. Curtin (2011), "Forecasting
Gasoline Prices Using Consumer Surveys," American Economic Review: Papers and
Proceedings, 101 (3 ), 110-114.
Barla, Phillippe, Bernard Lamonde, Luis F. Miranda-Moreno, and Nathalie Boucher (2009),
"Traveled distance, stock and fuel efficiency of private vehicles in Canada: Price elasticities and
rebound effect," Transportation, 36: 389-402.
Brand, Dan (2009), "Impacts of Higher Fuel Costs/' in: Innovations for Tomorrow's Transportation, U.S.
Federal Highway Administration (May), www.fhwa.dot.gov/policv/otps/innovation/issuel/impacts.htm
(accessed Feb. 21, 2013).
Brons, Martijn, Peter Nijkamp, Eric Pels, and Piet Rietveld (2008), "A meta-analysis of the price
elasticity of gasoline demand: A SUR approach," Energy Economics, 30: 2105-2122.
Dargay, J., 2007. The effect of prices and income on car travel in the UK. Transportation
Research Part A: Policy and Practice, 41(10), 949-960.Davis, Stacy C., Susan W.
Dargay, Joyce M., and Dermot Gately, 1997, "The Demand for Transportation Fuels: Imperfect
Price-reversibility?" Transportation Research - Part B, 3 l(l):71-82.
Dargay, Joyce M., Dermot Gately, and Martin Sommer (2007), "Vehicle Ownership and Income
Growth, Worldwide: 1960-2030. Energy Journal, 28(4): 143-170.
Diegel, and Robert G. Boundy (2011), Transportation Energy Data Book, 30th edition, Oak
Ridge National Laboratory, June, http://cta.ornl.gov/data/index. shtml
Gately, Dermot, and Hillard G. Huntington, 2002. "The Asymmetric Effects of Changes in Price
and Income on Energy and Oil Demand" Energy Journal, 23(1), pp. 19-55.
Gillingham, Kenneth (2011), The Consumer Response to Gasoline Price Changes: Empirical
Evidence and Policy Implications. Ph.D. Dissertation, Dept. of Management Science and
Engineering, Stanford University, June, http://purl. Stanford.edu/wz808zn3318
Gillingham, Kenneth (2013), "Identifying the Elasticity of Driving: Evidence from a Gasoline
Price Shock in California," working paper, Yale University,
February.http://www.yale.edu/gillingham/Gillingham_IdentifyingElasti cityDriving.pdf
Golan, Guy (2006), Inter-Media Agenda Setting and Global News Coverage: Assessing the
influence of the New York Times on three network television evening news programs,"
Journalism Studies, 7(2), 323-333.
Goldberg, P. K. (1998). The Effects of the Corporate Average Fuel Efficiency Standards,
Journal of Industrial Economics, 46 (4), 1-33.
72
-------
Goodwin, P., J. Dargay, and M. Hanly (2004). "Elasticities of Road Traffic and Fuel
Consumption with Respect to Price and Income: A Review," Transport Reviews, 24 (3), 275-
292.
Graham, D.J., Glaister, S., 2004, Road traffic demand elasticity estimates: a review. Transport Reviews,
24(3), 261-274.
Greene, D.L. (1992). "Vehicle Use and Fuel Economy: How Big is the Rebound Effect?" Energy Journal,
13(1), 117-143.
Greene, David L. (2012), "Rebound 2007: Analysis of U.S. Light-Duty Vehicle Travel Statistics," Energy
Policy, 41: 14-28.
Greene D.L., J.R. Kahn and R.C. Gibson (1999). "Fuel Economy Rebound Effect for US Households,
Energy Journal, 20 (3), 1-31.
Greening, L.A., D.L. Greene and C. Difiglio (2000). "Energy Efficiency and Consumption - The
Rebound Effect - A Survey," Energy Policy, 28,3 89-401.
Haughton, J. and S. Sarkar (1996). "Gasoline Tax as a Corrective tax: Estimates for the United
States (1970-1991)," Energy Journal, 17(2), 103-126.
Helfand, Gloria, and Ann Wolverton (2011), "Evaluating the Consumer Response to Fuel
Economy: A Review of the Literature," International Review of Environmental and Resource
Economics, 5, 103-146
Hughes, Jonathan E., Christopher R. Knittel, and Daniel Sperling (2008), "Evidence of a Shift in the Short-
Run Price Elasticity of Gasoline Demand/' Energy Journal, 29(1): 113-134.Hymel, Kent, Kenneth A.
Small, and Kurt Van Dender (2010). "Induced Demand and Rebound Effects in Road Transport,"
Transportation Research Part B - Methodological, 44( 10), 1220-1241.
Knittel, Christopher R., and Ryan Sandier (2012), "Carbon Prices and Greenhouse Gas
Emissions: The Intensive and Extensive Margins," in Don Fullerton and Catherine Wolfram,
eds., Design and Implementation of U.S. Climate Policy, University of Chicago Press, ch. 18, pp.
287-299.
Knittel, Christopher R., and Ryan Sandier (2013), "The Welfare Impact of Indirect Pigouvian
Taxation: Evidence from Transportation," Working Paper 18849, National Bureau of Economic
Research. http://www.nber.org/papers/wl8849Jones C.T. (1993). "Another Look at U.S.
Passenger Vehicle Use and the 'Rebound' Effect from Improved Fuel Efficiency," Energy
Journal, 14 (4), 99-110.
Litman, Todd (2010), "Changing Vehicle Travel Price Sensitivities: The Rebounding Rebound
Effect," Victoria Transportation Policy Institute, Victoria, B.C. www.vtpi.orgMackie P.J., A.S.
Fowkes, M. Wardman, G. Whelan, J. Nellthorp, and J. Bates (2003) Values of Travel Time
Savings in the UK: Summary Report, report to the UK Department for Transport. Leeds, UK:
Institute of Transport Studies, Univ. of Leeds.
http://www.dft.gov.uk/pgr/economics/rdg/valueoftraveltimesavingsinth3130?page=l#al OOP
73
-------
Mannering, F.L. (1986). A Note on Endogenous Variables in Household Vehicle Utilization Equations,
Transportation Research B, 20 (1), 1-6.
Molloy, Raven, and Hui Shan (2010), "The effect of gasoline prices on household location," working
paper 2010-36, Finance and Economics Discussion Series, Divisions of Research & Statistics and
Monetary Affairs, Federal Reserve Board, Washington, D.C. (June).
http://econpapers.repec.org/scripts/redir .pf?u=http%3A%2F%2Fwww.federalreserve.gov%2Fpubs%2Ffe
ds%2F2010%2F201036%2F201036pap.pdf;h=repec:fip:fedgfe:2010-36
Quantitative Micro Software (2004). EViews 5 User's Guide, QMS, Irvine, Calif.
Schrank, David, Tim Lomax, and Shawn Turner (2010). Urban Mobility Report 2010, Texas
Transportation Institute, December, http ://mobility.tamu.edu
Small K.A. and K. Van Dender (2005). A Study to Evaluate the Effect of Reduced Greenhouse Gas
Emissions on Vehicle Miles Traveled, Final Report to the California Air Resources Board, Contract No.
02-336.
Small K.A. and K. Van Dender (2007a). "Fuel Efficiency and Motor Vehicle Travel: The Declining
Rebound Effect," Energy Journal 28: 25-51.
Small K.A. and K. Van Dender (2007b). "Fuel Efficiency and Motor Vehicle Travel: The Declining
Rebound Effect," Working Paper No. 05-06-03, Department of Economics, University of California at
Irvine (revised).
Small K.A. and K. Van Dender (2007c). "If Cars Wee More Efficient, Would We Use Less Fuel?"
Access, 31 (Fall 2007), pp. 8-13. Berkeley: University of California Transportation Center.
Small K.A. and E.T. Verhoef (2007) The Economics of Urban Transportation, London and New York:
Routledge, forthcoming.
US Census Bureau (2009). Statistical Abstract of the United States: 2009. Washington, D.C.: US
Census Bureau.
US Council of Economic Advisors (2009). Annual Report of the Council of Economic Advisers.
Printed with Economic Report of the President. Washington, D.C.: Government Printing Office.
US EIA (2011). Annual Energy Outlook 2011: With Projections to 2035. Report Number: DOE/EIA-
0383(2011). Washington, D.C.: US Energy Information Administration, April 26.
http://www.eia.doe.gov/forecasts/aeo/index.cfm (accessed May 10, 2011).
US FHWA (2006) Highway Statistics 2005, Washington, D.C: US Federal Highway Administration, Report
FHWA-PL-06-009, December.
US FHWA (2008) 2008 Status of the Nation's Highways, Bridges, and Transit: Conditions and
Performance, U.S. Federal Highway Administration.
http://www.fhwa.dot.gov/policy/2008cpr/index.htm
74
-------
Wadud Z., D.J. Graham, and R.B. Noland (2007a). "Gasoline demand with heterogeneity in household
responses/' Working paper, Centre for Transport Studies, Imperial College London.
Wadud Z., D.J. Graham, and R.B. Noland (2007b). "Modelling fuel demand for different socio-economic
groups," Working paper, Centre for Transport Studies, Imperial College London.West, Sarah E. (2004),
"Distributional effects of alternative vehicle pollution control policies," Journal of Public Economic, 88:
735-757.
75
-------
Appendix A. Calculation of Dynamic Rebound Effect
The dynamic rebound takes into account that interacting variables, especially income and fuel price, are
changing over the course of the life of a vehicle—even its life beyond the projection period which ends
in 2035. It is calculated by projecting the dynamic adjustment process that is implied by the estimated
equations but allowing the "target" amount of travel to change each year according to actual or
projected conditions (income, fuel price, and urbanization and/or congestion) for that year—using
actual data from my data sources for 2000-2009 and data from the AEO projections for 2010-2035. (The
projection data are adjusted to match the estimation data for years 2008-2009, so that projections are
consistent with the estimated equations.)
This "target" is based on an adjustment to the typical mileage for a vehicle of a given age, as derived
from the National Personal Travel Survey (NPTS) and reported by the Transportation Energy Data Book,
ed. 29, Table 8.9. The adjustment occurs from two sources: changes in the interaction variables that
determine the long-run rebound effect, and the assumed unit change in fuel cost per mile resulting from
a policy. The adjustment is derived from the equations for the structural elasticity of mileage with
respect to fuel cost per mile (SM. PM in the source papers), which is influenced directly by the interaction
variables according to their estimated coefficients, and from the equation that converts SM. PM into a
long-run rebound effect.34 The actual mileage of a vehicle purchased in year t in a subsequent year M-r,
where ris the age of the vehicle, is projected as the weighted average of the previous year's mileage,
adjusted for the natural evolution due to the age-mileage profile, and the target mileage, which is based
on the age-mileage profile and the long-term rebound elasticity; the weights in taking this average are
ccm and (l-am), respectively, where ccm is the coefficient of the lagged dependent variable in the
estimated equation for vehicle-miles per adult. (This notation conforms with the two papers just cited in
the footnote.)
The actual procedure used to compute the dynamic rebound effect has three steps:
• First, the short-run rebound effect is recomputed for each year assuming that all variables except fuel
efficiency change as in the projection being considered.35 This projects the desired short-run response
34 Those equations are equation (7) in Small and Van Dender (2007) and equations (14a) and (15) in Hymel, Small,
andVanDender(2010).
35 Our projections are through year 2035. Vehicles sold in the lateryears of the projection will last beyond 2035, and
for those years we use 2035 values of interacting variables to compute the short-run rebound effect applying to these
vehicles as they age.
Appendix page 1
-------
that would occur for the owner of a vehicle whose fuel efficiency remains fixed as it ages, but who faces
other changes (income, fuel price, urbanization, congestion) that affect the owner's response.36 The
'"-'' rr '-'' rr '^~" rr
resulting change over the vehicle's lifetime is denoted by A0 t,i-u t+r-u t, where t is the year of
purchase and ris the vehicle's age.
• Simultaneously, these changes in short-run rebound as the vehicle ages are converted to the
corresponding change in structural elasticity using equation (lla) of Hymel et al. (2010), and that in turn
is converted to a change in long-run target response using equation (14a) of the same paper:
bLt+T = bLt+(bst+T-bst)-
D
where bL, is the long-run rebound for year t as already calculated, and D and DL are quantities defined
in Hymel et al.'s equation which account for effects of the equations for vehicle fleet size and vehicle
fuel efficiency when computing the short- and long-run rebound effects, respectively. As an
approximation, we assume the conversion factors D and DL are constant, although they actually change
very slightly over time. The ratio D/DL is actually very close to the simple multiplier, l/(l-of"), which
converts a short-run to a long-run response.37
• Finally, the baseline age-mileage profile mentioned earlier, denoted by M T for ages 7=0,1,..., 15, is
used as the starting point for changes in mileage over each year of the vehicle's age.38 The computation
assumes a unit increase in fuel cost per mile. (The size and sign of the change in fuel cost per mile is
immaterial because the equations are linear so they lead to the same answer once one divides by that
36 Because of the form of the estimating equations, which are linear in logarithms even accounting for interaction
variables, this calculation depends only very slightly on which year's fuel efficiency is chosen to hold constant:
namely, it depends on it through the truncation that occurs for those few state-year combinations that would
otherwise lead to a positive projected elasticity of VMT with respect to fuel cost (those values are truncated at zero).
Thus for the projections starting in 2010, the computation is simplified by assuming fuel efficiency is held constant
at its projected value for 2020; for the historical computations for 2000-2009, it is held constant at its actual value
for 2005.
37 The equations for D and Ef- in Hymel et al. (2010) are for the four-equation version of the model; they are also
valid for the 3 -equation version, simply by setting the coefficient of"1, which is absent in the latter, equal to zero.
38 The age-mileage profile is derived from the National Personal Travel Survey (NPTS) and reported in the
Transportation Energy Data Book, ed. 29, Table 8.9.
Appendix page 2
-------
change.) The projected mileage after response to the change in fuel cost per mile, for a new car
purchased in year t, is the weighted average described earlier:
MT = am
T-l
This is computed iteratively; for year 0 (the year the vehicle is purchased), the simple short-run response
as already projected is used:
s°
M0=(l-bt}M
In these equations, fa is a "rebound effect" defined as the negative of the relevant elasticity, so is
normally positive (or zero, if truncated); this is why it appears with a minus sign in the equation.
Appendix page 3
-------
Appendix B. Coefficient estimates
Table Bl. Coefficient estimates: Symmetric and asymmetric models
(a) Three-equation models
Equation
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
veh
veh
veh
veh
veh
veh
veh
veh
veh
veh
veh
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
Variable
intercept
income
adults per road mile
popratio
Urban
Railpop
D7479
Trend
vma(-1)
vehstock
pf+fint
pmA2
pm*inc
pm*Urban
pm*(dummy 2003-09)
pfcut
AR(1)
intercept
pnewcar
interest
income
Adults per road mile
licenses/adult
trend
vehstock(-l)
vma
pm
AR(1)
intercept
pf + vma
income
fint(-1)
Population Ratio
Trend66-73
Trend74-79
Trend80+
D7479
Urban
cafe
pfcut
Model 3.3
Coeff. Std. Error
1.6261
0.0781
-0.0149
0.0726
-0.0205
-0.0067
-0.0439
-0.0004
0.8346
0.0209
-0.0466
-0.0124
0.0528
0.0119
-0.1018
-0.2253
0.0400
-0.0008
0.0032
-0.0136
0.0345
0.0002
0.9318
0.0291
0.0013
-0.1461
-0.2447
-0.0050
-0.0016
0.9040
-0.0168
0.0005
-0.0068
-0.0007
-0.0070
-0.0905
-0.0345
-0.1773
0.1022
0.0117
0.0038
0.0322
0.0391
0.0043
0.0034
0.0002
0.0102
0.0067
0.0029
0.0059
0.0108
0.0094
0.0204
0.1452
0.0317
0.0042
0.0146
0.0060
0.0184
0.0007
0.0104
0.0147
0.0058
0.0230
0.0631
0.0041
0.0144
0.0100
0.0603
0.0011
0.0010
0.0003
0.0048
0.0467
0.0108
0.0201
Model 3.18
Coeff. Std. Error
1 .6771
0.0782
-0.0147
0.0836
-0.0372
-0.0053
-0.0436
-0.0003
0.8279
0.0238
-0.0464
-0.0113
0.0699
0.0078
-0.0251
-0.2188
0.0376
-0.001 1
0.0033
-0.0135
0.0344
0.0002
0.9323
0.0285
0.0009
0.0376
-0.2577
-0.0052
-0.0009
0.9036
0.0154
0.0006
-0.0060
-0.0007
-0.0082
-0.0869
-0.0402
-0.1756
0.1035
0.0117
0.0038
0.0325
0.0395
0.0043
0.0034
0.0002
0.0105
0.0068
0.0029
0.0060
0.0121
0.0096
0.0076
0.1451
0.0317
0.0042
0.0146
0.0060
0.0183
0.0007
0.0104
0.0147
0.0058
0.0317
0.0631
0.0041
0.0144
0.0100
0.0602
0.0011
0.0010
0.0003
0.0048
0.0467
0.0108
0.0201
Model 3.20b
Coeff. Std. Error
2.2568
0.0814
-0.0147
0.0804
-0.0211
-0.0080
-0.0432
0.0002
0.8256
0.0202
-0.0520
-0.0159
0.0569
0.0124
0.0124
-0.1038
-0.2174
0.0432
-0.0006
0.0037
-0.0137
0.0345
0.0003
0.9319
0.0281
0.0015
-0.1464
2.4538
-0.0185
-0.0048
0.9140
-0.0160
0.0005
-0.0058
0.0008
-0.0041
-0.0778
-0.0202
0.0316
-0.1822
0.4424
0.0117
0.0037
0.0329
0.0388
0.0043
0.0034
0.0004
0.0105
0.0067
0.0046
0.0061
0.0108
0.0093
0.0093
0.0205
0.1450
0.0317
0.0042
0.0146
0.0060
0.0183
0.0007
0.0104
0.0147
0.0058
0.0230
1 .0475
0.0057
0.0145
0.0109
0.0592
0.0011
0.0011
0.0007
0.0048
0.0470
0.0186
0.0124
0.0201
Model 3.21 b
Coeff. Std
3.1468
0.0770
-0.0151
0.0630
-0.0061
-0.0082
-0.0445
0.0013
0.8334
0.0161
pf+fint -0.0639
-0.0207
0.0577
0.0131
pfcut + fint 0.0340
-0.1021
-0.2188
0.0460
-0.0004
0.0038
-0.0137
0.0349
0.0004
0.9316
0.0281
0.0019
-0.1469
0.9282
pf + vma -0.0097
0.0000
0.8977
-0.0005
-0.0005
-0.0061
-0.0002
-0.0032
-0.0890
-0.0256
pfcut + vma 0.0143
-0.1804
. Error
0.3541
0.0118
0.0037
0.0323
0.0395
0.0042
0.0035
0.0004
0.0104
0.0067
0.0049
0.0061
0.0107
0.0093
0.0078
0.0204
0.1449
0.0317
0.0042
0.0146
0.0060
0.0183
0.0007
0.0104
0.0146
0.0058
0.0230
1.0517
0.0060
0.0146
0.0115
0.0586
0.0011
0.0011
0.0007
0.0048
0.0471
0.0183
0.0123
0.0202
Model 3.23
Coeff.
3.3926
0.0792
-0.0200
0.0732
0.0021
-0.0061
-0.0425
0.0013
0.8084
0.0203
pf+fint -0.0623
-0.0180
0.0535
0.0187
0.0284
-0.0978
-0.2232
0.0444
-0.0003
0.0036
-0.0138
0.0339
0.0004
0.9316
0.0286
0.0017
-0.1461
1.1934
pfrise -0.0133
-0.0041
0.9106
-0.0073
0.0001
-0.0057
0.0001
-0.0046
-0.0828
-0.0312
pfcut 0.0042
vma 0.0107
-0.1807
Std. Error
0.549(
0.01 2(
0.004
0.033'
0.040"
0.004!
0.003'
0.000!
0.012;
0.007(
0.005!
o.ooe;
0.011;
0.009S
0.009;
0.021!
0.145
o.osr
0.004;
0.014!
0.006(
0.018'
O.OOQ-
0.010'
0.014-
0.005!
0.023!
1.208
o.ooe;
0.015
0.012!
0.059'
0.001;
0.001
0.000"
0.004!
0.046:
0.018!
0.009!
0.016
0.020
Appendix page 4
-------
(b) Four-equation models
Model 4.3
Equation Variable Coeff. Std. Err.
vma intercept 1.6801 0.1066
vma inc 0.0835 0.0117
vma congestion 0.0014 0.0027
vma conglnc -0.0156 0.0032
vma cong*pm -0.0031 0.0022
vma D7479 -0.0430 0.0034
vma Trend -0.0003 0.0002
vma vma(-1) 0.8249 0.0105
vma vehstock 0.0276 0.0065
vma pm -0.0461 0.0030
vma pmA2 -0.0224 0.0060
vma pm'inc 0.0561 0.0111
vma popratio 0.1201 0.0384
vma urban -0.0842 0.0413
vma road miles/land area 0.0180 0.0065
vma pm*(dummy for 2003-09)
vma pfcut
vma
vma
vma
vma
vma
vma AR(1) -0.0900 0.0207
vehstock intercept -0.3535 0.1422
vehstock pnewcar 0.0418 0.0317
vehstock interest -0.0033 0.0040
vehstock income 0.0044 0.0146
vehstock urban -0.0420 0.0465
vehstock licenses/adult 0.0441 0.0178
vehstock trend 0.0000 0.0007
vehstock vehstock(-l) 0.9354 0.0102
vehstock vma 0.0384 0.0143
vehstock pm 0.0028 0.0057
vehstock rho -0.1468 0.0230
fint intercept -0.3202 0.0618
fint pf + vma -0.0074 0.0041
fint inc -0.0002 0.0143
fint fint(-1) 0.8894 0.0102
fint Trend66-73 0.0013 0.0009
fint Trend74-79 -0.0038 0.0008
fint Trend80+ -0.0010 0.0003
fint 7479 dummy -0.0118 0.0047
fint Urban -0.0847 0.0468
fint cafe -0.0607 0.0103
fint popratio 0.1096 0.0556
fint pfcut+vma
fint
fint rho -0.1694 0.0201
cong intercept -3.8401 0.9940
cong urban-lane-miles/adult -0.6926 0.1316
cong (vehicle miles/adult)+log(ui 0.2258 0.0885
cong population / state land aree 0.6121 0.0520
cong percent trucks 0.4597 0.2062
cong urban -4.3113 0.3550
Model 4.1 3
Coeff. Std. Err.
1.7249 0.1078
0.0839 0.01 1 7
0.0014 0.0027
-0.0146 0.0032
-0.0032 0.0022
-0.0429 0.0034
-0.0002 0.0002
0.8189 0.0107
0.0308 0.0066
-0.0460 0.0030
-0.0186 0.0061
0.0721 0.0121
0.1289 0.0386
-0.0980 0.0416
0.0173 0.0066
-0.0237 0.0071
-0.0856 0.0208
-0.3516 0.1422
0.0392 0.031 7
-0.0036 0.0040
0.0043 0.0146
-0.0424 0.0465
0.0440 0.0178
-0.0001 0.0007
0.9357 0.0102
0.0384 0.0143
0.0025 0.0057
-0.1471 0.0230
-0.3191 0.0619
-0.0075 0.0041
-0.0002 0.0143
0.8900 0.0102
0.0013 0.0010
-0.0037 0.0008
-0.0010 0.0003
-0.0119 0.0047
-0.0839 0.0468
-0.0601 0.0103
0.1130 0.0557
-0.1691 0.0201
-3.8457 0.9940
-0.6931 0.1316
0.2263 0.0885
0.6119 0.0520
0.4594 0.2062
-4.3124 0.3550
Model 4.20b Model 4.21 b Model 4.23
Coefficient Std. Error I Coefficient Std. Error
2.1693 0.4400
0.0847 0.0117
0.0032 0.0026
-0.0134 0.0031
-0.0013 0.0021
-0.0430 0.0034
0.0000 0.0005
0.8221 0.0107
0.0282 0.0066
-0.0498 0.0046
-0.0225 0.0061
0.0548 0.0111
0.1006 0.0419
-0.0694 0.0409
0.0181 0.0065
0.0100 0.0093
-0.0901 0.0207
-0.3569 0.1421
0.0430 0.031 7
-0.0032 0.0040
0.0043 0.0146
-0.0418 0.0465
0.0442 0.0178
0.0000 0.0007
0.9353 0.0102
0.0387 0.0143
0.0030 0.0057
-0.1468 0.0230
0.4210 0.9482
-0.0125 0.0055
0.0021 0.0144
0.8950 0.0106
0.0011 0.0010
-0.0028 0.0009
-0.0005 0.0006
-0.0088 0.0047
-0.0801 0.0470
-0.0678 0.0158
0.1293 0.0562
0.0085 0.0112
-0.1702 0.0201
-4.1046 0.9274
-0.6057 0.1102
0.2825 0.0860
0.5900 0.0490
0.4622 0.1983
-4.0385 0.3434
3.1388 0.3529
0.0807 0.0119
0.0016 0.0026
-0.0131 0.0031
-0.0016 0.0021
-0.0441 0.0035
0.0013 0.0005
0.8305 0.0107
0.0242 0.0066
-0.0629 0.0049
-0.0275 0.0061
0.0573 0.0110
0.1010 0.0410
-0.0589 0.0415
0.0155 0.0066
pfcut+fint 0.0340 0.0079
-0.0888 0.0206
-0.3554 0.1421
0.0445 0.0317
-0.0030 0.0040
0.0044 0.0146
-0.0416 0.0465
0.0445 0.0178
0.0000 0.0007
0.9351 0.0102
0.0384 0.0143
0.0032 0.0057
-0.1471 0.0230
-1.0263 0.9488
-0.0041 0.0058
0.0064 0.0144
0.8805 0.01 1 1
0.0001 0.0010
-0.0034 0.0009
-0.0014 0.0006
-0.0078 0.0047
-0.0919 0.0471
-0.0714 0.0155
0.1302 0.0556
pfcut+vma -0.0080 0.0112
-0.1691 0.0202
-4.0860 0.9273
-0.6058 0.1102
0.2799 0.0860
0.5908 0.0490
0.4634 0.1983
-4.0331 0.3434
Coefficien Std. Error
3.4021 0.4991
0.0781 0.0120
-0.0001 0.0028
-0.0166 0.0033
-0.0042 0.0022
-0.0441 0.0035
0.0014 0.0005
0.8229 0.0112
0.0274 0.0067
-0.0615 0.0054
-0.0245 0.0063
0.0534 0.0115
0.1437 0.0394
-0.0763 0.0419
0.0181 0.0067
pmcut_hat 0.0325 0.0091
-00932 00212
-0.3653 0.1422
0.0412 0.0318
-0.0030 0.0040
0.0041 0.0146
-0.0424 0.0466
0.0438 0.0178
-0.0001 0.0007
0.9348 0.0102
0.0396 0.0143
0.0028 0.0058
-0.1458 0.0230
0.7587 1 .0646
prfise -0.0122 0.0063
0.0005 0.0149
0.9108 0.0117
0.0010 0.0010
-0.0048 0.0010
0.0004 0.0006
-0.0033 0.0046
-0.0724 0.0462
0.0064 0.0158
0.1744 0.0542
pfcut 0.0024 0.0086
vma 0.0210 0.0152
-0.1753 0.0198
-4.6094 0.9904
-0.7682 0.1296
0.2914 0.0900
0.6424 0.0521
0.4061 0.2093
-4.6372 0.3616
Appendix page 5
-------
Table B2. Coefficient estimates: models with media and uncertainty variables
(a) Three-equation models
Equation
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
veh
veh
veh
veh
veh
veh
veh
veh
veh
veh
veh
Tint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
Variable
intercept
inc
Adults /road mile
popratio
Urban
Railpop
D7479
Trend
vma(-1)
vehstock
pf+fint
pm"2
pm*inc
pm*Urban
pfcut + fint
Media variable
pm*(dummy 2003-09)"
Fuel price variance
AR(1)
intercept
pnewcar
interest
income
adults /road mile
licenses/adult
trend
vehstock(-l)
vma
pm
AR(1)
intercept
pf + vma
inc
fint(-1)
popratio
Trend66-73
Trend74-79
Trend80+
D7479
Urban
cafe
pfcut
ARM)
Model 3.21 b
Coeff. Std. Error
3.1468
0.0770
-0.0151
0.0630
-0.0061
-0.0082
-0.0445
0.0013
0.8334
0.0161
-0.0639
-0.0207
0.0577
0.0131
0.0340
-0.1021
-0.2188
0.0460
-0.0004
0.0038
-0.0137
0.0349
0.0004
0.9316
0.0281
0.0019
-0.1469
0.9282
-0.0097
0.0000
0.8977
-0.0005
-0.0005
-0.0061
-0.0002
-0.0032
-0.0890
-0.0256
0.0143
-0.1804
0.3541
0.0118
0.0037
0.0323
0.0395
0.0042
0.0035
0.0004
0.0104
0.0067
0.0049
0.0061
0.0107
0.0093
0.0078
0.0204
0.1449
0.0317
0.0042
0.0146
0.0060
0.0183
0.0007
0.0104
0.0146
0.0058
0.0230
1.0517
0.0060
0.0146
0.0115
0.0586
0.0011
0.0011
0.0007
0.0048
0.0471
0.0183
0.0123
0.0202
"dummy is normalized
Model 3.35
Coeff.
2.9103
0.0830
-0.0142
0.0725
-0.0114
-0.0084
-0.0440
0.0011
0.8325
0.0162
pf +fint -0.0587
-0.0053
0.0583
0.0118
pfcut + fint 0.0286
pf *Media_dummy -0.0301
-0.0969
-0.2117
0.0449
-0.0002
0.0039
-0.0139
0.0348
0.0004
0.9316
0.0274
0.0016
-0.1469
pf+vma -0.0124
-0.0031
0.9070
-0.0391
0.0000
-0.0075
0.0005
-0.0015
-0.0872
-0.0023
pfCut + vma 0.0220
-0.1851
Std. Error
0.3668
0.0121
0.0038
0.0328
0.0400
0.0043
0.0035
0.0005
0.0106
0.0068
0.0052
0.0075
0.0109
0.0094
0.0081
0.0101
0.0206
0.1449
0.0317
0.0042
0.0146
0.0060
0.0183
0.0007
0.0104
0.0146
0.0058
0.0230
Model 3.37
Coeff.
3.1487
0.0828
-0.0145
0.0786
-0.0231
-0.0076
-0.0436
0.0014
0.8276
0.0181
pf +fint -0.0641
-0.0064
0.0711
0.0100
pfcut + fint 0.0332
pf *Media_dummy -0.0319
-0.0216
-0.0894
-0.1996
0.0434
-0.0004
0.0043
-0.0139
0.0346
0.0003
0.9319
0.0262
0.0012
-0.1475
1.0241 0.8319
0.0059
0.0145
0.0115
0.0590
0.0011
0.0011
0.0007
0.0048
0.0470
0.0172
0.0120
0.0202
pf + vma -0.0104
-0.0003
0.9009
0.0020
-0.0002
-0.0063
-0.0001
-0.0031
-0.0876
-0.0210
pfCut + vma 0.0129
-0.1810
Std. Error
0.3810
0.0123
0.0039
0.0334
0.0407
0.0044
0.0035
0.0005
0.0109
0.0070
0.0057
0.0075
0.0126
0.0097
0.0083
0.0101
0.0079
0.0209
0.1445
0.0317
0.0042
0.0146
0.0060
0.0183
0.0007
0.0104
0.0146
0.0058
0.0230
1.0025
0.0058
0.0145
0.0115
0.0585
0.0011
0.0011
0.0007
0.0048
0.0468
0.0169
0.0118
0.0202
Model 3.42
Coeff. Std
3.9416
0.0746
-0.0140
0.1462
-0.0132
-0.0065
-0.0429
0.0024
0.8321
0.0185
pf +fint 3.9959
-0.0126
0.0779
0.0091
pfcut + fint 0.0529
pf*Media_dummy -0.0316
-0.0265
pm*log(pf_var) 0.0028
-0.0960
-0.2249
0.0423
-0.0004
0.0033
-0.0136
0.0355
0.0003
0.9314
0.0289
0.0014
-0.1466
0.0017
pf + vma -0.0079
0.0050
0.8930
0.0070
-0.0017
-0.0045
-0.0009
-0.0049
-0.0920
-0.0592
PFCut + VMA 0.0031
-0.1786
Errc
0.4C
0.0
o.ot
o.o:
0.0'
o.ot
o.ot
o.ot
0.0
o.ot
o.ot
o.ot
0.0
o.ot
o.ot
0.0
o.ot
o.ot
o.o;
0.1'
o.o:
o.ot
o.o
o.ot
0.0
o.ot
0.0
0.0
o.ot
o.o;
0.9f
o.ot
0.0
0.0
o.o:
o.ot
o.ot
o.ot
o.ot
0.0'
0.0
0.0
o.o;
Appendix page 6
-------
(b) Four-equation models
Equation
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vma
vehstock
vehstock
vehstock
vehstock
vehstock
vehstock
vehstock
vehstock
vehstock
vehstock
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
fint
cong
cong
cong
cong
cong
cong
Variable
intercept
inc
cong
cong*income
cong*pm
7479 dummy
trend
vma(-1)
vehstock
pm
pm"2
pmlnc
popratio
urban
road miles/state land area
pfcut + fint
Media variable
pm*(dummy 2003-09)"
Fuel price variance
AR(1)
intercept
pnewcar
interest
income
urban
licenses/adult
trend
vehstock(-l)
vma
pm
intercept
pf + vma
inc
fint(-1)
Trend66-73
Trend74-79
Trend80+
7479 dummy
urban
cafep
popratio
pfcut+vma
rho
intercept
urban-lane-miles/adult
(vehicle miles/adult)+log(ui
population / state land aree
percent trucks
urban
Model 4.21 b
Coeff. Std. Error
3.1388
0.0807
0.0016
-0.0131
-0.0016
-0.0441
0.0013
0.8305
0.0242
-0.0629
-0.0275
0.0573
0.1010
-0.0589
0.0155
0.0340
-0.0888
-0.3554
0.0445
-0.0030
0.0044
-0.0416
0.0445
0.0000
0.9351
0.0384
0.0032
0 1471
-1.0263
-0.0041
0.0064
0.8805
0.0001
-0.0034
-0.0014
-0.0078
-0.0919
-0.0714
0.1302
-0.0080
-01691
-4.0860
-0.6058
0.2799
0.5908
0.4634
-4.0331
0.3529
0.0119
0.0026
0.0031
0.0021
0.0035
0.0005
0.0107
0.0066
0.0049
0.0061
0.0110
0.0410
0.0415
0.0066
0.0079
0.0206
0.1421
0.0317
0.0040
0.0146
0.0465
0.0178
0.0007
0.0102
0.0143
0.0057
0 0230
0.9488
0.0058
0.0144
0.0111
0.0010
0.0009
0.0006
0.0047
0.0471
0.0155
0.0556
0.0112
0.0202
0.9273
0.1102
0.0860
0.0490
0.1983
0.3434
Model 4.35
Coeff.
3.1737
0.0791
0.0011
-0.0144
-0.0025
-0.0445
0.0014
0.8314
0.0236
PM -0.0638
-0.0296
0.0575
0.1093
-0.0639
0.0148
pfcut+fint 0.0352
pf * Media_dummy 0.0061
-0.0913
-0.3577
0.0443
-0.0030
0.0043
-0.0417
0.0446
0.0000
0.9350
0.0387
0.0031
-0 1469
-0.6026
pf + vma -0.0060
0.0064
0.8833
0.0002
-0.0037
-0.0010
-0.0069
-0.0896
-0.0585
0.1330
pfcut+vma -0.0031
-0 1706
-3.9180
-0.6394
0.2546
0.6062
0.4554
-4.2241
Std. Error
0.3555
0.0119
0.0027
0.0032
0.0021
0.0035
0.0005
0.0106
0.0065
0.0050
0.0065
0.0110
0.0397
0.0415
0.0065
0.0080
0.0058
0.0206
0.1421
0.0317
0.0040
0.0146
0.0465
0.0178
0.0007
0.0102
0.0143
0.0057
0 0230
0.9380
0.0057
0.0144
0.0110
0.0010
0.0009
0.0006
0.0047
0.0470
0.0148
0.0553
0.0110
0.0202
0.9530
0.1176
0.0872
0.0502
0.2016
0.3484
Model 4.37
Coeff.
3.5432
0.0794
0.0006
-0.0128
-0.0028
-0.0444
0.0019
0.8221
0.0277
PM -0.0729
-0.0263
0.0825
0.1248
-0.0828
0.0133
pfcut+fint 0.0420
pf *Media_dummy 0.0071
-0.0359
-0.0840
-0.3557
0.0412
-0.0035
0.0042
-0.0421
0.0445
-0.0001
0.9354
0.0387
0.0028
-0 1474
-0.5382
pf + vma -0.0059
0.0066
0.8823
0.0000
-0.0035
-0.0010
-0.0068
-0.0894
-0.0583
0.1360
pfcut+vma -0.0022
-0.1704
-3.8896
-0.6352
0.2533
0.6088
0.4506
-4.2191
Std. Error
0.3653
0.0119
0.0027
0.0032
0.0021
0.0035
0.0005
0.0109
0.0066
0.0054
0.0066
0.0122
0.0399
0.0419
0.0066
0.0081
0.0058
0.0071
0.0207
0.1420
0.0317
0.0040
0.0146
0.0465
0.0178
0.0007
0.0102
0.0143
0.0057
0 0230
0.9373
0.0057
0.0144
0.0110
0.0010
0.0009
0.0006
0.0047
0.0470
0.0148
0.0553
0.0110
00909
0.9664
0.1217
0.0872
0.0504
0.2020
0.3485
Model 4.42
Coeff.
3.8758
0.0652
-0.0004
-0.0117
-0.0044
-0.0467
0.0024
0.8275
0.0268
PM -0.0706
0.0037
0.0944
0.0669
-0.0967
0.0111
pfcut+fint 0.0506
pf*Media_dummy -0.0080
-0.0308
pm*log(pf var) -0.0100
-0.0849
-0.3592
0.0403
-0.0038
0.0041
-0.0425
0.0444
-0.0001
0.9354
0.0391
0.0028
-0 1467
-0.5531
pf + vma -0.0049
0.0046
0.8749
0.0009
-0.0036
-0.0010
-0.0071
-0.0898
-0.0554
0.1700
pfcut+vma -0.0018
-0 1697
-3.8874
-0.6311
0.2552
0.6103
0.4493
-4.2251
Std. Err
0.37
0.01
0.00
0.00
0.00
0.00
0.00
0.01
0.00
0.00
0.00
0.01
0.04
0.04
0.00
0.00
0.00
0.00
0.00
0.02
0.14
0.03
0.00
0.01
0.04
0.01
0.00
0.01
0.01
0.00
0 02
0.93
0.00
0.01
0.01
0.00
0.00
0.00
0.00
0.04
0.01
0.05
0.01
0.02
0.96
0.12
0.08
0.05
0.20
0.34
"dummy is normalized
Appendix page 7
-------
Appendix C. Detailed yearly projections
Model 3.3:
2000 2001 2002 2003 2004 2005 2006 2007
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
Short Run Rebound
Dynamic Rebound
Long Run Rebound
2.3% 2.1% 2.1% 2.4% 2.6% 2.9% 3.0% 3.0% 3.3% 2.5% 2.8% 2.9% 2.8% 2.8% 2.8%
11.1% 11.3% 11.5% 11.8% 12.0% 12.0% 12.0% 11.8% 11.7% 11.4% 11.4% 11.1% 10.8% 10.5% 10.1%
14.7% 13.1% 13.0% 14.9% 16.4% 18.4% 18.8% 19.0% 20.7% 15.9% 17.6% 18.1% 17.7% 17.9% 17.4%
2.7% 2.5% 2.4%
9.6% 9.2% 8.8%
16.7% 15.9% 15.4%
2.4% 2.3% 2.2% 2.0% 2.0% 1.8% 1.8%
8.3% 7.9% 7.4% 6.9% 6.5% 6.1% 5.7%
14.9% 14.4% 13.7% 12.9% 12.3% 11.5% 11.0%
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
2.3% 2.1% 2.1% 2.4% 2.6% 2.9% 3.0% 3.0% 3.3% 2.5% 2.8% 3.3% 3.5% 3.6% 3.5% 3.4% 3.3% 3.3% 3.2% 3.2% 3.1% 2.9% 2.8% 2.7% 2.6%
11.5% 11.8% 12.3% 12.8% 13.2% 13.5% 13.7% 13.9% 14.0% 14.1% 14.4% 14.5% 14.4% 14.1% 13.7% 13.3% 12.9% 12.5% 12.0% 11.6% 11.1% 10.6% 10.1% 9.6% 9.3%
14.7% 13.1% 13.0% 14.9% 16.4% 18.4% 18.8% 19.0% 20.7% 15.9% 17.6% 20.8% 22.1% 22.6% 22.2% 21.7% 21.0% 20.8% 20.4% 19.9% 19.3% 18.6% 17.6% 17.0% 16.3%
2.3% 2.1% 2.1% 2.4% 2.6% 2.9% 3.0% 3.0% 3.3%
10.6% 10.6% 10.7% 10.7% 10.6% 10.4% 10.0% 9.5% 8.9%
14.7% 13.1% 13.0% 14.9% 16.4% 18.4% 18.8% 19.1
2.5% 2.8% 2.4% 2.2% 2.1% 1.9% 1.7% 1.5% 1.4% 1.3% 1.2% 1.2% 0.9%
8.9% 8.2% 7.8% 7.1% 6.5% 6.0% 5.5% 4.9% 4.5% 4.0% 3.5% 3.0% 2.6% 2.1%
20.7% 15.9% 17.6% 14.8% 13.8% 12.9% 11.8% 10.6% 9.6% 8.7% 8.1% 7.4% 7.4% 5.5%
0.8% 0.7% 0.6%
1.8% 1.4% 1.2%
4.7% 4.0% 3.7%
Model 3.21b:
.
2000 2001 2002 2003 2004 2005 2006 2007
Reference Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
0.4%
3.1%
2.2%
0.4%
3.7%
2.2%
0.4%
2.5%
2.2%
0.1%
3.4%
0.8%
0.1%
4.3%
0.8%
0.1%
2.6%
0.8%
0.1% 0.2%
3.7% 4.1%
0.6% 1.4%
0.1% 0.2%
4.9% 5.6%
0.6% 1.4%
0.1% 0.2%
2.8% 2.9%
0.6% 1.4%
0.5%
4.4%
3.2%
0.5%
6.3%
3.2%
0.5%
2.9%
3.2%
1.0% 1.2%
4.6% 4.6%
6.2% 7.3%
1.0% 1.2%
6.9% 7.4%
6.2% 7.3%
1.0% 1.2%
2.8% 2.5%
6.2% 7.3%
1.3%
4.6%
8.0%
1.3%
7.8%
8.0%
1.3%
2.1%
8.0%
2008
1 .7%
4.4%
10.5%
1 .7%
8.1%
10.5%
1 .7%
1 .6%
10.5%
2009 2010 2011 2012 2013 2014 2015
0.6%
4.2%
3.3%
0.6%
8.4%
3.3%
0.6%
1 .0%
3.3%
1.0%
4.2%
5.8%
0.9%
8.5%
5.7%
1.0%
2.0%
5.7%
1.1%
4.2%
6.5%
1 .7%
9.3%
10.6%
0.4%
1 .2%
2.5%
1 .0%
4.0%
6.1%
2.1%
9.4%
12.9%
0.3%
0.8%
1.6%
1.1%
3.8%
6.6%
2.3%
9.2%
14.0%
0.2%
0.5%
1.1%
1 .0%
3.5%
6.2%
2.2%
8.8%
13.7%
0.1%
0.3%
0.6%
0.9%
3.0%
5.6%
2.2%
8.4%
13.4%
0.1%
0.1%
0.3%
2016
0.8%
2.7%
4.9%
2.1%
8.0%
12.6%
0.0%
0.1%
0.2%
2017
0.8%
2.3%
4.5%
2.1%
7.5%
12.7%
0.0%
0.0%
0.1%
2018
0.7%
1 .9%
4.0%
2.0%
7.0%
12.2%
0.0%
0.0%
0.1%
2019 2020
0.6% 0.5%
1.6% 1.2%
3.7% 3.2%
1.9% 1.8%
6.4% 5.9%
11.8% 11.1%
0.0% 0.0%
0.0% 0.0%
0.0% 0.1%
2021
0.4%
0.9%
2.6%
1 .7%
5.3%
10.4%
0.0%
0.0%
0.0%
2022
0.4%
0.7%
2.2%
1.5%
4.7%
9.2%
0.0%
0.0%
0.0%
2023 2024
0.3%
0.5%
1 .7%
1 .4%
4.2%
8.6%
0.0%
0.0%
0.0%
0.2%
0.3%
1 .4%
1 .3%
3.8%
7.8%
0.0%
0.0%
0.0%
Model 3.35 (Reference case):
— Calculated using values of variables from historical data —
2000 2001 2002 2003 2004 2005 2006 2007
Short Run Rebound
Dynamic Rebound
1.1%
4.5%
6.6%
1 .0%
4.5%
6.1%
1 .0%
4.4%
6.3%
1.1%
4.4%
6.9%
1.2%
4.3%
7.0%
1.3%
4.2%
7.6%
1 .2%
4.0%
7.3%
2008
1 .2% 1 .2%
3.8% 3.6%
7.0% 7.5%
2009
1 .0%
3.4%
5.9%
2010
1.1%
3.3%
6.4%
2011
1.1%
3.0%
6.5%
2012
1 .0%
2.8%
6.3%
2013
1 .0%
2.5%
6.0%
2014
0.9%
2.2%
5.4%
2015
0.8%
1 .9%
4.7%
2016
0.7%
1 .6%
4.1%
2017
0.6%
1 .4%
3.7%
2018
0.6%
1 .2%
3.3%
2019
0.5%
1.0%
3.0%
ated using value
2020 2021
0.4%
0.8%
2.6%
0.4%
0.6%
2.2%
s of variables from A
2022 2023 20;
0.3%
0.5%
1.8%
0.3%
0.3%
1.5%
0.2
0.2
1 ?
Appendix page 8
-------
Model 4.3:
— Calculated using values of variables from historical data —
2000 2001 2002 2003 2004 2005 2006 2007
Reference Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
2.0% 1.6% 1.4% 1.9%
11.7% 12.0% 12.2% 12.5%
12.1% 9.2% 8.0% 11.4%
2.0% 1.6% 1.4% 1.9%
12.8% 13.5% 14.1% 14.9%
12.1% 9.2% 8.0% 11.4%
2.0% 1.6% 1.4% 1.9%
10.4% 10.3% 10.1% 9.8%
12.1% 9.2% 8.0% 11.4%
2.5% 3.1% 3.3%
12.7% 12.9% 12.8%
14.7% 18.6% 20.0%
2.5% 3.1% 3.3%
15.6% 16.3% 16.7%
14.7% 18.6% 20.0%
2.5% 3.1% 3.3%
9.5% 9.1% 8.2%
14.7% 18.6% 20.0%
3.4%
13.1%
20.8%
3.4%
17.7%
20.8%
3.4%
7.9%
20.8%
Calculated usin values of variables from AEO
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
3.9% 2.5%
13.5% 13.2%
23.5% 14.9%
3.9% 2.5%
17.7% 18.1%
23.5% 14.9%
3.9% 2.5%
8.6% 7.6%
23.5% 14.9%
3.0% 3.2% 3.1% 3.2% 3.2% 3.1% 2.9% 2.9% 2.8% 2.7% 2.6% 2.4% 2.3% 2.2% 2.1%
13.2% 13.1% 12.9% 12.5% 12.2% 11.7% 11.2% 10.7% 10.2% 9.7% 9.1% 8.6% 8.1% 7.6% 7.1%
18.2% 19.0% 18.7% 19.3% 19.0% 18.4% 17.6% 17.2% 16.6% 16.2% 15.4% 14.4% 13.9% 13.0% 12.5%
3.0% 3.9% 4.3% 4.5% 4.5% 4.5% 4.3% 4.4% 4.3% 4.2% 4.1% 4.0% 3.8% 3.7% 3.6%
18.6% 19.1% 19.2% 19.0% 18.6% 18.3% 17.8% 17.4% 16.8% 16.3% 15.7% 15.1% 14.5% 14.0% 13.5%
18.1% 23.8% 26.2% 27.5% 27.3% 27.1% 26.3% 26.5% 26.1% 25.7% 25.1% 24.3% 23.1% 22.5% 21.7%
3.0% 2.2% 2.0% 1.8% 1.6% 1.4% 1.2% 1.0% 0.9% 0.8% 0.9% 0.4% 0.3% 0.2% 0.2%
6.9% 6.0% 5.3% 4.7% 4.1% 3.4% 2.9% 2.4% 1.8% 1.3% 0.8% 0.4% 0.3% 0.2% 0.1%
18.1% 13.3% 11.7% 10.6% 9.4% 7.9% 6.8% 5.8% 5.1% 4.2% 4.7% 2.0% 1.4% 1.0% 0.9%
Model 4.21b:
Reference Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
High Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
Low Oil Price Case
Short Run Rebound
Dynamic Rebound
Long Run Rebound
— Calc
2000
0.2%
4.1%
0.9%
0.2%
5.5%
0.9%
0.2%
2.9%
0.9%
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
0.0% 0.0% 0.1% 0.4% 1.1% 1.4% 1.6% 2.1% 0.5% 1.1% 1.3% 1.2% 1.3% 1.3% 1.2% 1.1% 1.0% 0.9% 0.9% 0.7% 0.6% 0.5% 0.4% 0.4%
4.7% 5.1% 5.2% 5.2% 5.3% 5.1% 5.5% 5.6% 5.4% 5.4% 5.5% 5.3% 5.1% 4.8% 4.3% 3.8% 3.3% 2.8% 2.3% 1.8% 1.4% 1.0% 0.7% 0.5%
0.1% 0.0% 0.4% 2.2% 6.5% 8.4% 9.5% 13.0% 3.0% 6.4% 7.4% 7.1% 7.9% 7.7% 7.1% 6.3% 5.9% 5.4% 4.9% 4.2% 3.3% 2.9% 2.2% 1.9%
0.0% 0.0% 0.1% 0.4% 1.1% 1.4% 1.6% 2.1% 0.5% 1.1% 2.2% 2.7% 2.9% 2.9% 2.9% 2.8% 2.8% 2.8% 2.7% 2.6% 2.5% 2.3% 2.2% 2.0%
6.5% 7.4% 8.1% 8.7% 9.4% 9.9% 11.1% 10.7% 11.3% 11.8% 12.9% 13.1% 12.9% 12.5% 12.2% 11.7% 11.3% 10.6% 10.0% 9.4% 8.8% 8.1% 7.5% 7.0%
0.1% 0.0% 0.4% 2.2% 6.5% 8.4% 9.5% 13.0% 3.0% 6.3% 13.3% 16.4% 18.1% 18.0% 17.8% 17.0% 17.4% 16.9% 16.5% 15.9% 15.1% 13.8% 13.1% 12.2%
0.0% 0.0% 0.1% 0.4% 1.1% 1.4% 1.6% 2.1% 0.5% 1.1% 0.3% 0.2% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
3.1% 3.1% 2.8% 2.3% 2.0% 1.3% 1.2% 1.6% 0.8% 2.5% 0.9% 0.4% 0.2% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
0.1% 0.0% 0.4% 2.2% 6.5% 8.4% 9.5% 13.0% 3.0% 6.3% 1.6% 0.7% 0.4% 0.2% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
Appendix page 9
-------
Appendix D. Projections from model with structural break in 2003
35.0%
Projections of Rebound Effect:
Four-equation model estimated on 1966-2009 data with structural break
in 2003: projections assume break remains through 2035
2005 2010 2015 2020
2025
2030
2035
•Short run
•Dynamic
• Long run
Projections of Rebound Effect:
Four-equation model estimated on 1966-2009 data with structural break
in 2003: projections assume break is "turned off" starting 2010
oc no/
oO.U /o
on no/
oU.U /o
oc no/.
on no/.
1 O.Uvo
1 n no/.
5n%
Ono/
\
U-^
A " ^-^^
•v_ ^*"*-A-
^^"""M 1 1 ^^^^t""^
******^^*^^
S**>^a. ^^^^A ^i
— »— Short run
— •— Dynamic
—A— Long run
.U yo i i i i i • • • Y
2005 2010 2015 2020 2025 2030 2035
Appendix page 10
-------
EPA's Response to Peer Review
Comments
-------
The reviewers expressed overall support for the methodology described in the report, as well as support
for the implementation of the methodology to derive empirical estimates of the rebound effect for
passenger vehicles. Gillingham: "This is a thoughtful and careful effort aiming to address a difficult
question...tackles a difficult question using what is likely the best data publicly available...provides
estimates that appear to be reasonable...provides a valiant (and reasonable) attempt at forecasting the
VMT [vehicle miles traveled] rebound effect forward...It would be difficult to do much better given the
task at hand." Greene: "The S&H [Small and Hymel] analysis is very well done, uses appropriate models,
data and econometric methods...makes several important contributions to knowledge of the rebound
effect...results are consistent with both the central tendency of other estimates in the literature and
with the best studies in the peer-reviewed literature... projected rebound effects are useful and
plausible...results are useful to EPA as they now stand." Sallee: The Small/Hymel report "...uses an
appropriate methodology and defensible assumptions...Where I do disagree...! believe that the
preference of one method or specification over the other involves an element of subjective judgment
about how to weigh the costs and benefits of different approaches...! did not identify any issues that I
believe are objectively incorrect"
Comment: Gillingham commented that in the task given to the authors, the definition of VMT rebound
effect was vague and required that certain assumptions be made about how the adoption of fuel
efficient technologies will influence other vehicle characteristics. Gillingham noted how an increase in
vehicle price or a trade-off leading to less desirable vehicle characteristics could lead to fewer vehicles in
the fleet and a reduction in VMT. Gillingham also concludes that the authors' assumption that vehicle
price and other attributes remained fixed while vehicle efficiency improved would result in a slight
overestimate of the rebound effect.
EPA Response: The Small and Hymel methodology accounts for the VMT rebound effect from two
pathways. The first is the increase in the average fuel economy of the vehicle fleet, and that in turn
reduces the cost per mile of travel. The second is that the size of the vehicle fleet may increase because
vehicles are now more useful, in the sense that they can be driven more cheaply. This change in vehicle
fleet size may further affect the amount of overall driving. Empirically, they find that the first path is by
far the dominant one, so that one could ignore the second pathway as an approximation. The full effect
on vehicle sales and fleet size will also be influenced by any change in vehicle prices due to regulation.
This effect on fleet size would likely work in the opposite direction to that arising from a change in fuel
cost: if regulations result in manufacturers raising vehicle prices along with reduced fuel costs, those
higher prices would tend to mitigate any tendency for the size of the fleet to increase.
EPA agrees that there are potentially other aspects of our rule that may affect VMT besides the rule's
impact on fuel costs, which is the focus of the Small and Hymel report. For example, effects on vehicle
sales, due to changes in price, fuel economy, or other vehicle attributes could affect total VMT. In EPA's
analysis of the effects of greenhouse gas (GHG) standards on light-duty vehicle sales, it sees
counterbalancing forces: sales may increase due to improved fuel economy, but may decrease due to
increased costs. The agency has not concluded whether the net effect on vehicle sales will be positive
-------
or negative. In turn, vehicle sales impacts could affect rates of vehicle scrappage. Given the
complications in just assessing the directional impact of EPA standards on vehicle sales, focusing on the
direct effects of reduced fuel costs on VMT seems like a reasonable approach.
Comment: Greene noted that both 3-equation and 4-equation versions of the model do not consider
usage-induced capital depreciation, which will increase as efficiency technologies are adopted that
increase the price of the vehicle.
EPA Response: We are not sure of the causal linkage between higher vehicle prices and usage-induced
depreciation of vehicles. Furthermore, there is little evidence that drivers, in general, take capital cost
into account in decisions relating to amount of driving. Most choice models, for example those used in
practical urban transportation planning, assume that the cost variable affecting consumers' decisions
excludes capital cost entirely. In any case, we don't believe that this effect would have much
quantitative impact on VMT rebound estimates.
Comment: Greene and Gillingham commented that aggregate state input data may mask some of the
heterogeneity in the rebound effect, such as VMT rebound elasticities that vary by vehicle age.
EPA Response: We agree with the Greene and Gillingham that the use of aggregate VMT data does not
allow for the identification of some aspects of heterogeneity in the rebound effect. Alternative
approaches, focusing for example on travel survey data, may help address some of these heterogeneity
questions. But there are trade-offs between using more aggregate data (i.e., state level data) and travel
survey data. For example, Greene noted that, unlike travel survey data, studies based upon aggregate
data do not face the risk that observations of individual responses may fail to add up to a national-level
change. As another example, it can be difficult to control for important factors other than fuel prices and
fuel efficiency that may make one household drive more than the other (e.g., job location, after school
activities, etc.). EPA believes that the aggregate approach is appropriate for the purpose of estimating
the overall nationwide increase in driving that would result from a given reduction in fuel operating
costs.
Comment: Gillingham and Sallee raised the issue of the potential problem of measurement error with
the use of aggregate data in the report. Sallee commented that the VMT variable is not directly
measured, but is imputed based on fuel sales and estimates of fuel efficiency that may be inconsistent
across states and over time. Sallee suggested that using VMT readings from odometer readings from
smog checks may be one way to avoid the issues associated with using state level aggregate data.
EPA Response: As Small and Hymel state, there are potential problems with the VMT data collected by
the US Federal Highway Administration. These data are reported by states, which lack a uniform
methodology for estimating them. For example, some states rely on periodic vehicle counts, while
others multiply fuel consumption (measured from tax records) by an independent estimate of fleet fuel
efficiency. However, Small and Hymel do not think that these potential problems bias their results. They
posited that sources of measurement error are mostly unrelated to their independent variables. Even if
there are sources of measurement error, the use of fixed effects eliminates the spurious effect of any
cross-state relationship that is consistent over time. One might worry that errors in measuring fuel
-------
consumption by state could appear in both VMT data (in those states where the VMT estimate is based
on fuel consumption) and in fuel efficiency. This would bias OLS estimates, but not 2SLS and 3SLS, which
are specifically designed to eliminate asymptotic bias resulting from correlated errors in the dependent
variables.
We agree with Sallee that smog check odometer travel data would be another source of information
about travel behavior that could be useful in attempting to estimate VMT rebound effects. We also note
that Gillingham has undertaken work estimating the VMT rebound effect with California smog check
data. The Gillingham study is quite useful in providing another "data point" to assess VMT rebound
effects. However, there are trade-offs between using more aggregate data (i.e., state level data) and
disaggregate data (e.g., travel survey data, or smog check odometer readings). For example, as
mentioned above, Greene noted one advantage of using aggregate data for estimating the rebound
effect is that it covers the totality of vehicle travel. Unlike travel survey data, studies based upon
aggregate data do not face the risk that observations of individual responses fail to integrate to national-
level change. Also, only a limited amount of smog check data is available for specific states or cities,
which limits the usefulness of this approach to obtaining national estimates of VMT rebound.
Comment: Gillingham noted that while the report assumed no measurement error, if classical
measurement error in the regressors existed, the result would be a downward bias of the coefficients
(i.e. attenuation bias).
EPA Response: Attenuation bias is a common problem in econometric studies. The reviewer is correct
that measurement error in the regressors might lead to a downward bias in the coefficients estimated.
This is one of several sources of uncertainty in the results.
Comment: Gillingham commented that linear time trends are used instead of a standard panel data
approach with fixed time effects, and that if fixed time effects had been used, the model would control
for other changes more flexibly. Sallee commented similarly that there was apparently no attempt to
control for correlation across states in the error terms, and suggested clustering standard errors by year.
EPA Response: These comments raise the questions on how the time trend should be better
represented in the model and what the impact is on the standard errors if the time trend is treated
differently in the model. Due to limitations of the software used to estimate the model in the report, the
authors were not able to compute clustered standard errors along with the other two time-related
effects it takes into account, namely autoregressive errors and a lagged dependent variable. However,
subsequent to this report, the authors conducted two experiments to see if clustering the standard
errors makes a difference. First, they estimated a slightly different model, identical except omitting the
time trends. They also estimated this model with a different estimator, Generalized Method of
Moments (GMM), which was also used for a comparison in the Small and Van Dender paper. The
comparison was just to see whether this different method gave different results, which for the most
part it did not.
-------
Using this slightly modified model, standard errors were computed both with and without clustering.
The clustering was done at both at the state level, as the authors initially thought correlations across
time would be the main problem with standard errors, and also at the level of a year, as suggested by
the commenter. The results showed virtually no change in standard errors. This is somewhat surprising,
but with two time-related correlations already handled in the model (autoregressive errors and lagged
dependent variables), it is difficult to develop trustworthy intuition about what to expect from a time-
related clustering calculation.
The authors do not think the failure to cluster during the calculation of standard errors makes any
significant difference, and in particular they cannot find any evidence that the reported standard errors
are understated as a result. A more thorough description of these experiments is described in their
working paper.39
Comment: Gillingham suggested that when relying on time series variation over many years, testing for
autocorrelation and unit roots is a common approach, and noted that while the authors did consider 1st
order autocorrelation, that second-order order autocorrelation was not considered.
EPA Response: In the judgment of the authors, the time series variation in the data was too limited to
make it likely that both first-order and second-order correlation could be accurately measured.
Measuring first-order autocorrelation already is a major advance over much of the literature.
Comment: Sallee noted that panel identification may introduce the problem of omitted variable bias if
there are other factors that are correlated with gasoline price and VMT per adult, such as personal vs.
work driving, the quality of automobiles, commuting norms, fraction of two-earner families, expansion
of urban sprawl and that other factors that may also be correlated with VMT. Along this same line of
argument, Sallee noted that the price of gasoline is the most important variable in the analysis, and
adding time period fixed effects would remove the vast majority of variation in gasoline prices due to
fluctuations in global oil price. The remaining state-specific variations in price would be the result of
short term imbalances in supply and demand, and therefore may have limited impact on behavior. Thus,
Sallee suggested adding time dummies to distinguish between periods where there may be structural
breaks.
EPA Response: Using dummy variables for years better controls for changes over time for factors that
Sallee raises (e.g., quality of cars, commuting norms, etc), whereas a linear time trend will not be as
effective. But reducing omitted variable bias with fixed effects comes at a cost. As Sallee mentions, the
problem is that every time you add in a fixed effect, you are removing some kind of variation from the
data. Having both time and state dummies would mean that you are using only variation in prices within
a given state in a single year. And if most of the variation is coming from national trends, or fluctuations
Hymel, Kent, and Kenneth A. Small, "The Rebound Effect for Automobile Travel: Asymmetric Response
to Price Changes and Novel Features of the 2000s," Working paper 14-15-03, UC Irvine (May 2014).
http://www.economics.uci.edu/files/economics/docs/workingpapers/2014-15/14-15-03.pdf
-------
in the global oil price, then time fixed effects will remove that variation, leaving you with very little to
identify the coefficients.
At an earlier stage of the research the authors attempted to estimate a model with individual dummies
for each year. The result was very imprecise coefficient estimates, and sometimes failure of the iterative
nonlinear estimation routine to converge. It is for this reason that this approach was not included in the
results. An additional reason to forego year dummies is the possible anomalous causes of year-to-year
changes in state-level fuel prices, as noted by this same commenter. As for more complex time trends,
the authors did try a number of time-trend variables with structural breaks. No significant breaks were
found in time trends in the VMT equation, but there were identifiable breaks in the equation for fuel
intensity, resulting in the use of three time trend variables in the latter equation. This is not mentioned
in the text of the paper but can be seen in the detailed appendix results.
Consideration of uncertainties
Comment: All of the reviewers suggested that more treatment of uncertainty would be useful. Sallee
suggested that a fuller way of representing forecast and coefficient uncertainty would be to model the
uncertainty in the forecasted variables and provide a collection of different model results based on
random draws of these variables. If this was done, he suggested, it would make clearer which
parameters are pivotal, so users know where to focus their attention.
EPA Response: Incorporating uncertainty is a difficult challenge that has not been given much attention
in the literature on the VMT rebound effect. There are many different types of uncertainties: (1)
uncertainties due to data shortcomings, (2) issues with the experimental design available in the
historical record, (3) uncertainties due to model formulation, (4) uncertainties inherent in econometric
estimation, and (5) uncertainties about the future state of the world. Small and Hymel attempt to
address many of these issues by constructing alternative projections based on different assumptions.
While a formal uncertainty analysis might be useful to undertake in the context of the VMT rebound
issue, such an effort would be a significant and complex task in its own right. Given the complexities
associated with undertaking this type of analysis, a formal uncertainty analysis is beyond the scope of
this current effort.
Finding of asymmetric response
Comment: All reviewers believe Small and Hymel use a well-established approach to account for
asymmetric responses to increases/decreases in per mile fuel costs based on variation in fuel prices.
Greene and Sallee believe that there is sufficiently strong evidence of an asymmetric response in the
paper and the literature to use a model that allows for this difference, and agree with Small and Hymel's
decision to use 3.21b and 4.21b as their preferred model specifications. Gillingham, however, believes
that the saliency of gasoline prices may be different than the saliency of fuel price per mile, so he would
be more comfortable using results that assume a symmetric response.
EPA Response: Small and Hymel find a significant asymmetric VMT response to fuel cost increases and
decreases based upon fuel price changes. While we agree with Gillingham that the saliency of gasoline
-------
prices may be different than the saliency of fuel costs, Greene and Sallee suggest that an asymmetric
VMT response to fuel costs based on fuel price increases and decreases (i.e., models 3.21b and 4.21b)
seems reasonable in Small and Hymel's work. Small and Hymel suggest in their report that their
estimates of the impact on VMT from fuel costs may actually measure the response to changes in fuel
price rather than fuel efficiency because they are unable to find a statistically significant influence on
VMT from fuel efficiency alone. For this reason, as well as endogeneity concerns discussed in their
report, they prefer the model that captures asymmetry based on fuel price increases vs. decreases
(3.21b, 4.21b) rather than fuel costs (3.23 and 3.24). In models 3.21b and 4.21b, an increase in fuel
economy (which would by itself reduce fuel costs) behaves like a decrease in fuel price, with a smaller
response than when the fuel price increases. It should be noted that they also found a statistically
significant difference in the impact on VMT from fuel cost increases and decreases, too (models 3.23
and 3.24).
Appropriateness of dynamic rebound to account for variables that
change over vehicle lifetime
Comment: All three reviewers agree that the dynamic rebound effect (i.e., which accounts for how
income, congestion, and other variables that influence vehicle travel vary over time) is a useful way to
summarize the rebound effect through time. They also agree that the dynamic rebound effect should be
used to quantify the rebound effect over the period of a vehicle's lifetime.
EPA Response: We agree with the reviewers that the dynamic rebound effect is a useful summary
statistic for quantifying vehicle rebound effects over time and should be used to estimate the rebound
effect over the period of a vehicle's lifetime.
Appropriateness of methodology for projecting VMT
Comment: The reviewers are generally in agreement that there is strong evidence that the rebound
effect has changed over time and that changes are correlated with changes in income and fuel prices.
They also agree that there is a theoretical justification for including these effects, since income affects
the value of travelers' time and fuel prices affect the fuel cost share of the long-run cost per mile of
travel. Thus, they agree that it is appropriate to include these effects in the forecasting model.
Sallee suggested that one alternative to forecasting the rebound effect would be to take the best
available estimate of the rebound effect from recent years, say 2000 to 2007, and project these
estimates forward as a constant rebound effect over all future years without changing income and fuel
prices. Similarly, Sallee suggested that one way to judge the importance of the decline in the rebound
effect with income is to provide a comparison projection using a constant rebound effect.
The reviewers provided some additional considerations and recommendations regarding the
extrapolation approach. For example, Sallee raised questions about the way that fuel price volatility is
represented in the projections. Sallee and Greene suggested using a nonlinear extrapolation approach
that is asymptotic above zero. Greene commented that a linear extrapolation of the income and price
-------
effects could be improved upon by using a better functional form. Greene suggested an alternative
approach where the rebound approaches zero as income goes to infinity and fuel prices go to zero.
EPA Response: We agree with the reviewers that there is strong evidence that the rebound effect has
changed over time and that changes may be caused by changes in income and fuel prices. Thus, it is
appropriate to include these effects in the forecasting model. We believe that it is possible to judge the
importance of income without formal projections. One only needs to compare the rebound effect
estimated year by year with the estimated rebound effect for the entire time frame of the analysis to
see the impact of income on the rebound effect.
According to Small and Hymel, they attempted in earlier phases of this research to estimate a model
with a built-in nonlinear response that tends toward an asymptote of zero rebound (when incomes are
very high), but were unsuccessful. The procedure currently used truncates the rebound effect at zero
state by state, and has the effect of making the aggregate rebound nonlinear with a zero asymptote.
This seems like a reasonable way to project the rebound effect into the future when, as happened here,
using a nonlinear form ended up being intractable to implement. While the truncation procedure is
undoubtedly inaccurate at a fine level of detail (e.g., a given state in a given year), the errors are likely to
average out and so it can produce a satisfactory aggregate analysis. If an asymptotic value above zero is
used as suggested by Gillingham and Sallee, other issues would need to be addressed. For example,
what value should be chosen and what is the basis for the new chosen value? The choice of a positive
asymptotic value may be considered arbitrary and difficult to defend.
------- |