Consumer Willingness to Pay for
Vehicle Attributes:
What is the Current State of Knowledge?
oEPA
United States
Environmental Protection
Agency
-------
Consumer Willingness to Pay for
Vehicle Attributes:
What is the Current State of Knowledge?
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
Prepared for EPA by
RTI International
EPA Contract No. EP-C-16-021
Work Assignment No. 1-10
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.
NOTICE
&EPA
United States
Environmental Protection
Agency
EPA-420-R-18-016
July 2018
-------
ABSTRACT
As standards for vehicle greenhouse gas emissions and fuel economy have become more
stringent, concerns have arisen that the incorporation of fuel-saving technologies may entail
tradeoffs with other vehicle attributes valued by consumers including safety, comfort, and
performance. To the extent that such interactions are present, they may influence the rate of
consumer acceptance of fuel-saving technologies. Understanding and quantifying such
interactions, both positive and negative, is important for transportation policy analyses. Not only
will these estimates provide a better understanding of the role of fuel-saving technologies in
consumers' evaluation of new vehicles, and in consumer purchase decisions, but they will also
enable a better estimate of policy impacts on overall household welfare. Given the potential
importance of accounting for consumer willingness to pay (WTP) for changes in vehicle
attributes when conducting policy analyses, we conduct a detailed review and analysis of
literature that presents or can be used to calculate WTP for vehicle attributes in order to assess
the current state of knowledge in this area. We identified 52 relevant U.S.-focused papers
published since 1995 (with one exception) with sufficient data to calculate WTP values. We
identify 142 individual characteristics considered in the literature, which we consolidate into the
15 general categories of comfort, fuel availability, fuel costs, fuel type, incentives, model
availability, non-fuel operating costs, performance, pollution, prestige, range, reliability, safety,
size, and vehicle class. We then calculate WTP values for those characteristics based on the
coefficients and data reported in the papers. In addition to central tendency WTP estimates, we
present indicators of variability around each WTP value, based either on standard errors of the
estimated coefficients or the standard deviations in random coefficient models. We also examine
the implications of heterogeneous consumer characteristics (e.g., different levels of income,
household size, and other factors). Our findings suggest large variation in WTP values for
vehicle characteristics, both within and across studies. This variation may result in part because
of methodological difficulties in estimating how attributes affect consumer vehicle choices, such
as omitted variables, errors in variables, collinearity, and the use of proxies. We discuss the
implications of this variation in WTP estimates for estimating changes in consumer demand due
to a change in fuel efficiency technology.
Keywords: Consumer preference, fuel efficiency, vehicle demand, willingness to pay
JEL Codes: D12, 033, Q52, R40
in
-------
CONTENTS
Section Page
1. Introduction 1-1
2. Literature Review 2-1
3. Description of Studies and Attributes Analyzed 3-1
4. Methodology 4-1
4.1 Estimating Central Tendencies of WTP 4-1
4.2 Measuring Preference Heterogeneity and Estimation Uncertainty 4-5
5. Willingness to Pay for the Attributes of Vehicles 5-1
5.1 Comfort Grouping 5-6
5.1.1 Automatic Transmission 5-6
5.1.2 Rear-wheel Drive vs. Front-wheel Drive 5-6
5.1.3 Air Conditioning 5-7
5.1.4 Shoulder Room 5-8
5.2 Fuel Availability 5-9
5.2.1 Recharging Time 5-9
5.2.2 Fuel Availability 5-10
5.3 Fuel Cost 5-10
5.3.1 Reduction in Fuel Cost per Mile 5-10
5.3.2 Dollars per Year 5-15
5.3.3 Gallons per Mile 5-15
5.3.4 Miles per Dollar 5-16
5.3.5 Miles per Gallon 5-16
5.4 Fuel Type 5-17
5.4.1 Electric Vehicles (EVs) 5-17
5.4.2 Hybrid Vehicles 5-17
5.4.3 Flexible Fuel 5-19
iv
-------
5.4.4 Plug-in Electric Vehicles (PHEVs) 5-20
5.4.5 Methanol 5-21
5.4.6 Natural Gas 5-21
5.5 Performance 5-22
5.6 Pollution 5-26
5.7 Range 5-26
5.8 Size 5-28
5.8.1 Footprint 5-28
5.8.2 Luggage Space 5-30
5.8.3 Weight 5-31
6. Discussion: Why the Lack of Consensus on WTP? 6-1
7. Concluding Observations 7-1
8. References 8-1
Appendices
A: Bibliography of Papers Included in our Main Sample A-l
B: Willingness to Pay Results by Attribute Grouping B-1
C: Discussion of the potential Bias from Estimating WTP from Ratios of
Attribute and Price Derivatives C-l
D: Histograms of Untrimmed Central WTP Estimates by Attribute D-l
E: Untrimmed Distributions of Central WTP Estimates by Attribute E-l
F: Author Feedback received and response to comments F-l
v
-------
LIST OF FIGURES
Number Page
Figure 3-1. Distribution of Initial Pool of Papers Considered by Year of Publication 3-2
Figure 3-2. Distribution of Papers Considered by Time Period of Data Used 3-6
Figure 3-3. Number of Observations by Attribute Grouping 3-7
Figure 5-1. Rear-Wheel Drive Preference Variation, Range is +/- 1 Standard
Deviation 5-7
Figure 5-2. Population Taste Heterogeneity for Air Conditioning 5-8
Figure 5-3. WTP for a One Hour Reduction in Charge Time Across Studies 5-9
Figure 5-4. Willingness to Pay for $0.01/mile Decrease in Fuel Cost: All Estimates
(2015$) 5-11
Figure 5-5. Willingness to Pay for $0.01/mile Decrease in Fuel Cost: Trimmed
Sample (2015$) 5-12
Figure 5-6. Range of Reduction in Fuel Cost per Mile WTP Estimates Describing
Preference Heterogeneity 5-14
Figure 5-7. Range of Reduction in Fuel Cost per Mile WTP Estimates Describing
Estimation Uncertainty 5-14
Figure 5-8. Population Taste Heterogeneity for Electric Vehicles 5-18
Figure 5-9. Distribution of Trimmed Central WTP Estimates for Hybrid Vehicles 5-18
Figure 5-10. Population Taste Heterogeneity for Hybrid Vehicles (excluding outliers).... 5-20
Figure 5-11. Natural Gas Vehicle Preference Variation: Range is +/- 1 Standard
Deviation 5-22
Figure 5-12. Frequency Distribution of WTP Estimates: Normalized 0-60 Times 5-24
Figure 5-13. WTP for One Second Decrease in 0-60 mpg Time: Preference
Heterogeneity 5-25
Figure 5-14. WTP for One Second Decrease in 0-60 mpg Time: Estimation Error 5-26
Figure 5-15. WTP for Range in $/mile: Preference Variation, +/1 One Standard
Deviation 5-28
Figure 5-16. WTP for Range in $/mile: Estimation Error, +/1 One Standard Error 5-29
Figure 5-17. Trimmed Central WTP Estimates for Vehicle Footprint 5-29
Figure 5-18. WTP for Footprint: Preference Variation, +/- One Standard Deviation 5-30
Figure 5-19. Trimmed Distribution of Central WTP for Weight 5-32
Figure 6-1. Willingness to Pay for a $1 Present Value Decrease in Operating Cost: 1-
and 2-vehicle Households (Estimates derived from Brownstone et al., 1996) 6-6
vi
-------
LIST OF TABLES
Number Page
3-1. Literature Summary Statistics Based on our Main Sample 3-3
3-2. List of Papers Included and Description of Their Data 3-4
5-1. Summary Statistics from Pooled Central WTP Estimates* 5-4
5-2. Willingness to Pay for $0.01/mile Decrease in Fuel Cost—Combined GPM and
$0.01/mile Values 5-13
5-3. Willingness to Pay for a One Second Decrease in 0-60 mph Time: Combined
0-30, 0-60, and hp/lb Normalized Metrics 5-24
5-4. Comparison of Stated and Revealed Preference Estimates of WTP for One
Second Decrease in 0-60 mph Time 5-25
6-1. Willingness to Pay for $0.01/mile Decrease in Fuel Costs from Studies Using
the Same CA Survey 6-7
6-2. Willingness to Pay for a 1 Second Decrease in 0-to-60 Acceleration Time 6-8
6-3. Willingness to Pay for Alternative Fuel Availability Equivalent to Gasoline 6-9
6-4. Willingness to Pay for Reducing the Emissions of a Typical Gasoline Vehicle to
Zero 6-10
vii
-------
SECTION 1.
INTRODUCTION
As standards for vehicle greenhouse gas emissions and fuel economy become more
stringent, vehicle design modifications made to achieve environmental goals could potentially
impact other vehicle attributes valued by consumers including noise, safety, comfort, and
performance. To the extent that such interactions are present, they may influence the rate of
consumer acceptance of fuel-saving technologies. For instance, some analysts have argued that
consumers undervalue fuel savings and therefore underinvest in technologies that improve fuel
economy, but one possible explanation for consumer adoption patterns deviating from market
projections is that there are interactions with other vehicle attributes that consumers are
considering. Understanding and quantifying such interactions, both positive and negative, is
important for transportation policy analyses.
The research presented in this report has three main objectives.
Survey the econometric literature to identify the vehicle attributes for which estimates
of marginal willingness to pay (WTP) can be computed.
¦ Derive central tendency estimates of WTP for as many attributes as possible.
¦ Produce summary statistics describing the distribution of WTP estimates for all
attributes, with special attention to fuel cost and performance.
Developing consensus estimates of WTP for vehicle attributes is not a goal of the research
presented in this paper. A meta-analysis of WTP for fuel cost and performance using the
estimates developed in this study has also been completed (Greene et al., 2018).
This exploratory analysis of estimates of WTP for vehicle attributes is intended to
provide a better understanding of the role of vehicle attributes in consumers' evaluation of new
vehicles and in consumer purchase decisions, to eventually enable a better estimate of policy
impacts on overall household welfare when vehicle attributes change in response to a policy.
Given the potential importance of accounting for consumer WTP for changes in vehicle
attributes when conducting policy analyses, we conduct a detailed review and analysis of
literature that presents or can be used to calculate WTP for vehicle attributes in order to assess
the current state of knowledge in this area. We identified 52 relevant U.S.-focused papers
published since 1995 (with one exception1) with sufficient data to calculate WTP values. We
1 We retain Lave and Train (1979), the first application of a multinomial discrete choice model to automobile
choice, as a useful comparison point despite its publication year falling outside our primary restriction criteria.
1-1
-------
identify 142 individual characteristics considered in the literature, which we consolidate into the
15 general categories of comfort, fuel availability, fuel costs, fuel type, incentives, model
availability, non-fuel operating costs, performance, pollution, prestige, range, reliability, safety,
size, and vehicle class. We then calculate marginal WTP values for those characteristics based on
the coefficients and data reported in the papers.
Our method follows the first four steps of the procedure for meta-analysis of WTP data
recommended by Van Houtven (2008):
¦ Problem formulation: specifying research objectives and defining the scope of the
analysis,
¦ Data collection: via a formal literature search,
¦ Data evaluation and abstraction: insuring that the WTP are valid and acquiring them
along with descriptors (e.g., units) and study attributes,
¦ Data preparation: standardization of WTP and potential explanatory variables in
constant dollars and units to the extent possible.
The final two steps, data analysis and presentation of results, will be accomplished in a separate
paper focusing on fuel economy and performance (Greene, et al., 2018).
We limit the scope of our analysis to U.S. studies published between 1995 and 2015, with
the sole exception of Lave and Train (1979), the first use of a random utility model to vehicle
choice. We consider only U.S. studies because there are a sufficient number of them and because
our goal is to inform U.S. policy making. Introducing results from other countries with different
vehicle choices, consumer preferences, and government policies would require an analysis of the
impacts of those differences on the WTP estimates. In addition, consumers' preferences can
change over time. Focusing on more recent studies is intended to make our analysis more
relevant to current policy making. We focus on peer-reviewed studies but also include a smaller
number of studies from the grey literature, a procedure recommended for meta-analyses to
reduce publication bias (Van Houtven, 2008, p. 904). By means of a structured literature search
(described in Section 2), we identified 52 U.S.-focused papers with sufficient data to calculate
WTP values for various vehicle attributes. Within papers, we include all estimation results
presented unless they are identified by the authors as incorrect or erroneous. We do not include
only the authors' preferred model if alternatives are considered plausible. We do this to reduce
confirmation bias. For all plausible models we include all attribute estimates, statistically
significant or not, because the values of all parameter estimates are interdependent and because a
finding of statistical insignificance can also be meaningful.
1-2
-------
This report describes the central tendencies of the WTP estimates derived from the
literature. In addition to central WTP estimates, we present estimates of variability around each
WTP value. These ranges are based either on standard errors of the estimated coefficients or the
standard deviations in random coefficient models for models in which WTP depends on income.
The WTP estimates exhibit large variation in the implied values for vehicle characteristics, both
within and across studies. This variation may result in part because of methodological difficulties
in estimating how attributes affect consumer vehicle choices, such as omitted variables, errors in
variables, collinearity, and the use of proxies where the exact variables that the authors would
ideally like to include are not available. We discuss the implications of this variation in WTP
estimates for estimating changes in consumer demand due to a change in fuel efficiency
technology. A meta-analysis of the variability of WTP estimates is in Greene et al. (2018).
This report revises the Final Report prepared under previous contract EP-C-11-045, work
assignment 4-11. Since that report was written, we contacted the authors of the studies
comprising our main sample used in this report to ask for their feedback on our use of their study
results. As detailed in Appendix F, we received feedback from authors on 36 of the 52 papers in
our main sample. Responses for 20 of those papers either indicated agreement with our
calculations or suggested we verify certain calculations, which we did but that verification
resulted in no changes. For the other 16 papers where we received author feedback, we made
revisions to our calculations in response. The authors of two additional papers made suggestions
for additional clarification within the report that we incorporated. This report thus supersedes the
Final Report from Work Assignment 4-11.
The following section provides a brief overview of the econometric literature on
consumers' choices of vehicles and preferences for their attributes. This is followed by a
description of the studies analyzed and the attributes for which WTP estimates could be derived
in Section 3. Our methods for estimating WTP using coefficients and other information available
from the studies in our sample are described in Section 4. Descriptive statistics and analysis of
the results for the most prevalent attributes are presented in Section 5. Our discussion in Section
6 focuses on additional analyses of five specific studies within our data set that provide special
insights into the great variability of WTP estimates found in the literature. The studies present
varying results from the same database using different model formulations or estimation
methods. Finally, in Section 7 we conclude by reflecting on possible explanations for the
divergence of WTP estimates found in the literature and offering some recommendations for
future research.
1-3
-------
SECTION 2.
LITERATURE REVIEW
The econometric literature dealing with consumers' vehicle choices is extensive and rich
in terms of data sources, models and estimation methods. Vehicle choice models based on the
attributes of vehicles and consumers have their origins in economic theories and models
developed in the latter half of the 20th century. The theory that consumers desire the attributes of
goods and not the goods themselves and that a single good generally possesses multiple
attributes was proposed fifty years ago by Lancaster (1966). Among the earliest applications of
the new theory of consumer demand was an effort to predict choice of mode of transportation
based on the attributes, speed, frequency of service, comfort and cost (Quandt and Baumol,
1966). The new theory of consumer demand led to empirical efforts to estimate hedonic demand
equations, models for predicting consumers' willingness to pay for goods as a function of their
attributes (Rosen, 1974). Hedonic price modeling has also been applied to correcting price
indices for changes in the quality of goods over time (Grilliches, 1971). McFadden (1974)
applied the theory of demand for attributes to modeling consumers' choices among discrete
modes of transportation. Consumers were assumed to base their choices on indirect utility
functions comprised of an observable function of the attributes of the choices and of the
consumers and an unobservable random utility component. By specifying the distribution of
random utility as a type I extreme value distribution, McFadden derived the multinomial logit
model, variations of which still dominate the literature today. The first application of the
multinomial logit discrete choice model to automobile choice appeared in 1979 (Lave and Train,
1979). Lave and Train's model predicted consumers' choices among ten vehicle classes using
data from a survey of new car buyers in seven U.S. cities. The first estimation of an automobile
choice model using market shares data was a random coefficient model developed by Cardell et
al. (1977). Over the past 35 years, formulations of discrete choice models applied to vehicle
choices have increased in number and complexity. Methods have been developed for estimating
discrete choice models using market sales data (Berry et al., 1995) and for estimating models
from survey data with random coefficients to reflect variations in consumers' valuation of
different attributes (McFadden and Train, 2000). These theoretical and methodological
developments have engendered an extensive published literature that provides a rich resource for
analyzing consumers' willingness to pay for vehicle attributes.
Tardiff (1980) reviewed the earliest efforts to apply discrete choice models to automobile
choice in a special issue of Transportation Research devoted to automobile choice and its energy
implications. The earliest applications were efforts to predict the number of vehicles households
would own and their choice of transportation mode. Attributes typically included only the price
2-1
-------
of an automobile and such modal characteristics as travel times for the journey to work. The first
application of a random coefficient model to the choice of type of automobile that we identified
was Beggs and Cardell (1980), an analysis of consumers' likelihood of purchasing an electric
car. A number of studies published in the early 1980s extended Lave and Train's (1979) initial
multinomial logit (MNL) model of choice among vehicle classes to predict choices among
individual makes and models of vehicles and to represent consumers' decisions about how many
vehicles of which types to own or whether or not to purchase a vehicle. These efforts led to the
development of the nested multinomial logit (NMNL) model in which the choice of type of
vehicle was "nested" within the choice of how many vehicles to own. All of these models
estimated trade-offs between vehicle attributes and vehicle price, enabling the calculation of
marginal willingness to pay for various vehicle attributes.
Potoglou and Kanaroglou (2008) provide an overview of the more recent discrete choice
modeling literature as applied to households' automobile choices. The review covers models of
car ownership, vehicle type choice, as well as models of vehicle holdings and transactions.
During the 1980s the NMNL model came to be preferred by researchers over the simple MNL
model because of its ability to represent more flexible choice structures involving a larger
number of alternatives. Mixed Multinomial Logit (MMNL or MXL) models with random
coefficients representing heterogeneous preferences for vehicle attributes can approximate any
random utility model but must generally be estimated by numerical approximation or simulation.
Methods for estimation of random coefficient models from survey data were further developed
by McFadden and Train (2000) and Train (2009) and from vehicle sales data by Berry et al.
(1995). Because of differences in estimation methods and type of data used (individual survey
responses for MXL versus aggregated market sales data for the Berry, Levinson and Pakes, BLP,
method) we make a distinction between BLP and MXL models. Random coefficient models
greatly increased the potential to represent heterogeneous consumer preferences and more
complex preference structures. Not only could the means and standard deviations of coefficients
be estimated but also correlations among preferences.
While the econometric literature on vehicle choice is rich in terms of theory and
methodology, evaluations of the coefficient estimates and predictive ability of vehicle choice
models is relatively scarce. Haaf et al. (2014) observe that the bulk of the vehicle choice
literature is focused on explanation rather than prediction. Model validity is primarily judged by
goodness of fit measures and statistical significance and signs of coefficient estimates. However,
models that fit existing data best may not be best for prediction. Coefficients may be biased due
to misspecification, omitted variables or errors in variables or may be sensitive to overfitting
noise in the data instead of the signal. There is some evidence that this may be the case with
2-2
-------
vehicle choice models. In a study that appears to be unique in the literature, Haaf et al. (2014)
fitted 8,993 discrete choice models, including MNL, NMNL and MXL, to aggregate US sales
data for 2004-2006 using variables commonly included in models in the peer-reviewed literature
and choosing model formulations based on objective measures of goodness of fit to the within
sample data. They found that none of the models could outperform a static model which
predicted that market shares in 2007 would be equal to those of the most recent year in the
estimation data. Berry et al. (1995) similarly concluded that their random coefficient model, also
estimated on aggregate sales data but including some aggregate consumer data, had limited
ability to predict future market behavior. They report that it predicted market shares and new
vehicle average fuel economy well for the first two forecast years but that once new makes and
models with different attributes began to be introduced, the model's predictions ".. .became
markedly worse and deteriorated further over time" (Berry et al., 1995, p. 886).
Potoglou and Kanaroglou (2008) note that the early use of revealed preference data to
estimate consumers' likelihood of choosing alternative fuel vehicles (AFV) was problematic
because actual choices of AFVs were either rare or nonexistent in the marketplace. This led
researchers to develop stated preference (SP) surveys in which choices could be presented to
respondents using a structured experimental design, and the information given could be carefully
controlled. But SP surveys also had limitations. Most respondents had no direct experience with
the attributes of AFVs making their assessment of their values potentially unreliable. And, like
other surveys, SP surveys are susceptible to a variety of response biases, including "yea-saying"
in which respondents tend to give answers they believe are the ones wanted or social desirability
bias which can make respondents more likely to exaggerate their desire to purchase a low-
polluting vehicle. Indeed, early studies predicted a substantial willingness to purchase AFVs that
did not materialize in the marketplace. Hidrue et al. (2011) note that studies of electric and other
AFV choice based on SP survey data indicated a substantial willingness to pay to reduce
emissions and to save on fuel. The nature of survey response biases is such that they are likely to
affect certain types of willingness to pay estimates more than others. Combining SP and RP data
to estimate choice models has been proposed as a potential solution or means of ameliorating
response bias. In general, actual sales data are used to formulate constraints (moments) to be met
by the estimation algorithm. Although this method has merit, it is also limited by the degree to
which the RP data contains relevant information.
The recent literature includes many studies that model consumers' willingness to
purchase AFVs based on SP survey data. A large fraction aim at providing insights into the
market for electric drive vehicles. Tanaka et al. (2014) summarized the attributes included in 21
choice models focused on AFVs. All but one included purchase price, all included some measure
2-3
-------
of fuel cost, fifteen included measures of performance of which the most common was
acceleration. As noted by Hidrue et al. (2011), studies done in the 1990s and later include more
attributes specific to battery electric vehicles, such as emission reductions, refueling time and the
opportunity for home refueling. Although motorists are thoroughly familiar with the acceleration
performance of conventional internal combustion engine vehicles, the acceleration of an EV is
qualitatively different. Electric motors deliver almost full torque from 0 rpm and therefore
accelerate much more quickly from a full stop than an internal combustion engine vehicle.
Among other attributes of special relevance to EVs included in the studies were range (14), fuel
availability (12), emissions reduction (11) and fuel type (7). Again, while motorists are familiar
with the effect of range on the frequency of refueling, few have any experience with a vehicle
that takes hours rather than minutes to refuel but can conveniently be refueled at home. Drivers
of gasoline vehicles also lack experience with fuel availability as scarce as 1% to 10% that of
gasoline. This general lack of direct experience with novel vehicle technologies makes
interpretation of WTP estimates for attributes of AFVs uncertain.
Dimitropoulos et al. (2013) presents a meta-analysis of 33 SP studies that estimated WTP
for vehicle range based on surveys conducted between 1978 and 2011. WTP estimates varied
widely but the authors concluded that consumers were willing to pay, on average, between $66
and $75 (2005$) for a 1-mile increase in driving range. The distribution of estimates was
positively skewed, with a median value of $55 and a range of $8 to $317. The authors present
95% confidence intervals of $49 to $84 (unweighted), $48 to $101 (weighted by observations per
data set) and $29 to $104 (weighted by observations per data set and study sample size). The
meta-analysis produces several inferences concerning the effects of methods and study design.
Studies employing random coefficient models assuming log-normal distributions for both
purchase price and driving range produced much higher WTP values than other methods. Studies
that focused exclusively on BEVs, not including other types of alternative fuel vehicles,
produced higher estimates of WTP for range. In general, WTP for range was lower for studies
that included longer driving ranges. Studies that included the option of fast-charging for EVs
produced lower WTP estimates. Finally, US-based studies produced higher WTP values than
EU-based studies.
Dimitropoulos et al. (2013) point out two important shortcomings of existing studies. In
theory, the value of range should decrease at a decreasing rate with increasing range. However,
researchers generally formulated models that assumed a constant value per mile of range.
Consistent with this, the levels of driving range considered in a study were found to have an
important impact on WTP estimates. In addition, the value of range should not be independent of
the time required to refuel, a particularly important consideration for battery electric vehicles.
2-4
-------
Again, general practice is to estimate WTP for range independent of refueling time. Only three
studies considered the dependence of WTP for range on refueling time. These shortcomings
undoubtedly contributed to the conclusion that consumers would pay $3,800 (mean value) for an
increase in vehicle range from 100 to 150 miles and $17,200 for an increase from 100 to 350
miles. The median WTP values for such increases were $3,200 and $13,100, respectively.
Massiani (2013) makes similar criticisms of existing SP surveys of consumers' preferences
relative to electric vehicles. Additionally, he points out that concepts such as limited public
refueling availability or the convenience of home refueling may not be well understood by
respondents because they lack relevant experience. The potential role of an EV as a second or
third vehicle in a household portfolio also may affect preferences for different attributes but is
rarely considered, nor are relevant EV-specific factors such as the ownership of a garage.
Models embodying Lancaster's theory of consumers' demands for attributes of goods
have increased in mathematical complexity over the years, along with increasingly sophisticated
estimation methods. At the same time, increasingly diverse and detailed data sources have been
developed. Models of consumers' vehicle choices have generally been developed to explain
behavior or for policy analysis. Little attention has been given to model validation, either in
terms of predictive accuracy or the general plausibility of WTP values implied by model
parameters. The predictive accuracy of models is rarely reported in the literature. The few such
evaluations available indicate poor predictive ability. Researchers observed that revealed
preference data presented serious challenges for estimating vehicle choice models: 1) high
collinearity and limited variation in vehicle attributes, 2) problems defining choice sets from the
thousands of makes, models, drivetrain and trim configurations and, 3) uncertainty about the
attributes of greatest interest to consumers and difficulty in obtaining appropriate measures.
The potential for attribute-based models of consumer demand to predict demand for
novel products inspired numerous attempts to develop such models for alternative fuel vehicles.
The absence of revealed preference data on alternative fuel vehicle choices led to the
development of stated preference surveys. Because stated preference surveys could be structured
according to a rigorous experimental design they held the promise of overcoming the statistical
challenges presented by revealed preference data. Yet stated preference data has its own issues,
especially for estimating demand for novel products. These include well known survey biases
such as yea-saying and social desirability bias. Respondents also often have difficulty expressing
coherent preferences for attributes with which they are unfamiliar.
All of this makes the usefulness of WTP estimates derived from this literature for
conducting policy analyses an open question. This assessment attempts to address that question
by deriving WTP estimates from a large set of U.S. studies conducted since 1995.
2-5
-------
SECTION 3.
DESCRIPTION OF STUDIES AND ATTRIBUTES ANALYZED
We conducted a systematic literature review for peer-reviewed publications and grey
literature from academic or research institutions that suggested relevance to the following set of
search terms. We identified literature using three different search strategies. We reviewed search
engines such as Google Scholar, Science Direct, and Econlit directly using the below search
terms. In addition to these databases, we reviewed bibliographies of relevant literature for further
sources. Finally, we ran searches on relevant economics, energy, or environment-focused
academic journals. A fourth unanticipated strategy was receiving published or working paper
suggestions through correspondence with other authors during our data processing and analysis
stages.
Search parameters:
Types of literature: 1) peer reviewed publications, 2) grey literature from academic/research institutions
Search engines: Google Scholar, Econlit, Science Direct
Sample journals: Energy Economics, Econometrica, American Economic Review, Transportation Research
(Parts A-E), Resource and Energy Economics, Review of Economics & Statistics, Transportation Review Board
Publication Years: 1980-present
Region: primarily US
Search terms: willingness to pay, WTP, demand, stated preference, revealed preference, vehicle characteristics,
vehicle attributes, automobile, design, fuel, choice
We used the search parameters above to produce an initial pool of 160 papers. Figure 3-1
shows the distribution of these studies by publication year and highlights the relative surge in
interest and research output in recent years. We then discarded papers that focused primarily on
markets outside of the US (n=46), and all but one of those that studied US markets prior to 1995
(n=34 out of the 114 that were focused on the US), leaving us with 80 papers.2 This latter
restriction based on publication year enabled our final sample to better reflect modern vehicle
design, empirical modeling strategies, and consumer preferences.
During the calculation stage, we further discarded 28 papers from the remaining sample
of 80, as they did not provide enough data to enable calculation of willingness to pay estimates,
or proved to be irrelevant upon further examination. Our final sample included 52 relevant
papers with sufficient data to calculate WTP values. Nearly all were published from 1995
onward and focused on the U.S. We refer to these 52 papers as our "main sample" (see Appendix
A for a full bibliography of these studies).
2 We retain Lave and Train (1979), the first application of a multinomial discrete choice model to automobile
choice, as a useful comparison point despite its publication year falling outside our primary restriction criteria.
3-1
-------
Figure 3-1. Distribution of Initial Pool of Papers Considered by Year of Publication
14
1 ->
1 n
Q _
a _ _
A _ _
9
z,
n 1
II
1
1
Sample Description
From our final sample of 52 studies, we calculated 777 estimates of WTP for vehicle
attributes, within which there were 142 unique attributes. As Table 3-1 details, the majority of
the estimates came from peer-reviewed literature (86.4%); only seven papers from the main
sample came from grey literature. We found a mix of data types utilized. About 58.2% of the
estimates came from survey data: 19.6% used revealed preference surveys such as the National
Household Travel Survey that reflect respondents' actual vehicle purchases (see, e.g., Liu, 2014;
Liu, Tremblay, and Cirillo, 2014); 38.6% used stated preference surveys reflecting hypothetical
choices (e.g., Brownstone et al., 1996; Brownstone, Bunch, and Train, 2000). About 29.3% came
from market data (e.g., Berry, Levinsohn, and Pakes, 1995; Haaf et al., 2014), and another 12.5%
from other sources including joint revealed preference-stated preference (RP-SP) data (e.g.,
Axsen, Mountain, and Jaccard, 2009; Hess et al., 2011) and literature summaries (Greene, 2001;
Greene, Duleep, and McManus, 2004). Notably, newer studies tended to rely more heavily on
survey data, particularly stated preference surveys, as a mode of ascertaining taste for alternative
fuel technologies. The majority of available estimates came from logit models (MNL, NMNL, or
MXL).
3-2
-------
Table 3-1. Literature Summary Statistics Based on our Main Sample
Paper count
52
Observation count
777
Unique attribute count
142
Literature type (out of 52)
Peer-reviewed
86.4%
Grey
13.6%
Data type (out of 777)
Revealed preference (RP) survey
19.6%
Stated preference (SP) survey
38.6%
Market data
29.3%
Other
12.5%
Model type (out of 777)
Hedonic demand
8.8%
Multinomial logit (MNL)
.6%
Nested multinomial logit (NMNL)
13.6%
Mixed logit (MXL)
29.3%
Berry-Levinsohn-Pakes (BLP)
6.8%
Other
11.3%
Table 3-2 describes the data sources for each paper. Most of the data sources represent
the entire U.S. market, and vehicle purchases by households predominate. Studies of new vehicle
choices are the most common but some are based on only used vehicles and several include both.
Notably, sample sizes vary by orders of magnitude. Moreover, sample sizes are not directly
comparable across data types. Recently, surveys of household vehicle purchases have become
available that include millions of records. On the other hand, studies based on aggregate market
sales have smaller sample sizes (e.g., the sum of makes and models over several years) but
represent a complete accounting of all vehicle sales. Studies exploring choices of alternative
technology vehicles, such as battery electric vehicles, are typically based on stated preference
surveys because actual sales volumes have been too small to rely on stated preference survey
data. Sample sizes for stated preference surveys range from several hundred to several thousand
respondents. Because of the lack of comparability of sample sizes across the different types of
data, we do not attempt to weight WTP estimates by sample size.
3-3
-------
Table 3-2. List of Papers Included and Description of Their Data
Survey Type
Sample
Start
End
Citation
Region
Type of Data
(preference)
Market Segment
Size
Year
Year
Allcott and Wozny, 2014
U.S.
Market
Discrete
Households, used vehicles
1,068,459
1999
2008
Axsen, Mountain, and Jaccard, 2009
Canada
Survey
RP&SP
Households, new vehicles
9,630
2006
2006
Axsen, Mountain, and Jaccard, 2009
California
Survey
Revealed
Households, new vehicles
7,344
2006
2006
Beresteanu and Li, 2011
22 MSAs
Market
Discrete
Households, new vehicles
139,382
1999
2006
Berry, Levinsohn, andPakes, 1995
U.S.
Market
Aggregate
New vehicles, market
2,217
1971
1990
Brownstone and Train, 1999
California
Survey
Stated
Households, new vehicles
4,654
1993
1993
Brownstone et al., 1996
California
Survey
Stated
Households, new vehicles
1,156
1993
1993
Brownstone, Bunch, and Train, 2000
California
Survey
RP&SP
Households, new vehicles
5,253
1993
1995
Busse, Knittel, and Zettelmeyer, 2013
U.S.
Market
Discrete
Households, new and used
1,863,403
1999
2008
Dasgupta, Siddarth, and Silva-Risso, 2007
California
Market
Discrete
Households, new luxury
15,556
1999
2000
Daziano, 2013
California
Survey
Stated
Households, new vehicles
7,437
1999
1999
Dreyfus and Viscusi, 1995
U.S.
Survey
Revealed
Households, new and used
2,986
1988
1988
Espey and Nair, 2005
U.S.
Market
Aggregate
New vehicles, market
130
2001
2001
Fan and Rubin, 2010
Maine
Survey
Revealed
Households, new vehicles
2,623
2007
2007
Feng, Fullerton, and Gan, 2013
U.S.
Market
Discrete
Households, new vehicles
9,027
1996
2000
Fifer and Bunn, 2009
U.S.
Market
Discrete
Households, new vehicles
17,627
1996
2005
Frischknecht, Whitefoot, and Papalambros, 2010
U.S.
Market
Discrete
Households, new vehicles
6,563
2006
2006
Gallagher and Muehlegger, 2011
U.S.
Market
Discrete
Households, HEVs
4,781
2000
2010
Goldberg, 1995
U.S.
Survey
Revealed
Households, new and used
20,571
1983
1987
Gramlich, 2008
U.S.
Market
Discrete
Households, new vehicles
4,820
1971
2007
Greene and Duleep, 2004
U.S.
Lit Review
Greene, 2001
U.S.
Lit Review
Haafetal., 2014
U.S.
Market
Discrete
Households, new vehicles
3,000
2004
2006
Helveston et al., 2015
U.S.
Survey
Stated
Households, new vehicles
384
2013
2013
Hess et al., 2012
California
Survey
RP&SP
Households, new vehicles
3,274
2008
2009
Hess, Train, and Polak, 2006
California
Survey
Stated
Households, new vehicles
7,437
1999
1999
Hidrue et al., 2011
U.S.
Survey
Stated
Households, new vehicles
3,029
2009
2009
Kavalec, 1999
California
Survey
Stated
Households, new vehicles
4,747
1993
1993
Klier and Linn, 2012
U.S.
Market
Aggregate
New vehicles, market
64,671
1978
2007
Lave and Train, 1979
U.S.
Market
Discrete
Households, new vehicles
541
1976
1976
(continued)
-------
Table 3-2. List of Papers Included and Description of Their Data
Survey Type
Sample
Start
End
Citation
Region
Type of Data
(preference)
Market Segment
Size
Year
Year
Liu, 2014
U.S.
Survey
Revealed
Households, new and used
8,086
2008
2009
Liu, Tremblay, and Cirillo, 2014
DC, MD, VA
Survey
Revealed
Households, new and used
4,525
2009
2009
McCarthy and Tay, 1998
U.S.
Survey
Revealed
Households, new vehicles
33,284
1989
1989
McCarthy, 1996
U.S.
Survey
Revealed
Households, new vehicles
1,564
1989
1989
McFadden and Train, 2000
California
Survey
Stated
Households, new vehicles
4,654
1993
1993
McManus, 2007
U.S.
Market
Discrete
Households, new vehicles
445
2002
2005
Musti and Kockelman, 2011
Austin, TX
Survey
Stated
Households, new and used
608
2009
2009
Nixon and Saphores, 2011
U.S.
Survey
Stated
Households, new and used
835
2010
2010
Parsons et al., 2014
U.S.
Survey
Stated
Households, new vehicles
3,029
2009
2009
Petrin, 2002
U.S.
Market
Discrete
Households, new vehicles
2,407
1981
1993
Sallee, West, and Fan, 2015
U.S.
Market
Discrete
Households, used vehicles
1,429,677
1993
2009
Segal, 1995
U.S.
Survey
Stated
Households, new vehicles
662
1994
1994
Sexton and Sexton, 2014
Colorado
Market
Discrete
Households, new and used
1,053,000
2000
2010
Sexton and Sexton, 2014
Washington
Market
Discrete
Households, new and used
1,050,000
2000
2010
Shiau, Michalek, and Hendrickson, 2009
U.S.
Market
Discrete
Households, new vehicles
1,000
2007
2007
Skerlos and Raichur, 2013
U.S.
Market
Discrete
Households, new vehicles
NA
2008
2008
Tanaka et al., 2014
U.S.
Survey
Stated
Households, new vehicles
8,202
2012
2012
Tompkins et al., 1998
U.S.
Survey
Stated
Households, new vehicles
7,800
1993
1995
Train and Weeks, 2005
California
Survey
Stated
Households, new vehicles
500
2000
2000
Train and Winston, 2007
U.S.
Survey
Revealed
Households, new vehicles
458
2000
2000
Walls, 1996
U.S.
Market
Discrete
Households, new vehicles
79
1983
1988
Whitefoot, Fowlie, and Skerlos, 2011
U.S.
Market
Discrete
Households, new vehicles
473
2006
2006
Zhang, Gensler, and Garcia, 2011
U.S.
Survey
Stated
Households, new vehicles
7,595
2010
2010
Note: SP=stated preference, RP=revealed preference, MSA=metropolitan statistical area, HEV=hybrid electric vehicle.
-------
Although all but one of the studies in our sample was published after 1994, more than a
third of the studies make use of data series that began prior to 1995 (Figure 3-2). Altogether, the
studies' data span a 45-year period from 1971 to 2015. For studies based on a single year of
survey data, the start and end years are the same. The two literature review papers are not
included in Figure 3-2.
Figure 3-2. Distribution of Papers Considered by Time Period of Data Used.
Years Covered by Data Used in Studies
2020
2015
5 2010
01
^ 2005
C
^ 2000
re 1995
IS 1990
IA
¦Si 1985
Q.
1980
1975
1970
a
11- 11
OO
SO
OO
¦ >OOO0
88
i 11 11 11 11 11 11 11 111111111111111111111111 i 11111111111
5 11 16 21 26 31 36 41 46 51
Studies Ordered by First Year of Data
o Pub.Yr.
~ End
o Start
Given the diversity of attribute measures, a significant challenge was standardizing units
and measures across studies to enable cross comparison. We initially categorized attributes
broadly into fifteen groupings, listed in Figure 3-3, for the purposes of utility and illustration (see
Appendix B for a more detailed characterization).3 These groupings are intended to represent the
quality consumers seek or assess in vehicles, via the observed attribute. For example,
acceleration time and braking distance are both measures of performance. Miles per gallon is an
example of fuel costs. Appendix B lists all attributes under each grouping. We derive the
groupings from existing taxonomies in the literature, balancing against author interpretations.
We describe methodologies for standardizing attributes for comparison further in Section 5.
3 Not all the observations identified in Figure 3-3 could be included in our summary calculations (see Table 5-1) due
to unit conversion issues that prevented direct comparison with the other measures in that grouping.
3-6
-------
Figure 3-3. Number of Observations by Attribute Grouping
Comfort 44
Fuel availability 53
Fuel costs 122
Fuel type 84
Incentives 17
Model avail, 14
Non-fuel op costs ¦ 5
Performance 101
Pollution 19
Prestige 32
Range 40
Reliability m 6
Safety L 4
Size 71
Vehicle class 165
0 20 40 60 80 100 120 140 160 180
As shown in Figure 3-3, we find that grouping frequencies often do not map directly onto
consumer priorities. Most interesting to note is that key qualities such as safety, reliability, and
comfort rarely appear in the literature despite their expected relevance to consumer decision
making. In many cases, this is a result of limited data on these characteristics and few available
proxies. In other cases, some attributes may signal multiple qualities to consumers that may not
be captured in this taxonomy. Vehicle weight, for example, is a measure of size but also
correlates strongly with vehicle class.
We find that other core factors such as fuel cost, fuel type, and performance are
considered in many studies and provide grounds for comparative analysis. We also see that
vehicle class appears in several studies, though these attributes often serve to function as controls
or fixed effects rather than variables of interest.
3-7
-------
SECTION 4.
METHODOLOGY
This section provides an overview of the methodology we used to generate our estimates
of WTP based on the literature. Although some papers calculate and report WTP, many do not
though they provide sufficient information for WTP to be calculated.
4.1 Estimating Central Tendencies of WTP
The literature in the main sample presents three categories of empirical models from
which to derive willingness to pay (WTP) estimates:
1. Hedonic price models,
2. Multinomial logit (MNL) and nested multinomial logit (NMNL) models and,
3. Mixed logit (MXL) and other models (e.g., BLP4) with random distributions of
preferences.
In hedonic price models, vehicle price is the dependent variable and the vehicle's
attributes are explanatory variables. In the simplest form, the price of vehicle j ,pj, is a linear
function of its weighted attributes (xjk), with y/.-s as weights, as shown in Equation 4.1.
Pj = I'LiYkXjk (4.1)
Assuming the hedonic price function correctly represents a demand function, the
marginal value or willingness to pay for the kth attribute is the derivative of price with respect to
attribute xjk. In Equation 4.2 this is just the coefficient of xjk (Equation 4.2).5
dPj _
dxjk
Yk (4.2)
If attributes are interacted with other variables or if more complex functional forms are
used, the derivative will be more complex and may depend on the values of other variables. For
example, if all variables are entered as logarithms, the derivative of price would be yi/xjk, and a
mean value of Xk would be used to calculate the central tendency WTP.
4 Several models use the method of Berry, Levinson, and Pakes (1995) (BLP) to estimate random coefficient models
from aggregate sales data. We use the term MXL model to refer to random coefficient models estimated from
survey data.
5 Reduced form hedonic price models have a long-recognized identification problem when used to make inferences
about consumers' preferences (e.g., Nerlove, 1995; Rosen, 1974). Observed prices and quantities represent
solutions of supply and demand functions. Only with additional information can inferences about one or the
other be made with confidence. Many studies assume perfectly elastic supply at exogenous prices.
4-1
-------
In MNL and NMNL models, the indirect utility function6 of consumer i is a function of
vehicle attributes and, in general, other variables describing the consumer. The derivative of the
utility function with respect to an attribute gives the change in utility due to a marginal change in
one of its attributes. Purchase price is almost always one of the variables in the utility function.
However, the coefficient of any variable that is measured in present value dollars can be used if
price is not included. Because purchase price is measured in present value dollars, the negative
derivative of the utility function with respect to price is the marginal utility of a dollar of income
(since one dollar of price is equivalent to a negative dollar of income). It can be transformed into
a monetary utility function by multiplying through by l/(-/?), where fi is the coefficient of
purchase price, the minus sign being added so that utility is measured in positive dollars. This is
illustrated in Equation 4.3 for a simple linear utility function.
In Equation 4.3, the WTP (in dollars) for a change in attribute k is the derivative of Uy
with respect to Xjk, or ~Ub[>. Although simple linear utility functions such as Equation 4.3 are
sometimes encountered, in general, utility functions are more complex and include interactions
among variables and transformations of variables. In general, WTP is always obtained by
dividing the derivative of utility with respect to an attribute (3U/3x, whose units are utility per
unit of the attribute) by the negative of the derivative of utility with respect to a measure of
present value dollars such as vehicle price (-3U/3p, whose units are utility per dollar, present
value). Although we omit the consumer and vehicle subscripts in Equation 4.4, the derivatives
are often a function of consumer attributes and occasionally of vehicle attributes. In such cases,
we use measures of central tendency for those variables (e.g., mean household income) for the
population appropriate to the sample used in estimating the choice model.
In general, both a,j and P (or both derivatives in Equation 4.3) are random variables
because they are estimated with error. The first order Taylor series approximation to the ratio of
two random variables is just the ratio of the random variables. The mean of a ratio of random
variables is not generally equal to the ratio of the means because it is influenced by their
covariance. However, because published articles almost never provide the variance-covariance
matrix for coefficient estimates, we use the first order approximation in all cases to estimate
6 The utility function is called "indirect" because economists usually define utility as a function of quantities of
goods consumed. The indirect utility function is defined as the maximum utility a consumer with a given level of
income can achieve given the prices (and attributes) of goods.
Uij = Ppj + ELiajxjk => z^ = -vj + YJk=i-j*jk
(4.3)
WTPk
(4.4)
4-2
-------
WTP. We interpret this measure as the central tendency estimate of the marginal WTP for an
attribute conditional on the central tendency estimate of the price derivative. This differs from
the expected value of the ratio of the derivatives. On the other hand, it is computable from the
information provided in all the papers in our sample and has a meaningful interpretation.
The second order Taylor series approximation is useful for illustrating the potential
sources of error in the first order approximation. Consider the second order approximation of the
expected value, E[-a/p], of the ratio of two random variables, -a and P (Seltman, 2016)
(Equation 4.5).
If the coefficient estimates are uncorrected, the second right-hand-side term is zero; otherwise
the simple ratio WTP estimate will be biased if the coefficients are correlated. The third term's
effect could be either positive or negative. The direction of the bias introduced by excluding the
third term when using a first order rather than second order approximation depends on the sign of
-aj (P <0) and whether the variance of P is less than E3[P],
Daly et al. (2012, p. 336) show that if a,j and P are maximum likelihood estimators, their
ratio is also a maximum likelihood estimator (MLE) of the ratio of the true parameters.
Calculating the variance of the ratio, however, requires knowledge of the variance-covariance
matrix of the estimators, and this is almost never available in the published literature. Gatta et al.
(2015) demonstrate that the asymptotic property of the ratio of MLE estimators does not
preclude large errors when sample sizes are small. Fortunately, most of the papers we analyze
are based on relatively large samples (Table 3-2). Furthermore, when the price coefficient is far
from zero and its standard error is small, the ratio gives reliable results even when the sample
size is small. In the case of mixed logit models, even knowledge of the variance-covariance
matrix of estimated coefficients is generally not sufficient. Unbiased WTP estimates must be
obtained by simulation methods (e.g., Hensher and Greene, 2003).7 When authors provide WTP
estimates based on their own simulation analyses, we use the authors estimates. When authors do
not provide WTP estimates we use the ratio of derivatives method. As a consequence, in general,
our central tendency estimates of WTP, like nearly all those in the extant literature, should be
7 Concerning estimating WTP in mixed logit models, Hensher and Greene (2003, p. 163) state: "In deriving WTP
estimates based on random parameters one can use all the information in the distribution or just the mean and
standard deviation. The former is preferred but is more complicated. Simulation is used in the former case,
drawing from the estimated covariance matrix of the parameters." Unfortunately, the necessary information is
rarely provided in published articles.
E[-a] Cov[-a,p] Var[p]E[-a\
~E\f\ EHm~ eW]
(4.5)
4-3
-------
interpreted as conditional on the central tendency estimate of the price derivative. An extended
discussion of this issue can be found in Appendix C.
Carson and Czajkowski (C&C) (2013) point out that because coefficient estimates are
assumed to be normally distributed, there is always a theoretical probability that the denominator
of the WTP ratio will be zero, making the expected value undefined. While this is true in theory,
we consider it an artifact of the estimation methods with no practical importance, because it
implies that the marginal utility of income has a finite probability of being zero. The solution
proposed by C&C is to assume a different distribution for the coefficient estimates (e.g.,
lognormal) that has no probability density at zero. The problems associated with estimating WTP
from ratios of random variables can be avoided by estimating discrete choice models in WTP
space rather than preference, or attribute, space. However, only two papers in our sample used
the WTP space method (Train and Weeks, 2005; Helveston et al., 2015). Train and Weeks
(2005) observed that models estimated in preference space fit the data better.
Frequently, vehicle price is divided by household income (P/Y) in these specifications.
The marginal utility of income is expected to decline with increasing income, while sensitivity to
vehicle price is declining with income. In the utility equation (Equation 4.6), attributes of the
vehicle, Xjk, are interacted with attributes of the consumer, z,, such as income.
In this case, the WTP for attribute k depends on the central tendency values of the
coefficients and on both median income, F*, and the mean value of another consumer attribute,
Zi, as shown in Equation 4.7.
More complex formulations are frequently encountered but the WTP remains the negative of the
derivative of utility with respect to the attribute divided by the derivative of utility with respect to
vehicle price.
The same method used for MNL models is used for NMNL models. NMNL models are
more complex than simple MNL models because they represent a hierarchy of nested choices.
Choice of make and model may be nested inside (conditional on) choice of vehicle class.
However, the price of a vehicle and its attributes are located in the same nest. The derivatives of
the utility function at that level defines the tradeoff (marginal rate of substitution) between the
attribute and present value dollars (price). Thus, the utility functions of the nests that include the
attributes of vehicles and their prices are used in estimating marginal WTP using Equation 4.3. A
(4.6)
(4.7)
4-4
-------
given attribute may appear in several nests. We include the WTP measures from all nests in our
database.
To derive a central tendency estimate of WTP, the central value for income and the other
consumer attribute(s) must be known. Frequently, mean values for attributes are provided by a
paper's authors but none have provided the joint distributions of income and other consumer
attributes. The convention adopted in this paper is to use mean or median values (depending on
the data available) for all variables for the relevant population, at the midpoint year of the sample
data. When authors do not provide such data, it is often possible to find the appropriate data in
other sources (e.g., Census Bureau reports). In such cases, care has been taken to match the
relevant year and population whenever possible (e.g., new car buyers or all households? U.S.
households or those in California?).
In random coefficient models such as the mixed logit (MXL), some or all coefficients of
the indirect utility function are specified as random variables. Commonly, the papers use normal
distributions for coefficients of attributes whose marginal values may be either positive or
negative, and lognormal distributions are used when marginal values are believed to be either
always positive or always negative (e.g., fuel costs). The convention used in this paper is to use
mean values for normally distributed random coefficients and median values for lognormally
distributed coefficients for the central estimates of those coefficients. Mixed logit models can
become exceedingly complex when there are multiple, correlated random coefficients, and
vehicle attributes are interacted with several other variables. Some authors provide WTP
estimates they have calculated by simulation methods. In that case, we adopt the authors' WTP
estimates. Most authors provide sufficient information to derive central tendency WTP measures
using the convention describe above.
In this paper we focus exclusively on marginal WTP measures, that is, the willingness to
pay for one additional unit of an attribute. In some cases, it is more useful to estimate the WTP
for large changes in attributes (e.g., WTP for an increase in a battery electric vehicle's range
from 75 to 200 miles; see Dimitropoulos et al., 2013). The majority of papers in our sample are
derived from random utility models of consumers' vehicle choices. For many of these models
(e.g., MNL and NMNL) WTP measures for large changes in attributes can be readily estimated
using logsums (see, e.g., Zhao et al., 2012). For MXL models, simulation methods are required.
4.2 Measuring Preference Heterogeneity and Estimation Uncertainty
Although measures of the central tendencies of WTP for vehicle attributes are the first
goal of our research, all measures are subject to estimation error. In addition, many models
explicitly incorporate heterogeneity of preferences across consumers by estimating probability
4-5
-------
distributions for coefficients. When preference heterogeneity is not included in a model, we
estimate a range of WTP based on estimation error; otherwise we estimate a range of preference
heterogeneity but not estimation error. These two measures describe entirely different sources of
variability and are therefore presented and analyzed separately. Like our central tendency
estimates, our ranges of uncertainty suffer from a lack of knowledge about the covariance of the
attribute and price derivatives. In the absence of this information, we hold the price derivative
constant and vary only the attribute derivative. Thus, each range is conditional on the central
tendency estimate of the price derivative. Because of this, our ranges should not be interpreted as
probability or confidence intervals but rather as indicators of the degree of uncertainty in the
WTP estimates. A more detailed discussion can be found in Appendix C.
Attribute and price coefficients, as well as the attribute and price derivatives, are
estimated with error. Nevertheless, in models where there are no interactions of vehicle attributes
with consumer attributes, we calculate a range of uncertainty for WTP using +/- 1 standard error,
se, of only the attribute coefficient (Equation 4.8). This interval will be smaller than an interval
that included the error of estimation of the price derivative. However, including variability in the
price coefficient would require knowing the correlation between the price and attribute
coefficient estimates. In general, such data are not provided in the literature. Instead we focus on
the variability of the attribute coefficients, conditional on the central tendency estimate of the
price coefficient.
WTPLow = 2^, WTPmgh = ^ (4.8)
We use a single standard error range because, in practice, we have found that a two-
standard error range is frequently extremely wide, despite the fact that it includes no variability
in the price derivative. In our judgment, when a goal is to find a consensus among estimates, it is
more appropriate to use bounds that include two thirds of consumers rather than 95% of
consumers. Again, the potential correlation of a and P is not considered, nor is the uncertainty in
the estimate of p. Furthermore, because we are using only a first order approximation to the ratio
-a/p, the range of uncertainty should be considered only a general indication of the true
estimation uncertainty.
It is also important to understand how preferences may vary across the population of
vehicle buyers. Where variations in preferences can be reasonably estimated, we attempt to
approximate a range of ±1 standard deviation around the mean/median of the preference-related
variable (following the same rationale as that applied for using a one standard error range in
Equation 4.8). In general, articles do not provide sufficient information on the correlations
4-6
-------
among preference distributions to precisely estimate the preference heterogeneity implied by the
model. Our approach is intended to err on the side of underestimating ranges of heterogeneity.
For the many models in which the price coefficient, P, is interacted with income, we
calculate a range indicative of preference heterogeneity based solely on the distribution of
income, other consumer characteristics held constant. Of course, other consumer characteristics
vary with income, but the data necessary to accurately describe the covariances are either not
available for the sample population or would require substantial effort to estimate. Instead, we
vary income independently of other consumer characteristics as an indicator of the heterogeneity
of consumer preferences. Using Equation 4.7, we substitute the 25th and 75th percentiles of the
income distributions for median or mean income (depending on the data provided in the paper in
question). The same caveats noted above for estimation error apply to interpreting this range as a
true range of preference heterogeneity. In addition, in cases where attributes are interacted with
other consumer characteristics, we have not attempted to estimate the heterogeneity implied by
attributes other than income.
In MXL models, preference heterogeneity is a natural result of the distributions of
attribute coefficients. In MNL and NMNL models, we estimate preference heterogeneity from
the distributions of variables interacted with price and vehicle attributes, as described above. In
all cases, the range represents +/- 1 standard deviation of the attribute variable but not the price
variable. MXL models frequently assume that the price coefficient is not a random variable but
even when it is we use only its central tendency measure (mean or median).
For each paper an individual Excel workbook was used for the WTP calculations and to
generate a standard output table. This allowed us to send the worksheet to authors when
questions arose about the calculations. Having the correct units for all variables is critically
important but not all papers clearly state the units used in model estimation. The spreadsheet
format allows assumptions about units to be clearly documented and to be changed if so
indicated by an author's response to a query. We are grateful to the many authors who responded
promptly and helpfully to our queries.
The standard output for each paper included authors' names, date of publication, type of
data and description of sample, category of model, level of choice (e.g., make/model, powertrain,
vehicle class), constant dollar year, as well as attribute, price slope, estimated coefficients,
standard errors, standard deviations if a random coefficient model was used, and finally low,
central and high WTP estimates and the factor used to define the range (standard error, standard
deviation or variation in income). The standard Excel™ output tables were combined into a
Stata™ database for statistical analysis.
4-7
-------
SECTION 5.
WILLINGNESS TO PAY FOR THE ATTRIBUTES OF VEHICLES
In the following descriptive analysis, we present findings on the WTP values of key
attributes from the literature.8 Wherever possible, we have converted units to a common metric
to facilitate comparison; for example, a unit defined in terms of hundreds of miles per gallon was
standardized to miles per gallon. A few less straightforward conversions for fuel costs and
performance are explained below. For almost all of the WTP estimates we have calculated low,
central and high values. Because determining whether there are consensus values for attributes is
a goal of this study, most of the analysis focuses on the central tendency estimates. For estimates
based on random coefficients, or where attributes are interacted with each other and the
distribution of those attributes in the relevant population is known, the low and high values
measure the heterogeneity of preferences and represent +/- 1 standard deviation of the preference
distribution. For other estimates, the high and low values represent estimation uncertainty and
are equal to +/- 1 standard deviation of the attribute's coefficient estimate. Ranges based on
preference heterogeneity and estimation uncertainty are presented separately in figures or are
clearly labeled in tables.
As part of the effort to find consensus on attribute values we eliminated relatively few
"outliers" to create what we call "trimmed" samples. The National Institute of Standards and
Technology (2016) defines an outlier as ".. .an observation that lies an abnormal distance from
other values in a random sample from a population..." Our use of the term differs from this
definition in that we did not take a random sample of estimates of attribute values but rather
attempted to collect all estimates from U.S. studies published between 1995 and 2015. When we
omitted a study, it was because we were unable to calculate attribute values due to missing
information. In that sense, every value calculated belongs to the population of interest. There is
no rigorous statistical definition of "abnormal distance from other values." We have identified
outliers by creating histograms, visually identifying extreme values, and testing to ensure that,
once the extreme values were deleted, their distance from the mean of the trimmed sample was
greater than three standard deviations of the trimmed sample. For selected attributes, we have
included the full sample histograms in the main body of the report (see Appendix D and
Appendix E for figures representing untrimmed distributions of central WTP estimates for all
attributes, presented in two different ways). It was not possible to define clear rules for making
these adjustments; we are using professional judgment. Our intent was to remove a few
8 All values reported in dollars were converted to 2015$ using the CPI-U index.
5-1
-------
observations whose presence profoundly changes the estimated mean and variance of the set of
estimates in order to increase the likelihood of finding consensus among the remaining estimates.
As part of this study, we attempted to get feedback from authors of all papers included in
our main sample. Recognizing that there are some uncertainties involved in WTP calculations
based on the information available from their papers, we wanted to provide them with an
opportunity to comment on our methods for calculating WTP estimates based on their papers and
provide corrections/comments as appropriate. We started by contacting the corresponding author
using the contact information available in the publication or updated contact information when
one of the authors of this report was aware of an updated affiliation. In a number of cases, the
contact information provided for the corresponding author was no longer accurate and none of
the study authors knew their current affiliation, in which case we searched for an updated
affiliation and contact information and contacted them using that information where available. In
some cases, we could not locate current information and turned to contacting other study authors
for multi-authored papers. Detailed information on the comments received and our responses are
provided in Appendix F. We thank all authors that responded for their time and interest in our
study. Results presented in this section reflect our adjustments in response to all comments
received based on our interpretation of the comments received, but any remaining errors in
calculation or interpretation of WTP are the responsibility of the report authors and not the
authors of the individual studies.
Table 5-1 presents summary statistics for the central WTP values for the 32 individual
attributes (out of 142) that had five or more observations as well as aggregates for 1) aggregate
fuel cost per mile and 2) acceleration (0-60 mph) time reduction. The mean, standard deviation,
skewness, median, interquartile range, minimum and maximum describe the distribution of
estimates across studies and model formulations. In Table 5-1 and the other tables below, the
statistics presented describe the distributions of the central tendency estimates across studies.
Except where explicitly indicated, they are not the standard errors of individual estimates nor do
they reflect only heterogeneity of preferences. Instead, they reflect a combination of differences
due to time, place and populations included in the study, together with differences due to model
formulations, included and excluded variables, ways that attributes are measured and estimation
methods. Figures below represent high to low ranges of estimates due to estimation error or
preference variation; each line in these graphs represents an individual study and outliers are
included.9 WTP estimates for subcategories of the eight most commonly analyzed attribute
9 Estimation error and preference variation figures are truncated to focus on most study estimates. As such, lines
depicting outlier cases may extend outside the bounds of the graph area.
5-2
-------
categories (Comfort, Fuel Availability, Fuel Cost, Fuel Type, Performance, Pollution, and
Range) are described below in more detail in Sections 5.1 through 5.8, respectively. We include
figures showing the variation in WTP estimates across observations for selected vehicle
characteristics as illustrative examples of the variability present across observations (see
Appendix D and Appendix E for additional figures). Detailed WTP estimates for all 15 of our
general categories by study by model specification can be found in Appendix B.
Although many models include indicator variables for vehicle class, we do not include
the WTP estimates for vehicle class in Table 5-1. If all studies defined vehicle classes in the
same way, it would be possible to normalize estimates of WTP for vehicle classes by always
comparing to the same vehicle class. Unfortunately, definitions of vehicle classes vary
considerably across studies, making it impossible to compare the estimates. In contrast,
alternative fuel vehicles such as battery electric vehicles, plug-in hybrids and flex-fuel vehicles
are consistently compared with conventional gasoline vehicles. In that sense, the WTP estimates
are comparable across studies. Studies differ, however, in the way alternative fuel vehicles are
described, the alternatives included in the choice sets, and in the design of choice experiments.
5-3
-------
Table 5-1. Summary Statistics from Pooled Central WTP Estimates"
Raw
Trimmed
Inter-
Grouping
Attribute
N
Units
Out-
liers
Mean
SD
Min
Max
Mean
SD
Min
Max
Median
quartile
Range
Skew
Comfort
Auto-transmission
9
0/1
1
1,818.8
3,739.3
-2,987.0
9,260.6
888.6
2,660.8
-2,987.0
5,321.4
1,090.3
3,262.9
0.8
Rear-wheel drive
6
0/1
0
32,030.9
18,030.7
10,069.8
62,928.8
32,030.9
18,030.7
10,069.8
62,928.8
26,778.8
16,189.4
1.2
Air conditioning
13
0/1
0
3,484.2
9,627.1
-15,380.0
19,818.5
3,484.2
9,627.1
-15,380.0
19,818.5
3,961.6
7,474.0
0.9
Shoulder room
12
$/inch
1
1,085.1
1,393.7
178.4
5,266.6
705.0
478.7
178.4
1,800.1
545.9
764.2
1.3
Fuel
Recharging time
27
$/hr
0
2,194.8
2,923.1
-227.4
11,947.1
2,194.8
2,923.1
-227.4
11,947.1
930.9
2,689.9
2.4
availability
Fuel availability
18
$/%
2
834.5
2,133.1
48.8
9,133.3
227.8
201.4
48.8
789.6
161.7
166.7
1.9
Fuel costs
Aggregate fuel cost
117
$/cpm
7
-8,330.7
97,820.1
-1,052,470
64,499.0
1,879.7
6,875.4
-7,425.0
64,499.0
990.6
2,194.2
1.9
(value of
reduction in
costs)
per mile
Cost per mile
60
$/cpm
0
1,366.3
3,318.0
-7,425.0
19,415.1
1,366.3
3,318.0
-7,425.0
19,415.1
1,146.5
2,546.5
1.2
Cost per year
15
$/($/yr)
1
-63,204.0
274,176.8
-1,052,470
64,499.0
7,457.8
17,256.0
-3,224.1
64,499.0
1,133.9
4,723.8
6.6
Gallons per mile
24
S/0.01
gpm
3
46.1
3,090.7
-7,472.2
4,701.1
1,066.3
1,483.6
-1,094.8
4,701.1
1,027.1
1,816.8
1.0
Miles per dollar
8
$/
(10mi/$)
3
-14,917.7
20,600.8
-59,618.7
-676.5
-2,328.4
1,795.8
-4,985.1
-676.5
-2,097.0
2,349.3
1.1
Miles per gallon
7
$/mpg
0
991.4
1,404.0
-325.2
3,848.9
991.4
1,404.0
-325.2
3,848.9
800.3
1,451.5
1.2
Other converted
3 $/GGE or
0
896.4
702.1
115.6
1,475.9
896.4
702.1
115.6
1,475.9
1,907.6
1,360.3
0.8
units
$/gallon
Fuel type
Electric vehicle
27
0/1
1
-10,525.6
22,711.8
-77,780.3
30,651.0
-7,938.8
18,670.1
-43,983.9
30,651.0
-8,454.0
28,385.9
0.9
Hybrid
27
0/1
2
-12,671.0
44,878.6
-180,394.4
18,860.1
-1,436.6
18,573.8
-55,816.5
18,860.1
2,374.7
11,880.8
-0.6
Flex fuel
6
0/1
0
5,166.3
5,692.3
-4,409.9
10,975.3
5,166.3
5,692.3
-4,409.9
10,975.3
6,114.4
6,298.4
0.8
PHEV
5
0/1
0
12,337.8
12,061.0
-7,959.3
23,809.8
12,337.8
12,061.0
-7,959.3
23,809.8
14,740.0
4,877.4
0.8
Methanol
5
0/1
0
11,134.3
2,884.9
6,989.4
13,962.7
11,134.3
2,884.9
6,989.4
13,962.7
12,587.2
3,461.5
0.9
Natural gas
7
0/1
2
-5,619.6
23,691.2
-55,978.3
12,956.4
6,187.4
3,850.6
3,295.7
12,956.4
5,006.2
439.2
1.2
(continued)
-------
Table 5-1. Summary Statistics from Pooled Central WTP Estimates" (continued)
Raw
Trimmed
Grouping
Attribute
N
Units
Out-
liers
Mean
SD
Min
Max
Mean
SD
Min
Max
Median
Inter-
quartile
Range
Skew
Model
Availability
Make-model
availability
14
$/# of
models
2
898.8
2,281.3
0.5
6,841.5
5.9
7.6
0.5
22.1
2.1
8.2
2.8
Performance
Aggregate
acceleration (0-60)
time reduction
48
$/s
0
953.7
1,259.2
-1,546.9
5,543.5
953.7
1,259.2
-1,546.9
5,543.5
1,004.9
1,199.5
0.9
Acceleration (0-30)
time reduction
11
$/s
0
1,045.2
1,122.8
-1,546.9
3,287.5
1,045.2
1,122.8
-1,546.9
3,287.5
1,140.6
608.9
0.9
Acceleration (0-60)
time reduction
$/s
0
1,095.6
627.4
34.6
2,200.1
1,095.6
627.4
34.6
2,200.1
1,182.7
497.8
0.9
Horsepower/ weight
29
S ii.li I hp
lb
0
879.8
1,448.5
-860.4
5,543.5
879.8
1,448.5
-860.4
5,543.5
198.4
1,558.1
4.4
Horsepower
11
$/hp
0
53.6
108.8
0.0
355.0
53.6
108.8
0.0
355.0
9.2
38.2
5.8
Top speed
9
$/mph
0
100.1
58.3
27.6
209.7
100.1
58.3
27.6
209.7
54.2
86.5
1.8
Pollution
Emissions reduction
19
$/10%
0
48,007.8
69,595.8
-66,982.0 168,535.7
48,007.8
69,595.8
-66,982.0
168,535.7
1,491.3 132,083.1
32.2
Range
Range
40
$/mi
0
86.3
51.5
-20.1
242.6
86.3
51.5
-20.1
242.6
87.3
62.5
1.0
*Attributes in italics are combined into aggregate measures.
-------
5.1 Comfort Grouping
5.1.1 Automatic Transmission
There were nine WTP observations for automatic transmission in the surveyed literature,
pulled from four studies. A dummy indicator reflected preference for automatic transmission as
opposed to manual, or stick-shift, transmission. After dropping one extreme value greater than
$9,000, we found a trimmed mean of $889 for automatic transmission, although a relatively large
spread remained: the interquartile range spanned -$983 to $2,280. The slight negative skew
comes primarily from the two negative estimates from the Haaf et al (2014) study. The
remaining estimates reflect the anticipated positive sign and cluster close to the median value of
$1,090. WTP estimates for automatic transmission draw from either market data or revealed
preference surveys, and were produced using a variety of estimation strategies (e.g., hedonic,
MNL, MXL, NMNL). No studies attempted to capture population heterogeneity in taste for
transmission systems.
5.1.2 Rear-wheel Drive vs. Front-wheel Drive
We find six estimates for WTP for rear-wheel drive, all of which come from the same
Petrin (2002) study.10 Petrin employs a BLP model on market data from 1981-93. Estimated
WTP for rear-wheel drive is consistently positive and very large, with a mean of $32,031, and an
interquartile range of $16,189 (Figure 5-1). The transition during this period to very high
penetration of FWD, which has persisted, is difficult to reconcile with the generally large WTP
estimates for real-wheel drive. It seems likely that this parameter is aliasing other factors with
which it is correlated, such as the use of rear-wheel drive in high-performance vehicles.
Differences among the six estimates are due to different price coefficients for each of three
income tertiles, which vary by a factor of four, and from two different estimation methods whose
attribute coefficients vary by more than a factor of two. In general, WTP for rear-wheel drive is
greatest for the lowest income tertile and least for the highest. Given that all six estimates come
from the same study, it is difficult to make a judgment on the typical WTP for rear-wheel drive,
particularly as WTP is sensitive to the estimation method and may be aliasing other factors. The
estimates do indicate substantial consumer heterogeneity related to income; high and low WTP
for a given model specification differ by more than $10,000 on average. The Low (-1 std. dev.),
Central, and High (+1 std. dev.) estimates from each observation are shown in Figure 5-1.
10 Petrin (2002) directly includes a dummy variable for FWD, but we used the opposite of the sign to represent the
WTP for having rear-wheel drive because we were trying to standardize having as many of our WTP measures
represent positive valuations for attributes as possible.
5-6
-------
Figure 5-1. Rear-Wheel Drive Preference Variation, Range is +/- 1 Standard Deviation
590,000
$80,000
570,000
560,000
550,000
540,000
530,000
520,000
510,000
SO
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute.
5.1.3 Air Conditioning
There are 13 observations for air conditioning. The mean is $3,484, with no clear outliers
but a high standard deviation of $9,627. Estimates cross over both positive and negative values,
but the majority have positive values. All negative values come from the Petrin (2002) paper.
Data are either from revealed preference surveys or are market data. Aside from one study (Haaf
et al2014, which used data from 2004-2006), all data fall between 1971 and 1993, reflecting
developments in vehicle design as air conditioning became a standard feature and thus a weak
source of variation amongst vehicles manufactured in the last few decades.
No clear divergences emerge in the observations due to estimation strategy. A variety of
models are tested: BLP, NMNL, hedonic, MNL. In models that allowed variation in population
taste, high and low estimates vary considerably. Two studies present near zero variation in
population taste; others produce differences in population taste on the order of several thousand
dollars. This latter variation is particularly notable for the Petrin (2002) paper.
ffrtsffengmce Variation
Oentral
5-7
-------
WTP for air conditioning is generally positive and valued in the thousands of dollars (see
Figure 5-2), though the data are outdated in the surveyed literature and show no clear
convergence in value. Valuation for this attribute is particularly challenging and perhaps
inconsequential in studies on new vehicles.
Figure 5-2. Population Taste Heterogeneity for Air Conditioning
$30,000
$20,000
$10,000
SO
-$10,000
-$20,000
-$30,000
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
5.1.4 Shoulder Room
WTP for shoulder room is presented as dollars per additional inch. There are twelve
observations, all of which are from the Liu et al. (2014) paper using revealed preference survey
data and a multinomial logit estimation technique. The initial mean is $1,085 per additional inch;
we remove one extreme value greater than $5,000 to produce a trimmed mean closer to $700 and
a lower standard deviation of $478. Liu et al. (2014) produce their range of estimates by
estimating WTP by characterizing households by vehicle fleet size (1-4 car households) and
low-, medium-, and high-income segments. As expected, WTP rises for higher income brackets;
no clear pattern emerges in WTP based on household fleet size, though more extreme values
occur in fringe cases (e.g., high income households with one car have an average WTP of $5,267
per inch of shoulder room).
WTP for Air Conditioning: Range is 1 Standard Deviation
Preference Variation
5-8
-------
5.2 Fuel Availability
5.2.1 Recharging Time
There were 27 estimates in the literature from which we could calculate willingness to
pay for recharging time. These electric vehicle studies relied almost entirely on stated preference
surveys (aside from one market data set). Thirteen of the estimates came from the same 2008-9
web survey (Parsons et al., 2014; Hidrue et al., 2011). Units were normalized to willingness to
pay for a one-hour reduction in charging time. For simplicity, we assumed a linear relationship
between willingness to pay and percentage reductions in charging time when performing unit
conversions.
The 27 estimates span from -$227 to $11,947, as shown in Figure 5-3 below. The
variation may be due to differences in willingness to pay for specific charge times. Most
estimates were not continuous and had been converted from dummy variables (e.g., charge time
of 15 hours versus 5 hours). We find that the value of further reducing charge time from lower
charge times (e.g., less than 10 hours) produce willingness to pay values beyond $1,000 per hour,
while the value of reducing charge time for most 15- and 20-hour charge times are valued
between $400 and $950 per hour. It is reasonable to expect that marginal willingness to pay for
charge times at higher levels of range would be lower, reflecting a decreasing marginal utility of
reducing charging time as range increases. This literature could benefit from the study of
additional data sets, but we tentatively find that there are increasing returns to reducing EV
charge times below ten hours.
Figure 5-3. WTP for a One Hour Reduction in Charge Time Across Studies
$14,000.00
$12,000.00
$10,000.00
$8,000.00
$6,000.00
$4,000.00
$2,000.00
$0.00
-$2,000.00
~~~
~~~~~~~~
Note: Each point represents one of the 27 studies estimating the WTP for reduction in charge time (normalized to
one hour reduction).
5-9
-------
5.2.2 Fuel Availability
There were eighteen recoverable estimates of willingness to pay for fuel or station
availability within the literature. As with charge time, we assume a linear relationship between
WTP and percent increases in fuel availability to convert to common units. Final units are in
WTP for one percent increases in station availability. After dropping two outlier values of $2,242
and $9,133, we find a median value of $161.74 for a 1% increase in station availability. Half of
the estimates fall between $105.70 and $272.43 per 1% increase in availability.
Fuel availability data came from the Brownstone et al. (1996) California household
survey (12 observations), other stated preference studies (3 observations, 1 dropped), and a
literature review (3 observations, 1 dropped). Models were specified as nested multinomial logits
or mixed logits. Mixed logits tended to produce central WTP values on the lower end of the
distribution, from $48.75 to $144.35. We only have enough information to estimate preference
heterogeneity for two of the trimmed sample observations. In both of these cases, the high-low
range of one standard deviation spans over $200, crossing over zero.
5.3 Fuel Cost
In the literature reviewed, we focus on fuel cost measured in five different ways (see
Table 5-1 and subsections below).11 Willingness to pay for reductions in fuel cost in $ or cents
per mile (fuel price/mpg), fuel cost per year ($/year) and gallons per mile of fuel consumption
(1/mpg) are expected to be positive, as is WTP for increases in miles per gallon. Results for each
of these measures are presented below. To increase the number of estimates that can be directly
compared, we have also converted gallons per mile to cents per mile by multiplying by the price
of fuel and report values both in the native units used in the studies as well as for a common cost
per mile metric in Table 5-1.
5.3.1 Reduction in Fuel Cost per Mile
The effective sample of estimates can be increased by converting different fuel cost
metrics to a common metric, when it can be done straightforwardly and transparently. There
were a total of 60 observations of estimates of WTP for reductions in fuel costs per mile. The
most frequently used metrics are fuel cost per mile (fuel price/mpg) and gallons per mile
(1/mpg). WTP for an increase in fuel consumption of 0.01 gal./mi. can be converted to WTP for
a $0.01/mile decrease in fuel cost by dividing by the price of gasoline. Between 1985 and 2014,
the annual average price of gasoline in 2015 dollars ranged from a low of $1.62 to a high of
11 We dropped five observations from the aggregate fuel cost per mile calculation (going from 122 to 117) because
they could not be converted from their native units to comparable $/cpm measures.
5-10
-------
$3.81 per gallon. A five-year backward moving average ranges from $1.82 to $3.55 and the
simple average for the 1985-2015 period is $2.45. We round to $2.50 and assume that as our
expected gasoline price for all studies.12 In the figures and tables below, the WTP for 0.01
gallons per mile (gpm) has been converted to WTP for $0.01/mile by dividing by 2.5.
The distribution of the combined central WTP estimates is shown in Figure 5-4. The
discounted present value of fuel consumption for a typical U.S. light-duty vehicle provides a
useful reference point for identifying outliers. Using NHTSA (2006)'s estimated expected miles
by vehicle age for passenger cars and light trucks, discounted at 6% per year, the "present value"
miles are 110,382 for a passenger car and 123,458 for a light truck, and the simple average for
the two vehicle types is 116,920. Thus, a reasonable reference point for the value of a $0.01/mile
decrease in fuel costs would be $1,169. Seven estimates less than -$50,000 or greater than
$20,000 were deleted as outliers, resulting in the trimmed distribution shown in Figure 5-5.
Figure 5-4. Willingness to Pay for $0.01/mile Decrease in Fuel Cost: All Estimates (2015$)
WTP for Reduced Cents per Mile: Untrimmed
-1000000.00 -800000.00 -600000.00 -400000.00 -200000.00 0.00
2015 dollars
12 Ideally, one would want to use gasoline prices that align with those used in each individual study, but we do not
have that information for all studies.
5-11
-------
Figure 5-5. Willingness to Pay for $0.01/mile Decrease in Fuel Cost: Trimmed Sample
(2015$)
WTP for Reduced Cents per Mile: Trimmed
o
CO ~
11 "i 1 1 1 1 r~
-30000.00 -20000.00 -10000.00 0.00 10000.00 20000.0i
2015 dollars
Statistics for the combined metric ($0.01/mile) are shown in Table 5-2. The central
tendency estimates vary widely, even after being "trimmed" of outliers. Standard deviations
range from about 70% of the mean for evidence from studies combining revealed and stated
preference (RP & SP) surveys to almost eight times the mean for trimmed estimates based on
market sales data. Second, although the distributions of most estimates are less skewed after
trimming, in most cases the skewness is still great enough to favor use of the median over the
mean as a measure of central tendency. In general, the interquartile ranges (75th percentile-25th
percentile) are also large relative to the median values.
All the means and medians of the trimmed samples are positive (decreased fuel cost has
positive value) as expected. For most types of data, the magnitudes of the medians are between
zero and two times a reference estimate of the value of fuel costs to a typical light-duty vehicle in
the U.S. ($1,169). The total, median trimmed estimate is $991. A closer examination reveals that
the median estimates based on stated preference or SP & RP data provide a much greater
willingness to pay ($1,400 to $1,900) than those based on RP surveys ($580 to $690) or market
sales data ($100 to $275). Mean RP survey and market sales estimates are much greater than
medians, though still less than the SP survey mean.
5-12
-------
Table 5-2. Willingness to Pay for $0.01/mile Decrease in Fuel Cost—Combined GPM and
$0.01/mile Values
Standard Skew- Number of
Data Type
Mean
Deviation
ness
Median
P75-P25
Minimum
Maximum
Observations
Literature Review (no
outliers removed)
916.12
552.69
0.69
635.30
992.62
560.22
1,552.84
3
RP & SP Surveys (no
outliers removed)
1,712.86
1,166.82
1.00
1,417.13
1,845.78
602.49
3,918.48
7
RP Survey (untrimmed)
-66,796.13
27,275.00
-3.47
583.27
3,229.95
-1,052,470.00
19,415.06
15
RP Survey (trimmed)
3,609.11
6,579.91
1.89
691.80
3,191.61
-325.22
19,415.06
14
SP Survey (no outliers
removed)
3,809.16
12,214.37
4.23
1,888.55
2,817.04
-7,425.04
64,498.98
29
Market Sales (untrimmed)
-1,554.82
8,868.82
-5.07
97.71
1,946.94
-59,618.71
4,701.15
63
Market Sales (trimmed)
544.429
1,632.17
-0.54
274.61
1,394.43
-4,985.11
4,701.15
57
Total Untrimmed
-8,330.70
97,820.05
-10.52
737.76
1,912.29
-1,052,470.00
64,498.98
117
Total Trimmed
1,879.67
6,875.42
7.18
990.63
2,194.24
-7,425.04
64,498.98
110
Note: To differentiate between summary statistics based on trimmed and untrimmed samples for data types where
outliers were removed (RP Survey and Market Sales), we present both sets of values in the table, italicizing the
untrimmed values.
The Low, Central and High estimates for each observation (combination of study and
model specification) are shown graphically in Figures 5-6 and 5-7. Each paper is represented by
a single line. Outliers have not been removed. Not all papers provided sufficient information to
calculate a range of estimates so the sample sizes are smaller. Figure 5-6 contains estimates from
random coefficient models and the range from low to high represents +/- 1 standard deviation of
the estimated distribution of preferences. For estimates with varied income, the range
approximates an interquartile range for the relevant income distribution. Figure 5-7 shows the
estimates from fixed coefficient models and illustrates the uncertainty due to estimation error.
The range from low to high is +/- 1 standard error of the attribute coefficient. Although the range
across estimates is large for both preference heterogeneity and estimation error, for a given paper
the range of estimation error is generally smaller with a few exceptions. Preference heterogeneity
in fuel costs should be expected if for no other reason than the variation in vehicle usage across
households.
5-13
-------
Figure 5-6. Range of Reduction in Fuel Cost per Mile WTP Estimates Describing
Preference Heterogeneity
$30,000
$20,000
$10,000
$0
-$10,000
-$20,000
-$30,000
-$40,000
WTP for $0.01/Mile Reduction in Fuel Cost:
Range is +/-1 Standard Deviation
Preference Variation
•
|||J ugh
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
Figure 5-7. Range of Reduction in Fuel Cost per Mile WTP Estimates Describing
Estimation Uncertainty
WTP for $0.01/Mile Reduction in Fuel Cost:
Range is +/-1 Standard Error
Estimation Error
$15,000
$0
L
-$5,000
-$10,000
-$15,000
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
5-14
-------
In general, the median estimates of willingness to pay for a reduction in fuel cost per mile
fall within a range of two times the reference estimate ($1,169 x 2 = $2,338) to minus one times
the reference estimate (i.e., within a range of-$1,169 to $2,338). In general, the range of
estimates is large relative to measures of central tendency. Estimates are typically skewed,
suggesting that the median is a better measure of central tendency than the mean. Median WTP
estimates from stated preference surveys provide a much greater willingness to pay for fuel
savings than median estimates based on revealed preference data ($1,889 versus $692, see
Table 5-2). The median of central tendency WTP estimates based on stated preference data is
60% greater than the present value of lifetime fuel costs for a typical new light-duty vehicle in
the U.S. The corresponding median WTP estimate based on revealed preference data is about
60% of the discounted present value of lifetime fuel costs for a typical new light-duty vehicle in
the U.S. This pattern suggests that hypothetical bias (e.g., Loomis, 2014) may be present in the
inferences from stated preference surveys. The median WTP estimate from studies using market
sales data is smaller still ($275), only about one-fourth of the reference value.
5.3.2 Dollars per Year
Fifteen observations measured fuel costs in dollars per year ($/yr). Six of the 15 come
from Axsen et al.'s (2009) study based on Californian and Canadian survey data. The remaining
nine come from seven papers, of which three (accounting for 4 of the 9 estimates) made use of
the same California stated preference survey. When considering WTP estimates, the valuation
that might be expected from an economically rational consumer provides a useful reference
point. However, it should not be considered the correct value because the assumptions used to
generate it will always be uncertain to a greater or lesser extent. For a rational consumer, the
value of reducing fuel cost by one dollar per year would be calculated by summing across the
savings provided over the expected years of vehicle life, discounted to present value.
Discounting vehicle survival probabilities from (NHTSA, 2006) at 6% per year, the discounted
expected life of a passenger car is 8.9 years (12.8 years undiscounted) and 9.2 years for a light
truck (14.6 undiscounted). A reasonable reference point for WTP for a $l/year decrease in fuel
costs would therefore be about $9. Removing estimates less than -$400 leaves 14 data points
with a mean of $65 and a standard deviation of $150. The median value is $10 with an
interquartile range of $41.
5.3.3 Gallons per Mile
There are twenty-four estimates of willingness to pay for a 0.01 gallon per mile (gpm)
decrease in fuel consumption. Twenty-two are from studies using market sales data, and one
each from studies based on revealed preference and stated preference surveys. The estimates
5-15
-------
based on market sales include three data points less than -$12,500; we identify these as outliers.
Including the outliers, the mean estimate of WTP for a 0.01 gpm reduction in fuel consumption
is $215 with a standard deviation of $7,577. In the presence of outliers, the median is a better
measure of central tendency, $1,954. Trimming three extreme values produces a mean estimate
of $2,697 with a standard deviation of $3,744, and the median becomes $2,835. Even in the
trimmed sample, the interquartile range is $4,542.
5.3.4 Miles per Dollar
Eight estimates of the WTP for tens of miles per dollar come from two papers (Berry et
al., 1995; Petrin, 2002). Both papers estimate random coefficient models using the method of
Berry et al. (1995). In theory, the WTP for 10 miles per dollar can be derived from the WTP for
mpg by multiplying by 10 and dividing by the price of fuel. Thus, if the WTP for 1 mile per
gallon is $450 and gasoline costs $2.50/gallon, the WTP for 10 miles per dollar would be $1,800.
The estimated mean WTP from the full sample is -$18,006. Removing three outliers of less than
-$29,000 results in a mean estimate of-$3,270 with a standard deviation of $2,953. The median
estimate is also negative, at -$2,486. One would expect a positive WTP for an increase in miles
per dollar; the negative values may indicate that the variable is aliasing less desirable other
factors correlated with miles per dollar.
5.3.5 Miles per Gallon
Seven observations used miles per gallon (mpg) to represent vehicle fuel consumption.
Because the marginal value of a mile per gallon depends strongly on the initial mpg, estimates
should be expected to vary over time and from one consumer to another, as well as with the price
of fuel. Using expected annual vehicle travel by vehicle age from NHTSA (2006) and
discounting at 6% per year, a typical US passenger car would accumulate 110,332 discounted
lifetime miles, while the corresponding figure for a typical light truck would be 123,458 for a
simple average of 116,920. If gasoline costs $2.50 per gallon and the typical light-duty vehicle
gets 25 miles per gallon, the economically rational reference point would be a WTP of $450 for
one additional MPG. Again, this should not be interpreted as the correct WTP but merely as a
known reference point.
The mean estimate of WTP for an additional mpg based on the full sample of estimates is
$991 with a standard deviation of $1,404. The distribution of estimates is skewed (1.24) and the
median estimated WTP is $800. The interquartile range is also very large relative to the median:
$1,452.
5-16
-------
5.4 Fuel Type
As noted above, the WTP estimates for alternatively fueled vehicles are all relative to a
conventional gasoline vehicle.
5.4.1 Electric Vehicles (EVs)
The sizeable sample of 27 estimates of WTP for electric vehicles produced a trimmed
mean of approximately -$7,940 and a standard deviation of about $19,000. The estimates are
distributed across negative and positive values; they are primarily negative. The interquartile
range spans from -$20,378 to $8,008. The wide variation suggests little agreement in the
literature on consumer valuation of EVs.
All data are from survey data, primarily stated preference surveys. Over half of the
estimates make use of the same Brownstone et al. (1996) survey data from a phone-based
California study. Of the remaining observations, several others draw from California surveys.
The majority of studies employ mixed logit models. Notably, the few positive estimates come
from studies in which authors restricted the sample using 'early adopter' indicators, or
designated 'EV-oriented' classes based on consumer characteristics.
Valuation of EVs varies considerably within a sample. Figure 5-8 below reflects low,
central, and high WTP estimates for each study that allowed some variation across the
population, either using random coefficients, or in some cases, including income interactions.
Each line represents high and low WTP values produced by adding or subtracting one standard
deviation from the EV coefficient. We see high slopes indicating large variation across a
population for many of these studies, even within the California-centric data.
5.4.2 Hybrid Vehicles
Out of 27 estimates for WTP for hybrid vehicles in our sample, we found two extreme
negative values. After restricting results to greater than -$100,000, the mean increases to
-$1,437. Results are nonetheless scattered, as reflected in a trimmed standard deviation of over
$18,000 (Figure 5-9). The median may be a more appropriate measure of central tendency at
$2,375, given the strong negative skew still remaining after trimming the sample. Within the
interquartile range, values are largely positive, falling between -$425 and $11,456.
5-17
-------
Figure 5-8. Population Taste Heterogeneity for Electric Vehicles
WTP for Electric Vehicle: Range is +/-1 Standard Deviation Preference
Variation
$60,000
$40,000
$20,000
$0
-$20,000
-$40,000
-$60,000
-$80,000
-$100,000
-$120,000
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
Figure 5-9. Distribution of Trimmed Central WTP Estimates for Hybrid Vehicles
WTP for Hybrids: Trimmed
-60000.00
-40000.00
-20000.00
2015 dollars
0.00
20000.Oi
The relatively large sample of WTP estimates for hybrid vehicles may represent the surge
in interest in alternatives to conventional gas vehicles in the past fifteen years. All data are post-
2000, given the recent introduction of the technology. Studies primarily rely on MNL and MXL
models. There is some mix of stated and revealed preference surveys and market data. There is
high inconsistency in the results produced by different data types and models utilized. Three
studies use national market sales data within the same period from 2006-2008, all utilize MXL
5-18
-------
models, and produce widely scattered central WTP estimates of-$35,000, -$7,000, and
$11,000. Notably, most positive values come from stated preference surveys, ranging from
approximately $2,000-$3,000 to $10,000. In many of these cases, the authors analyze revealed
preference data from the same sample and find strong negative valuations of hybrid vehicles,
suggesting some dissonance between hypothetical and practical preferences for hybrids among
consumers. Estimating the WTP for hybrids from market data is challenging because hybrids
were a relatively novel technology during the time period of the studies. In general, market-based
studies did not explicitly control for consumers' aversion to the risk of novel technologies and
their general unfamiliarity with hybrid vehicles. These perceptions are likely to change as
hybrids become more common, which would make early market-based WTP estimates
misleading if applied to future markets.
The estimated range of preference heterogeneity varies across the studies that allow
variation in taste (Figure 5-10). In general, high and low estimates of WTP based on random
coefficient models are considerably spread, on the order of several thousand dollars. Several
observations produce near-zero slopes—indicating limited variation in taste across a population;
each of these observations with limited variation come from Liu (2014). Liu produced separate
sets of estimates by income subset, and so the variation in taste within each income subset is
minimal.
Despite widespread interest in alternative vehicle technologies, the literature has yet to
agree on a central valuation for hybrid vehicles among consumers. Future work should account
for differences due to the data type (particularly for survey data), modeling strategy, and study
sample.
5.4.3 Flexible Fuel
We find six estimates of willingness to pay for flexible fuel vehicles in the literature.
Each estimate comes from a stated preference survey. Five out of six of the studies were
conducted between 1996 and 1999; Hess et al. conducted the only recent study in 2012
incorporating flexible fuel vehicles.
There is considerable variation in the central values across papers, ranging from -$4,410
to $10,975. Only one study (Tompkins et al., 1998) finds a negative central value, but two other
studies find negative WTP values within one standard error from the mean. Researchers used a
range of modeling strategies—multinomial logits, nested logits, mixed logits—that produced
values spanning the distribution. None of the studies interacted household characteristics with
the flexible-fuel dummy. We do not find emergent patterns in the limited pool of observations to
explain the variation.
5-19
-------
Figure 5-10. Population Taste Heterogeneity for Hybrid Vehicles (excluding outliers)
WTPfor Hybrid Vehicle: Range is +/-1 Standard Deviation
Preference Variation
$60,000
$40,000
$20,000
$0
-$20,000
-$40,000
-$60,000
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
5.4.4 Plug-in Electric Vehicles (PHEVs)
We find five WTP estimates of plug-in electric vehicles (PHEVs) from four different
stated preference studies. Each study was published after 2010, reflecting the relatively recent
development and popularity of this technology. There is one negative central WTP value; the
four other estimates range from $13,112 to $23,810.
The five estimates provide some insight into heterogeneity in consumer preferences
within each sample. Three of the estimates were income-interacted and two were estimated using
a mixed logit specification. Varying the income interaction by one standard deviation produces
differences in willingness to pay of $915 to $2,400. Tanaka et al.'s (2014) mixed logit
specification produces differences of $344 for one standard deviation. Zhang & Gensler's (2011)
mixed logit produces much larger differences of $13,333. This latter study produced the only
negative central WTP value of-$7,959. We see that some subset of the sample does positively
value PHEVs despite the negative central tendency.
As with other alternative fuel technologies, we are limited to a small pool of studies and
reliance on stated preference surveys. The literature does suggest that individuals tend to
positively value PHEVs, even after considering error bounds on the central tendencies (not
shown in Table 5-1).
CBmTSl
5-20
-------
5.4.5 Methanol
There were five estimates of WTP for methanol vehicles. The estimates came from three
studies (Kavalec, 1999, Brownstone et al., 1999, Brownstone et al., 2000) that analyzed the same
1996 survey data of California households. Every study used a mixed logit model, varying in its
incorporation of SP and RP data, the variables included, and socioeconomic interactions (e.g.,
age, college education). Central tendencies ranged from $6,989 to $13,963. Consumer
preferences varied considerably within each sample. One standard deviation from the mean
produces differences of $9,000 to $15,000.
5.4.6 Natural Gas
Seven estimates of WTP for natural gas vehicles were identified. We trim an initial mean
of-$5,620 to $6,187 by removing two extreme values lower than -$9,000. In the trimmed
sample of five estimates, we see a narrow interquartile range from $4,620 to $5,059. The
majority of estimates are positive.
All estimates draw from survey data—mostly stated preference and a few revealed
preference. Data are primarily from California; in some studies, separate results are presented for
California and the US excluding California. These latter estimates reveal stark differences in
consumer WTP based on the study sample: Tompkins et al (1998) find a WTP of-$9,000 for
natural gas vehicles in a national survey (excluding California), and WTP of approximately
$3,000 for California. Their estimate for California accords well with the remaining WTP
estimates for California, which cluster around the same value of $3,000 despite varying
modeling strategies.
Even with confluence in central WTP values, we find high variation in population taste
from models employing random coefficients (Figure 5-11). Estimates span both positive and
negative values; the range between high and low estimates is on the order of tens of thousands of
dollars. This large variation arises at the same time that there is little diversity in study samples
and the data used is primarily from the 1990s.
5-21
-------
Figure 5-11.Natural Gas Vehicle Preference Variation: Range is +/-1 Standard Deviation
WTP for Natural Gas Vehicle, Range is +/-1 Standard Deviation
Preference Variation
$50,000
$40,000
$30,000
$20,000
$10,000
$0
-$10,000 L
-$20,000
-$30,000
-$40,000
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
5.5 Performance
With 68 estimates,13 performance is the third most frequently measured vehicle attribute,
after vehicle price and fuel economy. We use five different measures of performance (Table 5-1)
that have at least 5 observations each. Three of the metrics are useful measures of acceleration
performance. Willingness to pay for reductions in the number of seconds required to accelerate
from 0-30 mph (11 observations) and 0-60 mph (8 observations) and WTP for
horsepower/weight (hp/lb, 29 observations) can each be used as measures of acceleration
performance, and WTP for each should be positive. WTP values for top speed (mph) (9
observations) and horsepower (11 observations) are also expected to be positive. Horsepower is
an ambiguous measure of performance since horsepower must increase with vehicle mass and
size to maintain constant acceleration. It thus partially measures vehicle size. The mean
willingness to pay for 1 additional horsepower based on estimates from 11 papers is $54 with a
standard deviation of $109. The median WTP estimate of $9 is considerably less than the mean
but the interquartile range of $38 is much larger than the median. Another less than ideal
measure of performance is top speed. The mean WTP for an additional 1 mph of top speed is
$100 and the median is $54, indicative of a mild skewness of the distribution of the 9 estimates.
13 Of the 101 original performance estimates, we dropped 33 estimates that could not readily be converted to
reduction in 0-60 acceleration time. There are a number of measures, each of which generally uses units with
low numbers of observations: percent improvements in acceleration (total of 6); braking distance (1); cylinders
(2); displacement (3); "high" or "low" performance (total of 8); horsepower when measured in other units (% of
base vehicle or change, hp/cid; total of 12); and turning circle (1).
Central
5-22
-------
The interquartile range is $86. All four papers that estimated a WTP for top speed used data from
stated preference surveys.
As with fuel economy, the effective sample size can be increased by converting to a
common metric when it is reasonable to do so. Willingness to pay for seconds 0-30 can be
approximately converted to WTP for seconds 0-60 by dividing by 2.5. In general, it takes longer
to accelerate from 30-60 mph than from 0-30 mph so the conversion factor should be greater
than 2.0. The ratios of 0-60 to 0-30 mph acceleration times for 15 recent model year GM, Ford
and Chrysler vehicles measured by the State of Michigan (2016) averaged 2.54 with a standard
deviation of 0.1. The ratio of rated engine horsepower to vehicle weight has been shown to be an
accurate predictor of 0-60 mph acceleration times (EPA, 2015). EPA (2015) provides 0-60
acceleration times and hp/wt ratios for light-duty vehicles by model year from 1978 to 2014. A
power function fit of hp/wt to seconds 0-60 mph produced the following equation:
hp/wt = 0.3542(seconds 0-60)"0 88 R2 = 0.97 (5-1)
Solving the equation for the change in seconds 0-60 corresponding to an 0.01 increase in
hp/wt from the 1995 to 2014 average for light-duty vehicles (EPA, 2015, Table 3.5) gives an
approximate value for the reduction in 0-60 mph acceleration time of 1.68 seconds. In Table 5-3,
the WTP for seconds 0-30 mpg is converted to WTP for seconds 0-60 by dividing by 2.5. The
WTP for a 0.01 increase in hp/lb is converted to WTP for seconds 0-60 by dividing by 1.68.
Even after conversion to seconds to accelerate from 0-60 mph, the hp/lb metric differs
from the 0-30 and 0-60 metrics. The median estimate based on hp/lb is only $198, only one fifth
of the median willingness to pay implied by the 0-30 and 0-60 metrics. Six of the eleven 0-30
mph estimates were inferred from stated preference survey data, as were three of the eight 0-60
estimates. All of the more numerous hp/lb estimates are based on market sales data or revealed
preference survey data. All but the 0-60 mph estimates are skewed (Table 5-3). The large
difference between the mean and the median of the 0-60 mph estimates is due to the small
sample size.
The distribution of central WTP estimates for the 0-30 mph, 0-60 mph and hp/wt metrics
converted to 0-60 seconds is shown in Figure 5-12. There is no obvious reference point for WTP
for acceleration performance.
5-23
-------
Table 5-3. Willingness to Pay for a One Second Decrease in 0-60 mph Time: Combined
0-30, 0-60, and hp/lb Normalized Metrics
Standard Number of
Native Attribute
Mean
Deviation
Skewness
Median
P75-P25
Minimum
Maximum
Observations
Seconds 0-30 mph*
$1,045
$1,123
-0.46
$1,141
$609
-$1,547
$3,288
11
Seconds 0-60 mph
$1,096
$627
-0.01
$1,183
$498
$35
$2,200
8
hp/lb**
$880
$1,449
1.91
$198
$1,558
-$860
$5,544
29
*A one second reduction in 0-30 mph acceleration is assumed to correspond to 2.5 seconds reduction for 0-60 mph
acceleration time. Thus, the WTP for a one second reduction in 0-30 mph time is divided by 2.5 to obtain the
value of a one second reduction in 0-60 mph time.
**An increase of 0.01 hp/lb at the 1995-2014 average hp/wt of 0.0507 is estimated to correspond to a reduction in
0-60 mph time of 1.68 seconds. Thus, the WTP for an increase of 0.01 hp/lb is divided by 1.68 to estimate the
value of a one second reduction in 0-60 mph time.
Figure 5-12. Frequency Distribution of WTP Estimates: Normalized 0-60 Times
WTP for a 1 Second Decrease 0-60 mph
-2000.00 0.00 2000.00 4000.00 6000.0C
2015 dollars
Although there may be greater consistency in the measures of central tendency for
performance than for fuel economy, the dispersion of estimates is still large relative to the central
tendency measures. As was the case for fuel cost, the performance WTP estimates based on
market data indicate lower WTP than those based on stated or revealed preference survey data
(see Table 5-4). Tests for differences in the median estimates by data type rejected the null
hypothesis of equal medians at the 0.03 level.
5-24
-------
Table 5-4. Comparison of Stated and Revealed Preference Estimates of WTP for One
Second Decrease in 0-60 mph Time
Willingness to Pay for One Second Decrease in 0-60 mph Time by Type of Data—Normalized Metrics
Data Type
Mean
Std. Dev.
Skewness
Median
P75-P25
Minimum
Maximum
N. Obs.
Stated Preference
$918
$892
-2.4
$1,227
$202
-$1,547
$1,514
10
Revealed Preference
$1,838
$1,826
1.05
$1,380
$2,073
$22
$5,544
8
Revealed and Stated
Preference (Combined)
$1,050
$2
0
1,050
$3
$1,049
$1,052
2
Market Data
$723
$1,199
1.81
$198
$1,267
-$860
$4,946
26
Literature Review
$497
$51
0
$497
$72
$461
$533
2
Total
$947
$1,198
1.0
$950
$1,135
-$1,547
$5,544
48
The variation in estimates of WTP for acceleration performance are shown in
Figures 5-13 and 5-14. In most cases, the variation from low to high for a given estimate (the
slope of each line) is far smaller than the variation across estimates (vertical spread of the lines).
The scale of the variation across estimates hides some of the relative variation within an
estimate. For example, in one case the high estimate is $15 while the low estimate is less than
half that, $7, but the set appears to be a level line in Figure 5-13. On the other hand, it is not
uncommon to find a range of estimation error on the order of $1,000 (Figure 5-14).
Figure 5-13. WTP for One Second Decrease in 0-60 mpg Time: Preference Heterogeneity
WTP for 1 Second Reduction in 0-60 mph Acceleration Time: Range
is +/-1 Standard Deviation
Preference Variation
$8,000
$6,000
$4,000
$2,000
$0
-$2,000
-$4,000
-$6,000
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
High
5-25
-------
Figure 5-14. WTP for One Second Decrease in 0-60 mpg Time: Estimation Error
$8,000
$6,000
$4,000
$2,000
$0
-$2,000
-$4,000
WTP for 1 Second Reduction in 0-60 mph Acceleration Time: Range is +/-1
Standard Error
Estimation Error
Centra
Low
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
5.6 Pollution
We found 19 estimates of WTP for emissions reductions. Units were converted from
WTP for a variety of different levels of vehicle emissions reductions relative to a traditional
gasoline engine into WTP for a consistent measure, defined as a 10% decrease in emissions
relative to a contemporary gas vehicle to allow for comparison across studies. We assumed WTP
to vary linearly for this transformation (e.g., multiply by 10 to convert from a 1% reduction or by
0.2 for a 50% reduction), though recognizing that may be a strong assumption. Central estimates
range widely from -$66,982 to $168,536. Studies using older stated preference data from 1993—
1996 tended to find higher willingness to pay for emissions reductions. Newer studies using data
from 2009 and 2012 (Hidrue et al., 2011 and Tanaka et al., 2014, respectively) found WTP
values ranging from $297 to $582 for a ten-percent reduction in emissions. The Hidrue et al.
study found that WTP for that 10 percent reduction generally increased for higher-level
reductions (e.g., 95% lower emissions), suggesting the assumption of linearity when converting
to common units may hide variation in consumer tastes. Additional surveys or the study of
market data could provide additional insight on current preferences.
5.7 Range
We use 40 estimates of WTP for range in the literature, all of which were estimated in
terms of dollars per mile or based on a certain number of miles of range. There were three papers
5-26
-------
(Helveston et al., 2015; Hidrue et al., 2011; and Parsons et al., 2014) that had a total of 27
coefficients for range. However, all the range variables in these three studies were (0,1)
representing, for example, a vehicle with a 75 mile range, a 150 mile range and a 200 mile range
(the exact numbers actually vary across the papers). We do not feel there is a reasonable way to
estimate the marginal value of adding a mile of range ($/mi) for the lowest range included in any
of these papers. The reason is that we are estimating value per mile of changes in range, but do
not think that it is reasonable to calculate value per mile in that way when examining the
difference between a range of 0 miles (useless for transportation) and a positive range. Instead,
we include only estimates for ranges where we can take the difference between values reported
for different levels of range and calculate the average change in value per mile of range for that
change in range. This is an average value over the difference in range rather than a true marginal
value, but the best estimate we could calculate and a meaningful measure of changes in range
valuation with the magnitude of range.14 The observations produce a mean of $86 per mile, and
an interquartile range of $54-117. Greene (2001) derived a value of range based on the
assumption that is it a savings in refueling time and effort and shows that its value is a function
of the inverse of range. None of the papers estimate the value of range in this form, however. In
this framework, variations in the value of range would depend chiefly on the consumers' value of
time and the range of the vehicles under consideration.
The observations are drawn from 16 different papers and all but one (Greene, 2001)
utilize survey data. Approximately half of the estimates use the same two-round phone survey in
California, first published by Brownstone et al (1996). Several other estimates draw from
California surveys. Authors primarily employ MNL and MXL estimation strategies, though no
significant divergences are obvious by either estimation strategy, data type, or time frame.
In models that permit variation in taste across a population, we find varying levels of
heterogeneity (Figure 5-15). Several studies have high and low values that cross zero and
represent a spread of several hundred dollars per mile. As several of these observations utilize
the same data (i.e., the Brownstone et al. [1996] survey), these divergences seem to emerge from
the formulation employed by the authors. Figure 5-16 summarizes variation in estimates for
those available values that reflect estimation error.
14 The effect of this adjustment on the number of observations included from Helveston et al. (2015), Hidrue et al.
(2011), and Parsons et al. (2014) is that we drop 9 of their 27 reported observations (lowest range with values
reported) and include the 18 where we can calculate differences and convert to a $/mile of range estimate
consistent with the other observations. This adjustment is reflected throughout all figures and tables in the report
presenting summary statistics.
5-27
-------
5.8 Size
5.8.1 Footprint
We remove one extreme value of $680,000 per square foot from our sample of 19
observations and produce a trimmed mean of $3,398 per square foot, and a standard deviation of
$4,381. The distribution is somewhat balanced with an interquartile range of $477-$4,411
(Figure 5-17).
For one paper, units are unclearly marked so we assume them to be square feet as
standard in most market data sources.15 Our trimmed sample includes two more extreme cases,
which we retain nonetheless as their inclusion does not skew the findings regarding the central
tendency of the sample.
Figure 5-15. WTP for Range in $/mile: Preference Variation, +/1 One Standard Deviation
WTP for an Additional Mile of Vehicle Range:
Low to High is +/-1 Standard Deviation
Preference Variation
$1,000
$800
$600
$400
$200
$0
-$200
-$400
-$600
Centra
Low
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
15 We requested feedback from the study author, including a specific clarifying question about the units used in their
study, but did not receive a response.
5-28
-------
Figure 5-166. WTP for Range in $/mile: Estimation Error, +/1 One Standard Error
$800
$600
$400
$200
$0
-$200
-$400
-$600
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
Figure 5-177. Trimmed Central WTP Estimates for Vehicle Footprint
WTP for Footprint Increase: Trimmed
aj
3
CT
a)
-*-i 1 1 1 1-
-5000.00 0.00 5000.00 10000.00 15000.01
2015 dollars
WTP for an Additional Mile of Vehicle Range:
Low to High is +/-1 Standard Error
Estimation Error
Central
5-29
-------
The eighteen estimates in the trimmed sample are based primarily on market data (there
is one observation based on RP survey data) and reflect a variety of estimation strategies,
including BLP, MNL, NMNL, and MXL. Six of these estimates come from Petrin (2002), which
uses different subsets of sales data from 1981-93 to produce WTP values ranging from $3,500-
$14,500. Haaf et al. (2014) produce four estimates of WTP for vehicle footprint using different
modeling formulations; their outputs cluster more closely between $900-$l,500.
Population heterogeneity is relatively limited across the studies that allow variation in
taste. The largest difference between low and high values is $1,842, but for most, the range is a
few hundred dollars (Figure 5-18). We generally find that there is low variation in valuation of
footprint size across populations, and that additional square footage is positively valued.
Figure 5-188. WTP for Footprint: Preference Variation, +/- One Standard Deviation
$16,000
$14,000
$12,000
$10,000
$8,000
$6,000
$4,000
$2,000
$0
$/ft2 Footprint: Range is +/-1 Standard Deviation
Preference Variation
Low
Central
High
Note: Each line represents the range calculated for an observation of an estimate of WTP for a particular vehicle
attribute (combination of study and model specification; a single study can have multiple observations).
5.8.2 Luggage Space
From an initial pool of twelve estimates for luggage space, we remove one extreme value
near $30,000. The remaining estimates produce a balanced distribution centered around $1,445
per cubic foot. We see a moderate positive skew, with a median of $1,100, and an interquartile
range between $619 and $2,365.
The trimmed estimates draw from four different studies, one of which is a literature
review (Greene, 2001). The latter produces a lower estimate of $270 per cubic foot. One other
estimate is of an unanticipated negative sign. The majority of remaining estimates are from Liu
5-30
-------
et al. (2014)'s study using revealed preference survey data and a MNL modeling strategy. Liu et
al. (2014) produce estimates based on household fleet size and income segment. For the most
common cases (1 or 2 cars, medium income) and nearby cases, WTP ranges between $1,000-
$2,000 per cubic foot. One other estimate from McCarthy (1998) using 1989 data produces a
similar value of WTP. In general, consumers positively value additional space, with preferences
strengthening for households owning fewer vehicles in the study by Liu et al (2014); households
owning only one car were estimated to have a greater WTP than households with multiple
vehicles.
5.8.3 Weight
Although weight may have value to some consumers, it is often used by modelers as a
convenient proxy for vehicle capacity and size which complicates the interpretation of WTP
estimates. We remove one outlier from an initial pool of 19 observations and produce a trimmed
mean of $5.70 per pound, from an initial mean of $10 per pound. Despite trimming, the
distribution is still highly skewed. There are 25% of estimates between $0.41 and $0.50, with the
next quartile spreading over $0.51 to $10.23 dollars (Figure 5-19). The cluster of values near
zero comes solely from Klier and Linn (2012), a study using a linear instrumental variable least
squares technique. They produce varying estimates based on inclusion and exclusion of different
covariates. The other observations of WTP for weight come from hedonic studies and one MNL
model. These estimates tend to fall between $10-15 per pound, and draw from a mix of market
and revealed preference survey data. Within this further restricted sample of data points,
formulations focusing solely on trucks tend to produce lower WTP estimates. Disregarding the
Klier and Linn estimates, we find tentative consensus in the literature of consumer weight
valuation around $10-15 per pound. These figures are limited by the small sample of studies. We
are unable to estimate taste heterogeneity because our sample has only fixed coefficient models.
Modelers tend to employ weight as a proxy for qualities that may be more difficult to quantify,
such as size, comfort or handling.
5-31
-------
Figure 5-199. Trimmed Distribution of Central WTP for Weight
WTP for Weight: Trimmed
0.00 5.00 10.00 15.00 20.00 25.00
2015 dollars
5-32
-------
SECTION 6.
DISCUSSION: WHY I II I LACK OF CONSENSUS ON WTP?
This project suggests some challenges for research that models consumer demand for
vehicles and their attributes. Modeling results seem to be highly sensitive to a number of factors,
including sources of underlying data, modeling techniques, included and omitted variables, and
functional form. As discussed above, results vary widely not only across studies, but even within
individual papers. This field of research will inspire more confidence in its policy relevance if it
can identify greater convergence of values, or at least greater understanding of the factors that
contribute to the wide variation in values.
The most conspicuous feature of the central tendency WTP estimates for nearly all
attributes is their dispersion. This generalization holds true for the trimmed as well as the
untrimmed estimates. Outliers are common but not numerous, and removing them still leaves a
diverse set of estimates. For 24 of the 34 individual attributes and both of the aggregate measures
shown in Table 5-1, one standard deviation of the trimmed distribution is larger than the mean
value, and in only one case is it less than half of the mean. The distributions of estimates are
skewed, as a rule, making the medians better measures of central tendency than the means. These
two facts make it difficult to interpret overall measures of central tendency, even when at least
one of the measures of central tendency corresponds reasonably well to a constructed reference
point estimate.
A non-negligible number of the central WTP estimates violate prior expectations, e.g.,
willingness to pay less for a vehicle with improved fuel economy or higher horsepower/weight.
Still, in the great majority of cases the signs of WTP estimates agree with prior expectations.
The most commonly estimated attribute value categories, fuel cost and acceleration
performance, were examined in greater detail, taking advantage of the ability to convert variables
to a comparable metric. In the case of fuel costs, median estimates based on stated preference
survey data indicated a greater preference for reduced fuel costs compared to revealed preference
data. Estimates based on market sales data indicated an even lower willingness to pay for lower
fuel costs. However, the same result was not found for the estimates of WTP for acceleration
performance. It appears that the relationship between stated and revealed preference WTP
estimates differs depending on the attribute in question.
High and low estimates for each observation were calculated based on +/- 1 standard
error in the case of fixed parameter models and +/- 1 standard deviation in the distribution of
preferences in the case of random coefficient models or models in which preferences depended
6-1
-------
on household income. These results indicate both wide variation in consumers' preferences and
substantial uncertainty in estimation.
This report has focused on presenting descriptive statistics for the various attributes found
in the literature and has not conducted systematic analysis of why those differences arise.
Subsequent research will attempt to analyze why WTP estimates vary both across studies and for
model formulations within studies. It is possible that further meta-analysis of the WTP estimates
may help explain their variability.
Probably the most salient feature of the WTP estimates found in the literature is their
variation. On the one hand, variability is to be expected because circumstances and preferences
do vary from person to person and over time. On the other hand, wide variation is seen in
estimates of central tendency for entire populations across studies, and many estimates designed
to reflect differences across individuals are puzzling and contradict expectations. Studies that
provide estimates from various model formulations or estimation methods using the same data
set may provide insights into the reasons for the great variability of WTP estimates.
Haaf et al. (2014) estimated a variety of discrete choice models, including MNL, NMNL
and random coefficient models, using the same data set of sales of makes and models in the U.S.
from 2004-2006. The estimated coefficients of vehicle price (in 10,000s of $) ranged from -0.19
to -0.61, except for one model estimated using the method of BLP which produced a coefficient
estimate of-1.56. Coefficients of vehicle attributes were even more varied. In the six models
that represented fuel cost as gallons per mile, three coefficients had a negative sign (as expected)
while three had a positive sign. In addition to model form and estimation method, the models
differed with respect to the measures of vehicle size included. Those with positive signs used
width and length*width/height while those with negative signs included only length*width.
Length*width/height has been used in a few studies as a proxy for "style." In fact, it measures
the flatness of a vehicle. The variables were chosen based on objective measures of model fit and
adequacy, rather than the modelers' judgment. The results illustrate how inferences about the
values of attributes based on aggregate, revealed preference data can be strongly influenced by
the selection of attributes to include, how they are measured, model form and estimation method.
Klier and Linn (2012) provide results for 14 different estimations of vehicle choice
models at the make and model level using U.S. sales data for 2000-2008. The authors compare
estimates made by means of ordinary least squares (OLS) with estimates using two different sets
of instrumental variables (IV). The OLS estimates are very different from the IV estimates: the
price coefficient is about one fifth as large, the coefficient of cost per mile is positive and that of
hp/wt is more than two orders of magnitude smaller than that obtained by the IV methods. The
6-2
-------
authors clearly state that they believe that the OLS estimates are biased. Nonetheless, the
exercise demonstrates how strongly the method of estimation can influence results. So also can
the instruments of the IV method. Instruments similar to those used by BLP produced a
coefficient estimate for hp/wt of 9.53, while those created by the authors based on engine
characteristics produced an estimate of 38.75. Other coefficients do not vary as much between
the two sets of IVs (e.g., the coefficient of log of price is -1.86 for the BLP instruments and
-1.28 for the engine instruments) but the focus of the Klier and Linn paper is understanding the
tradeoff between performance and fuel economy so a difference greater than a factor of three is
important.
Some estimates may be robust to changes in a model's formulation while other are highly
sensitive. Klier and Linn (2012) also compare seven sets of coefficient estimates from models
differing with respect to variables included, except for one that employed a different error
structure. The four that measured performance as hp/wt had similar coefficient estimates for both
hp/wt (47.20, 38.75, 40.74 and 42.18) and the log of price (-1.79, -1.28, -1.34 and -1.49, in the
same order). On the other hand, the four models showed much greater differences in the
estimated coefficients of fuel cost per mile (-13.24, -11.05, -22.94 and -3.29). Models
including hp and weight as separate variables produced a different set of estimates for the
coefficients of the log of price (-0.99, -0.60 and -1.06) and fuel cost (-3.95, -0.98 and +0.43)
than when hp/wt was used. A consequence of the sensitivity of estimates to choice of variables
and instruments is that the Klier and Linn estimates of the WTP for 0.01 hp/lb fall into two
clusters: ($303, $264, $303 and $283) and ($52, $51 and $8). The authors express a preference
for the higher WTP estimates, which are consistent with their preferred model formulation using
IV estimation and the engine instruments they constructed.
Augmenting aggregate revealed preference data with data on consumer attributes can also
lead to dramatic changes in WTP estimates. Random coefficient models are compared with fixed
coefficient logit models by Petrin (2002) who also augments vehicle sales data with data from
the Consumer Expenditures Survey (CES) describing the average attributes of consumers
purchasing new vehicles by income group. Vehicle price is specified as the log of (consumer
income minus vehicle price). Four estimation methods were compared: OLS, IV, and a
Generalized Method of Moments algorithm, all used to estimate a random coefficient (RC) logit
model and the same model augmented (ARC) by fitting aggregate data on new vehicle
purchasers from the CES. In the RC models, separate price coefficients are estimated for three
income tertiles. Petrin reports that a Wald test rejects the OLS and IV estimated fixed coefficient
models in favor of the random coefficient models. While many of the coefficient estimates are
similar among the four models the estimated coefficient of Miles/Dollar (miles per gallon/fuel
6-3
-------
price) for the OLS, IV, RC, and ARC models are, respectively: 0.18, 0.05, -0.54 and -15.79.
Only the estimated coefficient of the ARC model is statistically significant, however. The
coefficient of miles per dollar is expected to be positive, and although the RC model estimates
are the mean of a probability distribution the estimated standard deviations are only 0.04 (RC)
and 2.58 (ARC), implying that nearly all consumers would prefer fewer miles per dollar.
Petrin's preferred model (random coefficients estimated by augmenting market sales data
with data from the CES) produced mean WTP estimates for miles per dollar for the lowest,
middle and highest income tertiles of-$33,890, -$13,313, and -$14,018, respectively. The
model estimated without using the CES data produced mean WTP estimates of-$1,771, -$1,192
and -$436 for the income tertiles. Estimated standard deviations describing taste heterogeneity
implied one standard deviation ranges of less than +/- 20% for the ARC model and less than +/-
10% for the RC model. For the ARC model, the mean WTP estimates for an increase of 0.01
hp/lb. (faster acceleration) were: -$607, -$239, and -$251 for low, middle and high income
tertiles. For the RC model the corresponding WTP estimates were: $1,115, $751 and $275, a
reversal of sign. However, only the standard deviation coefficient in the ARC model was
statistically significantly different from zero. In that model +/- one standard deviation of WTP
implied ranges of (-$1,539 to +$324), (-$605 to +$127) and (-$637 to +$134) for the three
income tertiles, implying that most consumers would prefer slower acceleration.
Modelers are aware of the challenges to obtaining robust estimates of attribute
coefficients in vehicle choice models. Brownstone et al. (2000) point out several severe
shortcomings of models estimated solely with RP data: 1) high collinearity and limited variation
in vehicle attributes, 2) problems defining choice sets from the thousands of makes, models,
drivetrain and trim configurations, and 3) problems accurately linking vehicle attributes to the
vehicles described by households. The authors note: "Under these difficult conditions RP model
estimates are often unstable, and can have theoretically incorrect signs." Because stated
preference surveys can be designed to avoid strong correlations among the attributes of an
alternative, can attempt to minimize the effects of unobserved attributes, and can clearly define
choice sets using a rigorous experimental design, they should be more likely to produce
consistent results. One hundred and forty-eight of the 786 WTP estimates considered in this
study come from five research papers that used the same survey of stated preferences for
alternative fuel vehicles conducted in California in 1993 (McFadden and Train, 2000;
Brownstone et al., 2000; Brownstone and Train, 1999; Kavalec, 1999; Brownstone et al., 1996).
A comparison of these studies sheds some light on the strengths and weaknesses of models
estimated using SP data. The studies are briefly described below, followed by a comparison of
their WTP estimates.
6-4
-------
The California survey began with an initial computer-aided telephone interview of 7,387
households, 4,747 of whom completed a follow-up mailed survey. Respondents were asked to
choose among four fuel types: gasoline, compressed natural gas, methanol and battery electric.
For each, two body types were offered, described by six attributes. Respondents were instructed
to assume all attributes not specifically described were identical for all vehicles. An orthogonal
main effects design was used to structure the choice alternatives.
Brownstone et al. (1996) estimated MNL models to predict vehicle transactions (add,
replace, dispose of a vehicle) rather than vehicle holdings. Separate models were estimated for
1,153 households owning one and 1,156 households owning two vehicles. Vehicle price was
interacted with household income category and the presence and age of children. Because WTP
estimates are derived by dividing the derivative of the utility function with respect to an attribute
by its derivative with respect to income, the interactions create a multiplicity of WTP estimates
for different income groups and household compositions. Estimates of the willingness to pay for
a $1 present value decrease in operating cost are shown in Figure 6-1. High and low WTP
estimates reflect +/- one standard error of the operating cost coefficient. The number of vehicles
owned by the household is shown in the horizontal axis labels, and luxury or lux indicates the
household owns at least one luxury vehicle. Although several of the estimates are close to $1, as
would be expected, others are negative, suggesting that at least three categories of consumers
would prefer higher operating costs. This result is due to positive coefficient estimates for
vehicle purchase price for three of the household categories. Positive coefficients on vehicle
price might represent a genuine preference to pay more for a vehicle (e.g., a Giffen good), but
more likely indicate shortcomings of the survey design, model specification and estimation. The
positive price coefficients create similarly anomalous WTP estimates for other attributes for
these household categories.
Kavalec (1999) used the 1993 California data to explore the effects of an aging
population on demand for gasoline through their vehicle purchase decisions. The focus was on
estimating the influence of age on consumers' preferences for different vehicle attributes.
Random coefficients were estimated for four fuel types and two vehicle size classes. The results
implied that the values of fuel cost and acceleration were only slightly affected by the
respondent's age but the value of top speed decreased steeply and linearly with increasing age
and the value of range first increased with increasing age but then decreased rapidly beyond the
age of 65.
6-5
-------
Figure 6-1. Willingness to Pay for a $1 Present Value Decrease in Operating Cost: 1- and
2-vehicle Households (Estimates derived from Brownstone et al., 1996)
Willingness to Pay for a $1 Present Value Decrease in Operating Cost
1- and 2-Vehicle Households (Brownstone et al., 1996)
~
I * *
X 1
~
I 1
1 1
T
(
<
1
1
Inc < $30K w Inc < $30K no $30K < Inc < Inc > $75K no Inc < $30K Inc < $30K no Inc > $30K Inc > $30K lux Inc > $30K no Inc > $30K no
child 1 child 1 $75K no child child 1 child 2 child 2 luxury child 2 no child 2 lux child 2 lux no child 2
1
Brownstone and Train (1999) used the same 1993 California survey data to estimate a
MXL model of alternative fuel vehicle choice. Because the mixed logit model represents
heterogeneity in consumers' preferences by means of random coefficients, the model includes
many fewer interactions between household attributes and vehicle attributes. The coefficient of
vehicle purchase price was assumed not to be a random variable but purchase price is divided by
the logarithm of household income and so varies systematically with income. Random
coefficients were estimated for choices between other vehicle types and electric and compressed
natural gas vehicles and for vehicle size class and luggage space.
Brownstone et al. (2000) combined the 1993 California survey data with revealed
preference data comprised of the actual purchases of 874 households who purchased a vehicle
between two waves of the survey. An MXL model was estimated. The authors note that the RP
data appeared to be essential to estimating a model that could realistically predict body type
choices and the appropriate volumes of purchases. Only the alternative fuel constants and fuel
cost were assumed to have random coefficients.
McFadden and Train (2000) estimated an MXL model using the 1993 California survey
data. They used the same variables and transformations as Brownstone et al. (1996). Random
coefficients were estimated for choices between other fuels and EVs and CNG vehicles, and for
size, luggage space, operating cost and refueling station availability. The latter two were
identified by new specification tests derived by the authors and expand on the set of random
coefficients estimated by Brownstone and Train (1999).
6-6
-------
Estimates of willingness to pay for fuel cost from the five studies are shown in Table 6-1.
Comparable estimates for Brownstone et al. (1996) are shown in Figure 6-1 (highlighted
individually to show the variation present within that single study). All but one of the models is
mixed logit. The ranges of WTP estimates are based either on +/- one standard error of the
coefficient estimate if an MNL model or on +/- one standard deviation of the random coefficient
estimate if an MXL model was used. In the case of the Kavalec model, +/- one standard
deviation of the age distribution of respondents was used since the coefficient was interacted
with the respondent's age. The central tendency estimates are relatively consistent, ranging from
$2,564 to $3,239 for the four models. In general, this seems to support the claim that SP survey
data should provide more consistent estimates than RP data. The low and high ranges are less
consistent, with the High estimates for MXL models ranging from $2,147 to $10,308. The same
models' Low estimates range from $2,058 for Kavalec's age-interacted WTP to -$4,275 for
McFadden and Train's random coefficient model. In the case of Brownstone et al. (1996) we
extracted the lowest, median and highest from the full set of estimates shown in Figure 6-1.
Table 6-1. Willingness to Pay for $0.01/mile Decrease in Fuel Costs from Studies Using
the Same CA Survey
WTP for SO.Ol/mile
Paper
Model
Variation
Low
Central
Hiuh
Brownstone & Bunch, 2000
MXL
Std.
Dev.
1,437.73
2,564.42
3,691.11
Brownstone & Bunch, 2000
MNL
Std.
Err
2,098.43
3,239.98
4,381.53
Brownstone & Bunch, 2000
MXL
Std.
Dev.
-4,068.99
2,799.65
9,668.29
Brownstone & Train, 1999
MXL
Std.
Dev.
1,012.73
2,749.85
4,486.98
Brownstone et al., 1996
MNL
Std.
Err.
5,440.98
7,739.32
10,037.66
Kavalec, 1999
MXL
Age
2,057.94
2,706.64
2,147.42
McFadden & Train, 2000
MXL
Std.
Dev.
-4,274.85
3,016.68
10,308.20
Estimates of WTP for a 1 second decrease in 0-60 mph acceleration time for the five
studies are shown in Table 6-2. The range of central tendency estimates is relatively larger than
that of the fuel cost estimates in Table 6-1: -$1,547 to $3,288. Only one of the nine Low-High
ranges includes zero. The effect of the few positive price coefficients in Brownstone et al. (1996)
can be seen in the set of positive estimates ranging from $690 to $3,288, implying that that
market segment would prefer vehicles with slower acceleration.
6-7
-------
Table 6-2. Willingness to Pay for a 1 Second Decrease in 0-to-60 Acceleration Time
WTP for 1 Second Decrease in 0-60 mph
Acceleration Time
Paper
Model
Variation
Low
Central
High
Brownstone & Bunch, 2000
MXL
Std.
Err.
$772.73
$1,048.70
$1,324.67
Brownstone & Bunch, 2000
MNL
Std.
Err.
-$519.61
$3,287.55
$7,094.71
Brownstone & Bunch, 2000
MXL
Std.
Err.
$776.74
$1,066.25
$1,355.76
Brownstone et al., 1996
NMNL
Std.
Err.
$267.44
$690.17
$1,112.90
Brownstone et al., 1996
NMNL
Std.
Err.
-$2,494.35
-$1,546.88
-$599.42
Brownstone et al., 1996
NMNL
Std.
Err.
$1,077.34
$1,513.61
$1,949.89
Brownstone & Train 1999
MXL
Std.
Dev.
$265.02
$1,299.02
$2,333.02
Kavalec, 1999
MXL
Age
$674.62
$1,140.64
$1,606.66
McFadden & Train, 2000
MXL
Std.
Err.
$1,009.62
$1,261.54
$1,514.43
There is similar variation among the five studies with respect to alternative fuel vehicle
attributes that are likely to be unfamiliar to respondents. Table 6-3 shows estimates of
willingness to pay for alternative refueling station availability equal to that of gasoline. Central
tendency estimates range from $94 to $314 per vehicle. The MXL central tendency estimates are
quite similar, ranging from only $108 to $144. Two of the MXL models indicate a negative Low
WTP estimate, suggesting that some consumers would prefer not to have greater availability of
refueling stations. The High WTP estimates range from $143 to $346. The central tendency
estimates seem low, especially for methanol and CNG vehicles which cannot operate without
fueling stations offering their fuel. Since that would render the vehicles useless from a practical
point of view, one might expect WTP for full availability versus no availability to be on the order
of the full price of a vehicle. The explanation for this may lie in the fact that the studies enter fuel
availability linearly, whereas Nicolas et al. (2004) have shown that the cost of limited fuel
availability in terms of access time is exponential in relative availability.
6-8
-------
Table 6-3. Willingness to Pay for Alternative Fuel Availability Equivalent to Gasoline
WTP for Availability =
Gasoline
Paper
Model
Variation
Low
Central
High
Brownstone & Bunch, 2000
MXL
Std. Err.
$73.08
$108.43
$143.78
Brownstone et al., 1996
NMNL
Std. Err.
$32.84
$179.12
$325.40
Brownstone et al., 1996
NMNL
Std. Err.
$179.94
$313.53
$447.11
Brownstone et al., 1996
NMNL
Std. Err.
$23.02
$93.53
$164.03
Brownstone et al., 1996
NMNL
Std. Err.
$44.94
$182.56
$320.18
Brownstone et al., 1996
NMNL
Std. Err.
$80.20
$231.33
$382.47
Brownstone et al., 1996
NMNL
Std. Err.
$36.14
$197.11
$358.08
Brownstone & Train 1999
MXL
Std. Dev.
-$33.09
$140.62
$314.33
Kavalec, 1999
MXL
Std. Err.
$106.17
$144.35
$182.54
McFadden & Train 2000
MXL
Std. Err.
-$111.58
$117.44
$346.43
There is greater variation in the estimates of the value of reducing the emissions of a
typical gasoline vehicle to zero (Table 6-4). The negative estimates (preference for higher
emissions) are shown in italics and come from Brownstone et al.'s (2000) model estimates that
used the RP survey wave of the 1993 California Survey. All the others are based on either the SP
data or a combination of the two. Perhaps one explanation for the reversal of signs on WTP for
reduced emissions for the SP results is the tendency of survey respondents to provide answers
they believe are the desired answers or the answers that reflect well on them (a.k.a., social
desirability bias).
The California SP studies all used the same database and most used the same set of
variables. Differences are due to small variations in the formulation of variables and estimation
methods. As expected, estimates of central tendency based on the SP surveys were more
consistent than seen above when RP surveys were used. WTP estimates for less familiar
attributes varied more than the estimates for familiar attributes. Estimates of the variation of
preferences across the population sometimes included counterintuitive preferences (e.g.,
preferring less fuel availability). A study that combined RP data with the California SP data
reversed the sign on WTP for emission reductions, suggesting that "yea-saying" bias may be
present in inferences about certain attributes, although there are other possible explanations.
6-9
-------
Table 6-4. Willingness to Pay for Reducing the Emissions of a Typical Gasoline Vehicle to
Zero
WTP for Reduction to Zero Emissions
Paper
Model
Variation
Low
Central
High
Brownstone et al., 1996
MXL
Std. Err.
$76,777
$168,536
$260,294
Brownstone et al.. 1996
NMNL
Std. Err.
$103,065
$144,803
$186,540
Brownstone et al., 1996
NMNL
Std. Err.
$8,357
$76,601
$144,846
Brownstone et al.. 1996
NMNL
Std. Err.
-$72,273
$8,213
$88,699
Brownstone et al., 2000
NMNL
Std. Err.
$47,823
$75,954
$104,085
Brownstone et al., 2000
NMNL
Std. Err.
-$83,602
-$66,982
-$50,362
Brownstone et al., 2000
NMNL
Std. Err.
$51,604
$81,736
$111,867
Brownstone & Train, 1999
MXL
Std. Dev.
-$28,709
$145,004
$318,716
Kavalec, 1999
MXL
Std. Err.
$102,201
$141,968
$181,735
McFadden & Train. 2000
MXL
Std. Err.
$100,047
$132,461
$164,892
6-10
-------
SECTION 7.
CONCLUDING OBSERVATIONS
A goal of this study was to identify consensus estimates of the values of various vehicle
attributes through a comprehensive analysis of empirical estimates in the published literature.
Unfortunately, we have found very little useful consensus. Frequently, standard deviations are of
the same magnitude as, or greater than, mean values. In general, medians differ markedly from
means and interquartile ranges are as large, or larger than, medians. Typically, estimates based
on stated preference surveys do not agree with estimates based on revealed preference surveys or
market sales data. For example, fuel cost is the most frequently included attribute in vehicle
choice studies after purchase price. The mean estimate of central tendency estimates of WTP for
a one cent per mile decrease in fuel costs, based on 27 stated preference estimates, is $3,914 with
a standard deviation of $12,655. The mean of 15 estimates using revealed preference data is
-$66,796 ($3,609 trimmed) with a standard deviation of $272,752 ($6,580 trimmed). The
medians and interquartile ranges for stated and revealed preference studies are, respectively,
$1,889 and $2,817 (SP), $583 ($692 trimmed) and $3,230 ($3,192 trimmed) (RP). This is the
variability of the central tendency estimates from the 42 (41 trimmed) estimates and does not
reflect standard errors of estimation nor preference heterogeneity in the population. Some
consistency can be found in the fact that most estimates are positive (consumers would prefer
lower fuel costs). This "consensus" however, encompasses such a wide range of values that it is
of little use for informing policy decisions. Unfortunately, the results for other attributes are
often just as divergent.
In the authors' judgment, the magnitude of uncertainty exhibited in the recent literature
reflects the inherent difficulty of estimating how much consumers value vehicle attributes. Motor
vehicles are complex, multi-attribute commodities. Consider just one of the more important
attributes for consumers: safety. Dimensions of safety include frontal, side, offset and rear
crashworthiness, occupant protection for the driver as well as front and rear passengers, rollover
propensity, handling, braking distance and anti-lock braking, traction control and, more recently,
an increasing array of intelligent warning and control systems. As a rule, it is not possible to
include all the relevant safety dimensions in a statistical model. Furthermore, safety is just one of
several important dimensions that include price, capacity to carry people and cargo, reliability,
performance, fuel economy, cost of maintenance and insurance, comfort, style, etc. All of these
major dimensions are themselves multi-dimensional. As a general rule, it will not be possible to
include all these measures nor will it be possible to find metrics that accurately reflect
consumers' perceptions. Styling is important to consumers and yet few studies attempt to
7-1
-------
explicitly include it.16 In statistical terms, there will inevitably be omitted variables and errors in
variables. Finally, many of the attributes are correlated. Performance, comfort and size, for
example, are correlated with each other and with purchase price. Omitted variables, errors in
variables and correlated variables cause coefficient estimates to be biased. The nature of the
biases will depend on which variables are included or excluded. Thus, the coefficient estimates
obtained will depend on how a model is formulated.
Finally, the model of rational economic decision making that underlies all the studies we
analyzed may not be an adequate representation of consumers' decision-making processes.
Especially when faced with multi-dimensional, complex choices, consumers often employ
simpler, heuristic choice methods (e.g., Kahneman, 2011). Consumers often focus on a small
number of key attributes and satisfice less salient ones. Behavioral psychology has shown that
the context of the consumers' decision strongly affects choices. The models reviewed generally
assume that context need not be considered.
Given the large number of possible explanatory variables, and especially in models that
include interactions between consumer and vehicle attributes, overfitting is likely. Some
statistically significant variables may be fitting quirks or idiosyncrasies in the sample data rather
than meaningful relationships. An overfitted model will statistically "explain" or fit well the data
on which it is estimated, but will not necessarily predict well beyond the sample. Haaf et al.
(2014) meticulously tested a variety of common vehicle choice model types on market sales data.
The experiment found none that could predict sales shares for the following year better than a
naive model that assumed that shares would remain unchanged, However, in that study, the
attribute-based models could predict better than the naive assumption that sales shares would
remain the same as the older year when predicting farther into the future or for new vehicle
designs.
The WTP estimates described in this report strongly suggest that the results obtained
depend importantly on decisions made by the analyst. WTP estimates for the same attribute vary
widely across and even within studies. WTP values vary with the type of data: stated preference
survey, revealed preference survey and market sales. Results are also sensitive to estimation
methods. Instrumental variables estimates are frequently strikingly different from OLS estimates.
Multinomial logit and nested logit forms with demographic variables interacted with vehicle
attributes to create heterogeneity often differ markedly from mixed logit models that represent
heterogeneity with random coefficients.
16 In a rare exception, one study included length*width/height, literally the flatness of a vehicle, as a measure of
style.
7-2
-------
This creates a dilemma for the analyst. On the one hand, theory can help distinguish
among functional forms and provide definitive expectations for signs and even magnitudes of
coefficients and WTP estimates. On the other hand, the premises embedded in theories can make
analysts susceptible to confirmation bias, ".. .the seeking or interpreting of evidence in ways that
are partial to existing beliefs, expectations, or a hypothesis in hand," usually by ".. .unwitting
selectivity in the acquisition and use of evidence" (Nickerson, 1993). The large variations in
WTP values suggest the possibility that analysts may focus on a few variables of special interest,
preferring formulations that provide values for these variables that are consistent with prior
expectations. Other variables in the model may be overlooked or unexpected values for these
variables may be explained by acknowledging that they are aliasing other attributes that are not
of interest. Alternative specifications that lead to conflicting inferences may not be presented. In
the previous section we discussed several studies that do provide numerous alternative results.
These studies provide valuable insights that help understand why estimates from different studies
can vary so markedly.
This paper has examined the willingness to pay for vehicle attributes that can be derived
from these studies. Although measures of central tendency generally agree on signs, the
variability in estimates across studies is almost always very large relative to the mean or median
of the WTP estimates for any given attribute. Further analysis of these existing studies is needed
to understand why such large differences in WTP estimates arise.
At this point we can only hypothesize about what might be causing the frequently
extreme dispersion of estimates. Our estimates come from 20 years of published literature using
data that cover an even greater period of time. While all are from the U.S., some pertain only to
California and a few others to a limited number of states or metropolitan areas. Some of the data
sources are stated preference surveys, others are revealed preference surveys or market sales, and
researchers occasionally use combinations of these. Researchers estimate different types of
models and use different estimation methods. Functional forms and the ways the same attribute
is measured differ. A statistical meta-analysis of the WTP database we have created may lead to
useful insights into the wide variability of existing estimates.
The lack of consensus we have found in the literature points to major challenges for
researchers attempting to model consumer preferences for vehicles and their attributes. Modeling
results seem to be sensitive to a number of factors, including sources of underlying data,
modeling techniques, included and omitted variables, and functional form. As discussed above,
results vary widely not only across studies, but even within individual papers. This field of
research will inspire more confidence in its policy relevance if it can identify greater
7-3
-------
convergence of values, or at least greater understanding of the factors that contribute to the wide
variation in values.
Recognizing the difficulty of the problem researchers in this area face, we offer a few
recommendations that might eventually lead to greater consensus. First, model and parameter
validation should become a key focus of future research. More studies should analyze the ability
of vehicle choice models to predict outside of the data on which they were estimated. The
robustness of coefficient estimates to alternative formulations of variables, model functional
forms and estimation methods should be an important criterion for evaluation.
We recommend that authors routinely provide WTP estimates implied by their models.
Authors have access to key information (e.g., variance-covariance matrices of coefficient
estimates or joint probability distributions of coefficients in random parameter models) that are
generally not available in published articles. In general, this enables authors to more accurately
estimate marginal WTP than we have been able to do. In this report we focused exclusively on
marginal WTP estimates. Estimates of WTP over intervals can also be calculated (e.g.,
Dimitropoulos et al., 2013) by means of logsums (e.g., Zhao et al., 2012). Routine reporting of
WTP estimates would facilitate comparisons across studies, as well as alerting researchers and
readers to possible model deficiencies.
It would also be helpful to pay special attention to aliasing effects, for example, by
identifying variables believed to be proxies for omitted variables and those believed not to be,
and providing supporting evidence. Researchers could pay greater attention to how attributes are
represented in their models and provide explicit interpretations of interactions between vehicle
and consumer attributes and the values they imply. In studies based on stated preference data,
researchers could attempt to establish how well consumers understand the attributes they are
asked to consider, and greater attention could be given to identifying potentially biased responses
and their implications. Finally, it may be useful to explore alternatives to the economically
rational, continuous trade-off model of consumer choice for understanding how consumers value
vehicle attributes.
7-4
-------
SECTION 8.
REFERENCES
1. Allcott, Hunt, and Nathan Wozny. "Gasoline Prices, Fuel Economy, and the Energy
Paradox." Review of Economics and Statistics 96, no. 5 (2014): 779-95.
2. Anderson, S., R. Kellogg and J. Sallee, 2011. What Do Consumers Believe about Future
Gasoline Prices?," NBER Working Paper 16974, National Bureau of Economic
Ressearch, Cambridge, MA.
3. Axsen, J., D.C. Mountain and M. Jaccard, 2009. "Combining stated and revealed choice
research to simulate the neighbor effect: The case of hybrid-electric vehicles," Resource
and Energy Economics, vol. 31, pp. 221-238.
4. Beggs, S., and S. Cardell 1980. "Choice of smallest car by multi-vehicle households and
the demand for electric vehicles," Transportation Research-A, vol. 14A, pp. 389-405.
5. Berry, S., J. Levinsohn and A. Pakes, 1995. "Automobile Prices in Market Equilibrium,"
Econometrica, vol. 63, no. 4, pp. 841-890.
6. Brownstone, D., D.S. Bunch, T.F. Golob and W. Ren, 1996. "A Transactions Choice
Model for Forecasting Demand for Alternative-fuel Vehicles," Research in
Transportation Economics, vol. 4, pp. 87-129.
7. Brownstone, D. and K. Train, 1999. "Forecasting new product penetration with flexible
substitution patterns," Journal of Econometrics, vol. 89, pp. 109-129.
8. Brownstone, D., D.S. Bunch and K. Train, 2000. "Joint mixed logit models of stated and
revealed preferences for alternative-fuel vehicles," Transportation Research B, vol. 34,
pp. 315-338.
9. Carson, R.T., Czajkowski, M. 2013. "A New Baseline Model for Estimating Willingness
to Pay from Discrete Choice Models," International Choice Modelling Conference,
Sydney, Australia.
10. Daly, A., S. Hess and G. de Jong, 2012. "Calculating errors for measures derived from
choice modelling estimates," Transportation Research Part B, vol. 46, pp. 333-341.
11. Dimitropoulos, A., P. Rieteld and J.N. van Ommeren, 2013. "Consumer valuation of
changes in driving range: A meta-analysis," Transportation Research-A, vol. 55, pp. 27-
45.
12. Gatta, V., E. Marcucci and L. Scaccia, 2015. "On finite sample performance of
confidence interval methods for willingness to pay measures," Transportation Research
Part A, vol. 82, pp. 169-192
8-1
-------
13
14
15
16
17
18
19
20
21
22
23
24
25
Greene, D.L., A. Hossain, J. Hofmann, G. Helfand and R. Beach, 2018. "Consumer
Willingness to Pay for Vehicle Attributes: What Do We Know?" RTI International,
Research Triangle Park, NC, unpublished manuscript, March 22, 2018.
Greene, D.L., 2001. TAFV Alternative Fuels and Vehicles Choice Model
Documentation, ORNL/TM-2001/134, Oak Ridge National Laboratory, Oak Ridge, TN,
July.
Griliches, Zvi, 1971. "Hedonic Price Indexes of Automobiles: An Econometric Analysis
of Quality Change," in Zvi Griliches (ed.), Price Indexes and Quality Change, Cambridge
University Press, Cambridge, UK.
Haaf, G., J.J. Michalek, W.R. Morrow and Y. Liu, 2014. "Sensitivity of Vehicle Market
Share Predictions to Discrete Choice Model Specification," Journal of Mechanical
Design, vol. 136, Dec., pp. 121402-1 to 121402-9.
Hensher, D.A. and W.H. Greene, 2003. "The mixed logit model: The state of practice,"
Transportation, vol. 30, pp. 133-176.
Hidrue, M.K., G.R. Parsons, W. Kempton and M.P. Gardner, 2011. "Willingness to pay
for electric vehicles and their attributes," Resource and Energy Economics, vol. 33, pp.
686-705.
Kavalec, C., 1999. "Vehicle Choice in and Aging Population: Some Insights from a
Stated Preference Survey for California," The Energy Journal, vol. 20, no. 3, pp. 123-
138.
Klier, T. and J. Linn, 2012. "New-vehicle characteristics and the cost of the Corporate
Average Fuel Economy standard," RAND Journal of Economics, vol. 43, no. 1, pp. 186-
213.
Lancaster, K.J., 1966. "A new approach to consumer theory," Journal of Political
Economy, vol. 74, no. 2, pp. 132-157.
Lave, C. and K. Train, 1979. "A Disaggregate Model of Auto-Type Choice,"
Transportation Research-A, vol. 13, no. 1, pp. 1-9.
Liu, Y., J.M. Tremblay and C. Cirillo, 2014. "An integrated model for discrete and
continuous decisions with application to vehicle ownership, type and usage choices,"
Transportation Research-A, vol. 69, pp. 315-328.
Loomis, J.B., 2014. "Strategies for overcoming hypothetical bias in stated preference
surveys," Journal of Agricultural and Resource Economics, vol. 39, no. 1, pp. 34-46.
Massiani, J., 2013. SP surveys for electric and alternative fuel vehicles: are we doing the
right thing?, Working Paper No. 01/WP/2013, Department of Economics, Ca' Foscari
University of Venice.
8-2
-------
26. McFadden, D.L., 1974. "The Measurement of Urban Travel Demand." Journal of Public
Economics 3: 303-328.
27. McFadden, D.L. and K. Train, 2000. "Mixed MNL Models for Discrete Response,"
Journal of Applied Econometrics, vol. 15, pp. 447-470.
28. National Highway Traffic Safety Administration (NHTSA), 2006. Vehicle Survivability
and Travel Mileage Schedules, DOT HS 809 952, U.S. Department of Transportation,
Washington, DC, January.
29. National Institute of Standards and Technology (NIST), 2016. Engineering Statistics
Handbook, accessed at http://www.itl.nist.gov/div898/handbook/prc/sectionl/prcl6.htm,
on July 21, 2016.
30. National Research Council, 2002. At What Price? Conceptualizing and Measuring Cost-
of-Living and Price Indexes. Panel on Conceptual, Measurement, and Other Statistical
Issues in Developing Cost-of-Living Indexes, Charles L. Schultze and Christopher
Mackie, Editors. Committee on National Statistics, Division of Behavioral and Social
Sciences and Education. Washington, DC: National Academy Press.
31. Nicholas, M., S.L. Handy and D. Sperling, 2004. "Using Geographic Information
Systems to Evaluate Siting and Networks of Hydrogen Stations," Transportation
Research Record 1880, pp. 126-134, Transportation Research Board, National Research
Council, Washington, D.C.
32. Nickerson, R.S., 1993. "Confirmation Bias: A Ubiquitous Phenomenon in Many Guises,"
Review of General Psychology, vol. 2, no. 2, pp. 175-220.
33. Petrin, A., 2002. "Quantifying the Benefits of New Products: The Case of the Minivan,"
Journal of Political Economy, vol. 110, no. 4, pp. 705-729.
34. Potoglou, D. and P.S. Kanaroglou, 2008. "Disaggregate Demand Analysis for
Conventional and Alternative Fueled Automobiles: A Review," International Journal of
Sustainable Transportation, vol. 2, no. 4, pp. 234-259.
35. Quandt, R.E. and W.J. Baumol, 1966. "The Demand for Abstract Transport Modes:
Theory and Measurement," Journal of Regional Science, vol. 6, no. 2, 13-26.
36. Rosen, S. (1974). "Hedonic prices and implicit markets: product differentiation in pure
competition". Journal of Political Economy 82 (1): 34-55.
37. Sallee, James M., Sarah E. West, and Wei Fan. "Do Consumers Recognize the Value of
Fuel Economy? Evidence from Used Car Prices and Gasoline Price Fluctuations."
Journal of Public Economics 135 (2016): 61-73.
38. Seltman, Howard, n.d. "Approximations for Mean and Variance of a Ratio."
http://www.stat.cmu.edu/~hseltman/files/ratio.pdf, accessed 5/23/2016.
8-3
-------
39. State of Michigan, 2014. "Summary of Vehicle Acceleration and Top Speed Testing,"
accessed at
http://www.michigan.gov/documents/msp/VehicleAcceleration_TopSpeedTesting20105_
470745_7.pdf on May 27, 2016.
40. Tanaka, M., T. Ida, K. Murakami and 1. Friedman, 2014. "Consumers' willingness to pay
for alternative fuel vehicles: A comparative discrete choice analysis between the US and
Japan," Transportation Research-A, vol. 70, pp. 194-209
41. Tardiff, T.J., 1980. "Vehicle Choice Models: Review of Previous Studies and Directions
for Further Research," Transportation Research-A, vol. 14A, pp. 327-335.
42. Tompkins, M., D.S. Bunch, D. Santini, M. Bradley, A. Vyas, and D. Poyer. 1998.
"Determinants of alternative fuel vehicle choice in the Continental United States,"
Journal of Transportation Research Board 1641:130-138.
43. Train, K., 2009. Discrete Choice Methods with Simulations, Cambridge University Press,
Cambridge, New York.
44. Train, K., and M. Weeks, 2005. "Discrete Choice Models in Preference Space and
Willingness-to-Pay Space," In: Scarpa, R. and A. Alberini (eds), Applications of
Simulation Methods in Environmental and Resource Economics, The Economics of Non-
Market Goods and Resources, vol. 6: 1-16. Springer: New York, doi: 10.1007/1-4020-
3684-1 1.
45. Van Houtven, G., 2008. "Methods for the Meta-Analysis of Willingness-to-Pay Data,"
Pharmocoeconomics, vol. 26, no. 11, pp. 901-910.
46. Zhao, Y., K. Kockelman and A. Karlstrom, 2012. "Welfare calculations in discrete
choice settings: An exploratory analysis of error term correlation with finite populations,"
Transport Policy, vol. 19, pp. 76-84.
8-4
-------
APPENDIX A:
BIBLIOGRAPHY OF PAPERS INCLUDED IN OUR MAIN SAMPLE
Allcott, Hunt, and Nathan Wozny. "Gasoline Prices, Fuel Economy, and the Energy Paradox."
Review of Economics and Statistics 96, no. 5 (2014): 779-95. doi:10.1162/rest_a_00419.
Axsen, Jonn, Dean C. Mountain, and Mark Jaccard. "Combining Stated and Revealed Choice
Research to Simulate the Neighbor Effect: The Case of Hybrid-electric Vehicles."
Resource and Energy Economics 31, no. 3 (2009): 221-38.
doi:10.1016/j.reseneeco.2009.02.001.
Beresteanu, Arie, and Shanjun Li. "Gasoline Prices, Government Support, And The Demand For
Hybrid Vehicles In The United States*." International Economic Review 52, no. 1
(2011): 161-82. doi:10.1111/j.l468-2354.2010.00623.x.
Berry, Steven, James Levinsohn, and Ariel Pakes. "Automobile Prices in Market Equilibrium."
Econometrica 63, no. 4 (1995): 841. doi: 10.2307/2171802.
Brownstone, David, and Kenneth Train. "Forecasting New Product Penetration with Flexible
Substitution Patterns." Journal of Econometrics 89, no. 1-2 (1998): 109-29.
doi: 10.1016/s03 04-4076(98)00057-8.
Brownstone, David, David S. Bunch, and Kenneth Train. "Joint Mixed Logit Models of Stated
and Revealed Preferences for Alternative-fuel Vehicles." Transportation Research Part
B: Methodological 34, no. 5 (2000): 315-38. doi: 10.1016/sO 191-2615(99)00031-4.
Brownstone, David, David S. Bunch, Thomas F. Golob, and Weiping Ren. "A Transactions
Choice Model for Forecasting Demand for Alternative-fuel Vehicles." Research in
Transportation Economics A (1996): 87-129. doi:10.1016/s0739-8859(96)80007-2.
Busse, Meghan R., Christopher R. Knittel, and Florian Zettelmeyer. "Are Consumers Myopic?
Evidence from New and Used Car Purchases." American Economic Review 103, no. 1
(2013): 220-56. doi:10.1257/aer,103.1.220.
Dasgupta, Srabana, S. Siddarth, and Jorge Silva-Risso. "To Lease or to Buy? A Structural Model
of a Consumer's Vehicle and Contract Choice Decisions." Journal of Marketing
Research 44, no. 3 (2007): 490-502. doi:10.1509/jmkr.44.3.490.
Daziano, Ricardo A. "Conditional-logit Bayes Estimators for Consumer Valuation of Electric
Vehicle Driving Range." Resource and Energy Economics 35, no. 3 (2013): 429-50.
doi:10.1016/j.reseneeco.2013.05.001.
Dreyfus, Mark K., and W. Kip Viscusi. "Rates of Time Preference and Consumer Valuations of
Automobile Safety and Fuel Efficiency." The Journal of Law and Economics 38, no. 1
(1995): 79-105. doi: 10.1086/467326.
Espey, Molly, and Santosh Nair. "Automobile Fuel Economy: What Is It Worth?" Contemporary
Economic Policy 23, no. 3 (2005): 317-23. doi:10.1093/cep/byi024.
Fan, Qin, and Jonathan Rubin. "Two-Stage Hedonic Price Model for Light-Duty Vehicles."
Transportation Research Record: Journal of the Transportation Research Board 2157
(2010): 119-28. doi: 10.3141/2157-15.
A-l
-------
Feng, Ye, Don Fullerton, and Li Gan. "Vehicle Choices, Miles Driven, and Pollution Policies."
Journal ofRegulatory Economics 44, no. 1 (2013): 4-29. doi:10.1007/slll49-013-9221-
z.
Fifer, Daniel P.C. and N.P. Bunn. "Assessing Consumer Valuation of Fuel Economy in Auto
Markets," Honors Thesis, Department of Economics, Duke University, Durham, North
Carolina (2009).
Frischknecht, Bart D., Katie Whitefoot, and Panos Y. Papalambros. "On the Suitability of
Econometric Demand Models in Design for Market Systems." Journal of Mechanical
Design J.Mech. Des. 132, no. 12 (2010): 121007. doi:l0.1115/1.4002941.
Gallagher, Kelly Sims, and Erich Muehlegger. "Giving Green to Get Green? Incentives and
Consumer Adoption of Hybrid Vehicle Technology." Journal of Environmental
Economics and Management 61, no. 1 (2011): 1-15. doi:10.1016/j.jeem.2010.05.004.
Goldberg, Pinelopi Koujianou. "Product Differentiation and Oligopoly in International Markets:
The Case of the U.S. Automobile Industry." Econometrica 63, no. 4 (1995): 891.
doi: 10.2307/2171803.
Gramlich, J. Gas Prices and Endogenous Produce Selection in the U.S. Automobile Industry,
manuscript, Department of Economics, Yale University, New Haven, Connecticut,
November, 20, 2008.
Greene, D. L. "TAFV Alternative Fuels and Vehicles Choice Model Documentation."
Department of Energy, 2001. doi:10.2172/814556.
Greene, D.L., K.G. Duleep, and W. McManus. "Future Potential of Hybrid and Diesel
Powertrains in the U.S. Light-duty Vehicle Market." Oak Ridge National Lab, 2004.
doi: 10.2172/885725.
Haaf, C. Grace, Jeremy J. Michalek, W. Ross Morrow, and Yimin Liu. "Sensitivity of Vehicle
Market Share Predictions to Discrete Choice Model Specification." Journal of
Mechanical Design J. Mech. Des. 136, no. 12 (2014): 121402. doi:10.1115/1.4028282.
Helveston, John Paul, Yimin Liu, Elea Mcdonnell Feit, Erica Fuchs, Erica Klampfl, and Jeremy
J. Michalek. "Will Subsidies Drive Electric Vehicle Adoption? Measuring Consumer
Preferences in the U.S. and China." Transportation Research Part A: Policy and Practice
73 (2015): 96-112. doi:10.1016/j.tra.2015.01.002.
Hess, Stephane, Kenneth E. Train, and John W. Polak. "On the Use of a Modified Latin
Hypercube Sampling (MLHS) Method in the Estimation of a Mixed Logit Model for
Vehicle Choice." Transportation Research Part B: Methodological 40, no. 2 (2006): 147-
63. doi: 10.1016/j.trb.2004.10.005.
Hess, Stephane, Mark Fowler, Thomas Adler, and Aniss Bahreinian. "A Joint Model for Vehicle
Type and Fuel Type Choice: Evidence from a Cross-nested Logit Study." Transportation
39, no. 3 (2011): 593-625. doi: 10.1007/sl 1116-011-9366-5.
Hidrue, Michael K., George R. Parsons, Willett Kempton, and Meryl P. Gardner. "Willingness to
Pay for Electric Vehicles and Their Attributes." Resource and Energy Economics 33, no.
3 (2011): 686-705. doi: 10.1016/j.reseneeco.2011.02.002.
A-2
-------
Kavalec, Chris. "Vehicle Choice in an Aging Population: Some Insights from a Stated
Preference Survey for California." The Energy Journal 20, no. 3 (1999).
doi: 10.5 547/issnO 195-6574-ej -vol20-no3 -5.
Klier, Thomas, and Joshua Linn. "New-vehicle Characteristics and the Cost of the Corporate
Average Fuel Economy Standard." The RAND Journal of Economics 43, no. 1 (2012):
186-213. doi: 10.1111/j. 1756-2171.2012.00162.x.
Lave, Charles A., and Kenneth Train. "A Disaggregate Model of Auto-type Choice."
Transportation Research Part A: General 13, no. 1 (1979): 1-9. doi: 10.1016/0191-
2607(79)90081-5.
Liu, Yangwen, Jean-Michel Tremblay, and Cinzia Cirillo. "An Integrated Model for Discrete and
Continuous Decisions with Application to Vehicle Ownership, Type and Usage Choices."
Transportation Research Part A: Policy and Practice 69 (2014): 315-28.
doi:10.1016/j.tra.2014.09.001.
Liu, Yizao. "Household Demand and Willingness to Pay for Hybrid Vehicles." Energy
Economics 44(2014): 191-97. doi:10.1016/j.eneco.2014.03.027.
McFadden, Daniel, and Kenneth Train. "Mixed MNL Models for Discrete Response." Journal of
Applied Econometrics 15, no. 5 (2000): 447-70. doi:10.1002/1099-
1255(200009/10)15:53.3.co;2-t.
McCarthy, Patrick S. "Market Price and Income Elasticities of New Vehicle Demands." The
Review of Economics and Statistics 78, no. 3 (1996): 543. doi: 10.2307/2109802.
McCarthy, Patrick S., and Richard S. Tay. "New Vehicle Consumption and Fuel Efficiency: A
Nested Logit Approach." Transportation Research PartE: Logistics and Transportation
Review 34, no. 1 (1998): 39-51. doi:10.1016/sl366-5545(97)00042-2.
McManus, Walter. "The Link Between Gasoline Prices and Vehicle Sales." Business Economics
42, no. 1 (2007): 53-60. doi: 10.2145/20070106.
Musti, Sashank, and Kara M. Kockelman. "Evolution of the Household Vehicle Fleet:
Anticipating Fleet Composition, PHEV Adoption and GHG Emissions in Austin, Texas."
Transportation Research Part A: Policy and Practice 45, no. 8 (2011): 707-20.
doi: 10.1016/j.tra.2011.04.011.
Nixon, Hilary, and Jean-Daniel Saphores. "Understanding household preferences for
alternatives-fuel vehicle technologies." Report 10-11. Mineta Transportation Institute,
San Jose, CA, 2011.
Parsons, George R., Michael K. Hidrue, Willett Kempton, and Meryl P. Gardner. "Willingness to
Pay for Vehicle-to-grid (V2G) Electric Vehicles and Their Contract Terms." Energy
Economics 42 (2014): 313-24. doi:10.1016/j.eneco.2013.12.018.
Petrin, Amil. "Quantifying the Benefits of New Products: The Case of the Minivan." Journal of
Political Economy 110, no. 4 (2002): 705-29. doi: 10.1086/340779.
Sallee, James M., Sarah E. West, and Wei Fan. "Do Consumers Recognize the Value of Fuel
Economy? Evidence from Used Car Prices and Gasoline Price Fluctuations." Journal of
Public Economics 135 (2016): 61-73. doi:10.1016/j.jpubeco.2016.01.003.
A-3
-------
Segal, Robin. "Forecasting the Market for Electric Vehicles in California Using Conjoint
Analysis." The Energy Journal 16, no. 3 (1995). doi:10.5547/issn0195-6574-ej-voll6-
no3-4.
Sexton, Steven E., and Alison L. Sexton. "Conspicuous Conservation: The Prius Halo and
Willingness to Pay for Environmental Bona Fides." Journal of Environmental Economics
and Management 67, no. 3 (2014): 303-17. doi:10.1016/j.jeem.2013.11.004.
Shiau, Ching-Shin Norman, Jeremy J. Michalek, and Chris T. Hendrickson. "A Structural
Analysis of Vehicle Design Responses to Corporate Average Fuel Economy Policy."
Transportation Research Part A: Policy and Practice 43, no. 9-10 (2009): 814-28.
doi:10.1016/j.tra.2009.08.002.
Skerlos, S., Raichur, V. "PRISM 2.0: Mixed Logit Consumer Vehicle Choice Modeling Using
Revealed Preference Data." Electric Power Research Institute, 2013.
Tanaka, Makoto, Takanori Ida, Kayo Murakami, and Lee Friedman. "Consumers' Willingness to
Pay for Alternative Fuel Vehicles: A Comparative Discrete Choice Analysis between the
US and Japan." Transportation Research Part A: Policy and Practice 70 (2014): 194-
209. doi: 10.1016/j.tra.2014.10.019.
Tompkins, Melanie, David Bunch, Danilo Santini, Mark Bradley, Anant Vyas, and David Poyer.
"Determinants of Alternative Fuel Vehicle Choice in the Continental United States."
Transportation Research Record: Journal of the Transportation Research Board 1641
(1998): 130-38. doi: 10.3141/1641-16.
Train, Kenneth, and Garrett Sonnier. (1995). "Mixed Logit with Bounded Distributions of
Correlated Partworths." Applications of Simulation Methods in Environmental and
Resource Economics The Economics of Non-Market Goods and Resources: 117-34.
doi: 10.1007/l-4020-3684-l_7.
Train, Kenneth, and Melvyn Weeks. (2005). "Discrete Choice Models in Preference Space and
Willingness-to-Pay Space." In: Scarpa, R. and A. Alberini (eds), Applications of
Simulation Methods in Environmental and Resource Economics. The Economics of Non-
Market Goods and Resources, vol. 6: 1-16. Springer: New York, doi: 10.1007/1-4020-
3684-1 1.
Train, Kenneth E., and Clifford Winston. "Vehicle Choice Behavior And The Declining Market
Share Of U.S. Automakers." International Economic Review 48, no. 4 (2007): 1469-496.
doi:10.1111/j. 1468-2354.2007.00471.x.
Walls, Margaret A. "Valuing the Characteristics of Natural Gas Vehicles: An Implicit Markets
Approach." The Review of Economics and Statistics 78, no. 2 (1996): 266.
doi: 10.2307/2109928.
Whitefoot, K., M. Fowlie, and S. Skerlos. "Product Design Response to Industrial Policy:
Evaluating Fuel Economy Standards Using an Engineering Model of Endogenous
Product Design." Working Paper, University of Michigan (2011).
Zhang, Ting, Sonja Gensler, and Rosanna Garcia. "A Study of the Diffusion of Alternative Fuel
Vehicles: An Agent-Based Modeling Approach." Journal of Product Innovation
Management 28, no. 2 (2011): 152-68. doi: 10.1111/j. 1540-5885.2011.00789.x.
A-4
-------
APPENDIX B:
WILLINGNESS TO PAY RESULTS BY ATTRIBUTE GROUPING
Comfort
hirst Author
Second Author
Pub Ycnr
Jourmil
Diilii Tvpe
Dolhir Yc;ir
SUit Model
Attribute
1 nteraction
Coeff.
SE
mu
si^nui
Slundiird Units
Low WIT
Central WIT
IliKh WIT
K:in^e Desc.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
air conditioning
0.58
0.63
0.58
1.22
0/1
-1031.32
976.18
2983.68
random coef.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
air conditioning
1.52
0.89
1.52
1.82
0/1
-239.51
1397.70
3034.92
random coef.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
AT
0.03
0.01
0/1
199.06
345.03
491.01
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
comfort rating
1024.10
149.07
scale f1,51
1171.35
1370.90
1570.45
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
AT
0.06
0.01
0/1
1707.61
2158.67
2609.74
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
AT
0.04
0.02
0/1
943.49
1522.10
2100.71
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
air conditioning
5.78
0.26
0/1
2858.51
3961.64
5370.75
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
air conditioning
3.47
0.53
0/1
5921.01
11004.47
18930.18
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
air conditioning
8.96
0.43
0/1
4714.37
6479.66
10603.26
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
AT
0.88
0.28
0/1
332.29
658.41
1303.16
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
AT
3.52
0.23
0/1
1704.37
2401.11
3329.70
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
AT
1.69
0.30
0/1
2816.23
5321.43
9416.98
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
powersteering
5.53
0.36
0/1
2854.54
3990.39
6651.30
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
powersteering
0.62
0.20
0/1
215.69
396.04
724.45
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
powersteering
-1.59
0.62
0/1
-10416.58
-4520.05
-1960.07
standard error
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
air conditioning
0.09
0/1
723.78
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
AT
-0.10
0/1
-2987.03
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
AT
-0.07
0/1
-2311.02
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.03
0.01
$/inch
0.15
0.18
0.21
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.03
0.01
$/inch
0.44
0.53
0.63
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.03
0.01
$/inch
4.34
5.27
6.19
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.05
0.00
$/inch
0.32
0.34
0.36
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.05
0.00
$/inch
0.70
0.73
0.77
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.05
0.00
$/inch
1.11
1.17
1.23
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.05
0.00
$/inch
0.29
0.31
0.32
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.05
0.00
$/inch
0.52
0.55
0.57
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.05
0.00
$/inch
1.70
1.80
1.90
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.07
0.01
$/inch
0.36
0.41
0.45
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.07
0.01
$/inch
0.57
0.64
0.70
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
shoulder room
income
0.07
0.01
$/inch
0.99
1.10
1.22
standard error
Petrin
2002
J. Political Economy
market data
1983
BLP
air conditioning
3.88
0.01
0/1
7765.51
7785.18
7804.84
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
air conditioning
3.88
0.01
0/1
8177.08
8197.79
8218.49
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
air conditioning
-1.97
0.95
0/1
-5574.90
-3785.78
-1996.67
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
air conditioning
-1.97
0.95
0/1
-15239.89
-10349.05
-5458.20
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
air conditioning
-1.97
0.95
0/1
-22648.48
-15380.04
-8111.60
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
air conditioning
3.88
0.01
0/1
19768.42
19818.47
19868.53
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
AWD
-5.24
1.61
0/1
28591.26
40909.35
53227.43
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
AWD
-12.32
4.42
0/1
40803.59
62928.77
85053.95
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
AWD
-12.32
4.42
0/1
16878.14
26030.08
35182.02
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
AWD
-12.32
4.42
0/1
16028.64
24719.95
33411.26
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
AWD
-5.24
1.61
0/1
13101.89
10069.80
7037.72
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
AWD
-5.24
1.61
0/1
19238.72
27527.41
35816.11
random coef.
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
AT
0.65
0.28
0/1
5355.25
9260.61
13165.96
standard error
Walls
1996
RE Stat
market data
1990
Hedonic
air conditioning
0.72
23.40
0/1
4019.92
14464.17
24908.41
standard error
B-l
-------
Fuel Availability
hirst Author
Second Author
Pub Vear
Journal
Data Tvpe
Dollar Veai*
Stat Model
Attribute
1 nteraction
Coeff.
SE
mu
si^ma
Standard IJnits
Low WTP
Central WIT
1 llfili WIT
Kan^e Desc.
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
refuel time reduction
0.00
0 00
$/hr
-207.15
920.66
2048 48
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
refuel time reduction
0.00
0 00
$/hr
-95.28
1071.92
2239 12
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
station availability 1%
fuel type
0.57
0.48
$/%
32.84
179.12
325.40
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
station availability 1%
fuel type
0.58
0.45
$/%
44.94
182.56
320.18
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
station availability 1%
fuel type
0.74
0.49
$/%
80.20
231.33
382.47
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
station availability 1%
fuel type
0.63
0.53
$/%
36.14
197.11
358.08
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
station availability 1%
fuel type
1.00
0.44
$/%
179.94
313.53
447.11
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
station availability 1%
fuel type
0.30
0.23
$/%
23.02
93.53
164.03
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
station availability 1%
0.91
0.30
$/%
73.08
108.43
143.78
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
station availability 1%
0.53
0.17
$/%
69.75
102.97
136.18
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
station availability 1%
0.67
0.83
$/%
-33.09
140.62
314.33
random coef.
Greene
2001
Grey
Lit. Review
1990
NMNL
home refuel
-168.40
$/(min or hr?)
37.26
Greene
2001
Grey
Lit. Review
1990
Other
home refuel
0.08
0/1
171.88
Greene
2001
Grey
Lit. Review
1990
Other
station availability 1%
$/%
2242.30
Greene
2001
Grey
Lit. Review
1990
Other
station availability 1%
$/%
365.71
Greene
2001
Grey
Lit. Review
1990
Other
station availability 1%
$/%
48.76
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
station availability 1%
midsize
$/%
578.75
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
station availability 1%
smallSUV
$/%
789.60
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
station refuel time 300h
smallSUV
$/hr
698.70
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
station refuel time 300h
midsize
$/hr
419.47
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
charge time reduction EV
0.03
26.24
$/hr
-26126.71
30.52
26187.75
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
charge time reduction EV
3.34
1.48
$/hr
1927.35
3400.85
4874.36
standard error
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
charge time reduction PHEV
3.33
8.88
$/hr
-5466.35
3388.65
12243.64
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
charge time reduction PHEV
3.94
1.33
$/hr
2686.30
4012.25
5338.21
standard error
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
Plug-in at work/other
0.12
0.12
0/1
-107.52
3476.54
7060.61
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
station availability 1%
0.31
0.13
$/%
5230.18
9133.30
13036.41
varied income
Hidrue
Parsons
2011
REE
SP survey
2009
Other
charge time reduction
2.20
0.52
$/hr
5495.55
7168.10
8840.66
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
charge time reduction
0.80
0.05
$/hr
11161.32
11947.10
12732.89
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
charge time reduction
0.55
0.05
$/hr
1059.52
1173.38
1287.23
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
charge time reduction
2.00
0.50
$/hr
702.85
930.92
1159.00
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
charge time reduction
0.07
0.05
$/hr
85.77
348.46
611.14
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
charge time reduction
1.60
0.55
$/hr
1150.49
1737.72
2324.95
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
home refuel
0.67
0.26
0/1
6133.75
9991.46
13849.17
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
refuel time reduction
0.00
0.00
$/hr
-419.85
268.43
956.70
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
refuel time reduction
-0.01
0.00
$/hr
2486.68
5189.59
7892.51
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
station availability 1%
0.97
0.26
$/%
106.17
144.35
182.54
standard error
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
station availability 1%
0.70
1.40
$/%
-111.58
117.44
346.43
random coef.
Nixon
Saphores
2011
Grey
market data
2010
MNL
charge time reduction
-0.42
$/hr
2063.63
8692.80
36617.50
random coef.
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
GV-oriented
0.07
0.19
$/hr
-859.09
-227.41
404.28
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
$/hr
199.27
765.58
1331.89
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
EV-oriented
-0.23
0.07
$/hr
1075.59
1587.78
2099.96
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
GV-oriented
-0.43
0.20
$/hr
209.06
399.12
589.18
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
$/hr
530.40
698.70
866.99
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
EV-oriented
-0.48
0.08
$/hr
796.47
946.75
1097.03
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
EV-oriented
-0.69
0.08
$/hr
704.69
793.89
883.09
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
$/hr
670.43
782.55
894.66
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
charge time reduction
GV-oriented
-1.42
0.26
$/hr
629.06
768.85
908.64
standard error
Segal
1995
Energy Journal
SP survey
1994
MNL
charge time reduction
-0.18
$/hr
1344.86
Segal
1995
Energy Journal
SP survey
1994
MNL
home or station refuel
0.12
0/1
859.22
Segal
1995
Energy Journal
SP survey
1994
MNL
home refuel
-0.54
0/1
-4019.63
Segal
1995
Energy Journal
SP survey
1994
MNL
refuel 12-6am
-0.06
0/1
-478.17
Segal
1995
Energy Journal
SP survey
1994
MNL
refuel any but 2-9pm
-0.71
0/1
-5282.31
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
station availability 1%
0.01
0.00
$/%
49.41
51.10
52.78
standard error
B-2
-------
Fuel Costs
First Author
Second Author
Pub Year
Journal
D ll 1 1>|H
Dollar Year
Stat Model
Attribute
Inleraction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Axsen
Mountain
2009
REE
RP survey
^006
MNL
fuel cost per mile reduction
0.00
0.00
$/cpm
555.52
8 .27
611.02
standard error
Axsen
Mountain
2009
REE
RP&SP
^006
MNL
fuel cost per mile reduction
0.00
0.00
$/cpm
1278.14
1 49.26
1820.38
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
fuel cost per mile reduction
0.00
0.00
$/cpm
675.16
718.64
762.11
standard error
Axsen
Mountain
2009
REE
SP survey
2006
MNL
fuel cost per mile reduction
-0.03
0.00
$/cpm
577.92
606.79
635.65
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
fuel cost per mile reduction
0.00
$/cpm
602.49
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
fuel cost per mile reduction
-0.04
0.00
$/cpm
2151.33
2258.79
2366.25
standard error
Allcott
Wozny
2014
RE Stat
market data
2005
Other
fuel cost per mile reduction
0.76
0.05
$/cpm
-1224.65
-1094.77
-964.90
standard error
Allcott
Wozny
2014
RE Stat
market data
2005
Other
fuel cost per mile reduction
0.55
0.03
$/cpm
-882.62
-792.27
-701.92
standard error
Allcott
Wozny
2014
RE Stat
market data
2005
Other
fuel cost per mile reduction
0.51
0.03
$/cpm
-816.53
-734.65
-652.77
standard error
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
fuel cost per mile reduction
-2.93
1.24
-2.93
2.11
$/cpm
76.03
256.09
436.15
random coef.
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
fuel cost per mile reduction
-8.62
1.18
5.77
1.24
$/cpm
309.40
899.19
1488.97
random coef.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
fuel cost per mile reduction
-0.49
0.16
-0.49
0.67
$/cpm
-11665.16
-4985.11
1694.94
random coef.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
fuel cost per mile reduction
-0.12
0.32
-0.12
1.05
$/cpm
-6382.44
-676.51
5029.43
random coef.
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
30K< Inc < 75K nochild
-0.08
0.02
$/cpm
2293.20
3221.85
4150.50
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
30K< Inc luxury child
-0.08
0.05
$/cpm
-10032.96
-6068.32
-2103.69
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
<30K child
-0.01
0.05
$/cpm
-909.79
233.28
1376.35
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
30K< Inc luxury no child
-0.08
0.04
$/cpm
-9940.07
-6557.68
-3175.30
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
<30Kno child
-0.08
0.02
$/cpm
1327.59
1888.38
2449.17
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
>75Kno child
-0.13
0.05
$/cpm
-10456.93
-7425.04
-4393.15
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
<30Kno child
-0.03
0.04
$/cpm
-424.92
1888.55
4202.03
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
>30Kno lux child
-0.08
0.02
$/cpm
5440.98
7739.32
10037.66
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
>30K no lux no child
-0.08
0.02
$/cpm
2413.86
3352.58
4291.31
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
fuel cost per mile reduction
<30K child
-0.01
0.04
$/cpm
-1578.77
404.81
2388.39
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
fuel cost per mile reduction
-0.13
0.03
0.26
0.06
$/cpm
1437.73
2564.42
3691.11
random coef.
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
fuel cost per mile reduction
-0.24
0.05
0.58
0.15
$/cpm
-4068.99
2799.65
9668.29
random coef.
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
fuel cost per mile reduction
-0.19
0.07
$/cpm
2098.43
3239.98
4381.53
standard error
Brownstone
Train
1999
J. Econometrics
SP survey
1993
MXL
fuel cost per mile reduction
-1.32
0.83
$/cpm
1012.73
2749.85
4486.98
random coef.
Busse
Knittel
2013
AER
market data
2008
Other
fuel cost per mile reduction
-1170.00
$/cpm
266.61
Busse
Knittel
2013
AER
market data
2008
Other
fuel cost per mile reduction
-1973.00
$/cpm
1027.08
Busse
Knittel
2013
AER
market data
2008
Other
fuel cost per mile reduction
-5801.00
$/cpm
969.34
Busse
Knittel
2013
AER
market data
2008
Other
fuel cost per mile reduction
-4571.00
$/cpm
1381.45
Dasgupta
Siddarth
2007
J. Marketing Research
market data
2000
MXL
fuel cost per mile reduction
-32721.40
9243.33
14301.50
4783.11
$/cpm
154.59
274.61
394.63
random coef.
Daziano
2013
REE
SP survey
1999
MXL
fuel cost per mile reduction
-0.05
$/cpm
155.44
5307.06
10236.63
random coef.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
fuel cost per mile reduction
0.69
0.03
$/cpm
-1096766.00
-1052469.50
-1008173.06
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
fuel cost per mile reduction
-257560.00
32074.72
$/cpm
1207.38
1379.12
1550.87
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
fuel cost per mile reduction
SUV
-0.87
0.26
$/cpm
-945.64
800.32
2546.28
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
fuel cost per mile reduction
van
1.10
0.25
$/cpm
2997.47
3848.93
4700.38
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
fuel cost per mile reduction
pickup
-1.20
0.26
$/cpm
-2079.69
-325.22
1429.25
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
fuel cost per mile reduction
0.17
0.06
$/cpm
291.11
439.30
587.50
standard error
Fifer
Burin
2009
Grey
market data
2002
Hedonic
fuel cost per mile reduction
$/cpm
747.12
1036.11
1325.09
standard error
Fifer
Burin
2009
Grey
market data
2002
Hedonic
fuel cost per mile reduction
van
-468153.00
103508.60
$/cpm
1921.84
2467.38
3012.92
standard error
Fifer
Bunn
2009
Grey
market data
2002
Hedonic
fuel cost per mile reduction
pickup
-549569.00
95167.13
$/cpm
2394.91
2896.48
3398.06
standard error
Fifer
Bunn
2009
Grey
market data
2002
Hedonic
fuel cost per mile reduction
cars+SUVs
-87349.10
40142.99
$/cpm
248.80
460.37
671.94
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
fuel cost per mile reduction
-0.88
0.03
1.03
0.03
$/cpm
-0.07
0.43
0.94
random coef.
Gallagher
Muehlegger
2011
JEEM
market data
2011
Other
fuel cost per mile reduction
0.00
0.00
$/cpm
-284.36
-469.79
-655.21
standard error
Gallagher
Muehlegger
2011
JEEM
market data
2011
Other
fuel cost per mile reduction
0.00
0.00
$/cpm
-1833.40
-3224.11
-4614.83
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
fuel cost per mile reduction
-1.38
0.74
$/cpm
23.86
10.04
4.27
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
fuel cost per mile reduction
0.23
0.93
$/cpm
32.36
-7.80
-54.30
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
fuel cost per mile reduction
-7.14
0.74
$/cpm
70.12
48.37
35.01
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
-0.075
0.17
$/cpm
-757.59
-231.91
293.76
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
small
-4.70
0.66
$/cpm
1399.31
1627.91
1856.52
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
CUV
-9.07
2.95
$/cpm
2119.75
3141.53
4163.30
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
van
-3.65
0.70
$/cpm
1021.78
1264.23
1506.69
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
large
-4.68
0.65
$/cpm
1395.85
1620.99
1846.12
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
SUV
-8.10
1.13
$/cpm
2414.16
2805.55
3196.95
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
specialty
-4.02
0.44
$/cpm
1239.99
1392.39
1544.79
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
midsize
-5.55
0.55
$/cpm
1731.82
1922.32
2112.83
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
truck
-4.17
0.70
$/cpm
1201.89
1444.34
1686.80
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
Luxury
-2.13
0.43
$/cpm
588.82
737.76
886.69
standard error
Gramlich
2008
Grey
market data
2007
NMNL
fuel cost per mile reduction
$/cpm
1860.93
1568.22
1275.50
standard error
Greene
2001
Grey
Lit. Review
1990
Other
fuel cost per mile reduction
-0.70
$/cpm
1552.84
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
fuel cost per mile reduction
smallSUV
$/cpm
635.30
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
fuel cost per mile reduction
midsize
$/cpm
560.22
B-3
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interact ion
Coefl.
SE
mu
sigma
Standard Units
LowWTP
Central WTP
High WTP
Range Desc.
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
fuel cost per mile reduction
50.67
$/cpm
-6356.81
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
fuel cost per mile reduction
87.68
$/cpm
-7457.56
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
fuel cost per mile reduction
-14.97
$/cpm
1788.63
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
fuel cost per mile reduction
-19.41
$/cpm
2563.25
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
fuel cost per mile reduction
-41.22
$/cpm
4701.15
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MXL
fuel cost per mile reduction
90.83
0.01
90.83
0.01
$/cpm
-7471.38
-7472.19
-7473.00
random coef.
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
fuel cost per mile reduction
0.03
$/cpm
-2.04
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
fuel cost per mile reduction
0.02
$/cpm
1188.93
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
fuel cost per mile reduction
-1.63
0.08
$/cpm
1578.37
1654.14
1729.91
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
fuel cost per mile reduction
-1.60
0.11
$/cpm
1519.98
1625.65
1731.33
standard error
Hess
Fowler
2012
T ransportmetrica
RP&SP
2009
NMNL
fuel cost per mile reduction
0.02
0.00
$/cpm
1054.11
1219.60
1385.08
varied income
Hess
Fowler
2012
T ransportmetrica
RP&SP
2009
NMNL
fuel cost per mile reduction
-0.13
0.05
$/cpm
5353.82
3918.48
2483.14
varied income
Hess
Fowler
2012
T ransportmetrica
RP&SP
2009
NMNL
fuel cost per mile reduction
-0.05
0.02
$/cpm
1861.38
1417.13
972.89
varied income
Hess
Train
2006
TR-B
SP survey
1999
MXL
fuel cost per mile reduction
0.04
0.04
$/cpm
2838.44
5045.82
8969.81
random coef.
Hidrue
Parsons
2011
REE
SP survey
2009
Other
fuel cost per mile reduction
-0.17
0.23
$/cpm
273.03
115.63
-41.76
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
fuel cost per mile reduction
-0.35
0.04
$/cpm
1207.40
1097.64
987.88
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
fuel cost per mile reduction
age
$/cpm
2057.94
2706.64
2147.42
varied interaction
Kavalec
1999
Energy Journal
SP survey
1993
MXL
home refueling cost reduction
-0.02
0.01
$/cpm
25.56
238.60
451.64
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-0.98
6.68
$/cpm
-1435.55
17.98
1471.51
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-12.96
4.32
$/cpm
-226.51
76.71
379.94
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-3.29
13.70
$/cpm
-1176.11
24.31
1224.73
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
Other
fuel cost per mile reduction
-14.11
2.59
$/cpm
-114.96
97.71
310.37
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-3.95
6.26
$/cpm
-781.61
43.93
869.47
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-11.05
8.40
$/cpm
-761.73
95.05
951.83
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-12.52
2.63
$/cpm
-116.11
77.88
271.87
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-11.05
20.64
$/cpm
-2010.18
95.05
2200.27
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
0.43
7.36
$/cpm
-910.97
-4.47
902.04
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-12.44
2.44
$/cpm
-110.06
83.01
276.07
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
fuel cost per mile reduction
-13.24
9.94
$/cpm
-643.55
81.44
806.43
standard error
Lave
Train
1979
TR-A
market data
1976
MNL
fuel cost per mile reduction
-0.35
-0.22
$/cpm
2498.16
3729.28
5676.85
varied income
Liu
2014
Energy Economics
RP survey
2009
MXL
fuel cost per mile reduction
-0.08
0.01
$/cpm
15912.01
18254.69
20597.38
random coef.
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
fuel cost per mile reduction
income
0.00
0.01
S/MPG
-0.06
0.03
0.11
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
fuel cost per mile reduction
income
0.00
0.01
S/MPG
-0.18
0.08
0.33
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
fuel cost per mile reduction
income
0.00
0.01
S/MPG
-1.74
0.77
3.28
standard error
McCarthy
1996
RE Stat
RP survey
1989
MNL
fuel cost per mile reduction
-0.52
-0.45
$/cpm
2925.20
19415.06
35904.92
standard error
McCarthy
Tay
1998
TR-E
RP survey
1989
NMNL
fuel cost per mile reduction
-0.52
0.12
$/cpm
111.31
144.57
177.82
standard error
McFadden
Train
2000
J. Applied Econometrics
SP survey
1993
MXL
fuel cost per mile reduction
-0.18
0.45
$/cpm
-4274.85
3016.68
10308.20
random coef.
McManus
2007
Business Economics
market data
2002
Hedonic
fuel cost per mile reduction
-768.00
4.82
$/cpm
1005.58
1011.93
1018.28
standard error
Nixon
Saphores
2011
Grey
SP survey
2010
MXL
fuel cost per mile reduction
0.18
$/cpm
-67.17
-423.28
-2667.27
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
fuel cost per mile reduction
-15.79
2.58
$/cpm
-27169.79
-23419.68
-19669.56
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
fuel cost per mile reduction
-15.79
2.58
$/cpm
-28609.77
-24660.90
-20712.03
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
fuel cost per mile reduction
-15.79
2.58
$/cpm
-69165.26
-59618.71
-50072.17
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
fuel cost per mile reduction
-0.54
0.04
$/cpm
-3342.58
-3116.36
-2890.13
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
fuel cost per mile reduction
-0.54
0.04
$/cpm
-822.77
-767.09
-711.40
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
fuel cost per mile reduction
-0.54
0.04
$/cpm
-2249.18
-2096.96
-1944.73
random coef.
Sallee
West
2015
NBER
market data
2008
Other
fuel cost per mile reduction
0.69
0.01
$/cpm
-403.81
-398.16
-392.50
standard error
Sallee
West
2015
NBER
market data
2008
Other
fuel cost per mile reduction
0.98
0.04
$/cpm
-588.12
-565.50
-542.88
standard error
Segal
1995
Energy Journal
SP survey
1994
MNL
fuel cost per mile reduction
-0.94
$/cpm
1475.91
Shiau
Michalek
2009
TR-A
market data
2007
MXL
fuel cost per mile reduction
-0.18
0.15
$/cpm
488.07
2270.95
4053.84
random coef.
Skerlos
Raichur
2013
Grey
market data
2008
Other
fuel cost per mile reduction
-17.31
7.83
1.17
1.32
$/cpm
959.25
1028.79
1098.33
random coef.
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
charger cost $100 reduction
-0.06
0.00
0/1
212.09
218.80
225.51
standard error
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
fuel cost % reduction
0.01
0.00
$/%
47.76
51.10
54.43
standard error
T ompkins
Bunch
1998
UC ITS
SP survey
1995
MXL
fuel cost per mile reduction
-0.08
0.01
$/cpm
2374.51
2770.26
3166.01
standard error
T ompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
home refueling cost reduction
EV
-0.13
0.02
$/cpm
3761.08
4554.55
5348.03
standard error
Train
Weeks
2005
Book
SP survey
2000
MXL
fuel cost per mile reduction
-0.49
0.05
$/cpm
57635.18
64498.98
71362.77
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
fuel cost per mile reduction
-0.04
0.05
$/cpm
-5505.50
9101.08
23707.66
random coef.
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
fuel cost per mile reduction
-32.00
-102.00
$/cpm
-3859.27
1817.20
7493.67
random coef.
Whitefoot
Fowlie
2011
Grey
market data
2006
BLP
fuel cost per mile reduction
-0.37
0.17
$/cpm
545.26
376.69
208.12
standard error
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
fuel cost per mile reduction
-92.98
-92.98
28.08
$/cpm
2003.74
2846.30
3688.86
random coef.
B-4
-------
Fuel Type
hirst Author
Second Author
Pub Vear
Journal
Data Tvpe
Dollar Veai*
Stat Model
Attribute
Interaction
Coeff.
SE
mu
si^nia
Standard Units
Low WTP
Central \\ Tl>
Nigh WIT
kaiij^c Dcsc.
Axsen
Mountain
2009
REE
RP & SP
2006
MNL
Hybrid
-2.79
0.23
0/1
-66335.73
-55816.47
-45297 21
standard error
Axsen
Mountain
2009
REE
SP survey
2006
MNL
Hybrid
0 22
0.06
0/1
3696.39
4155.70
4615.01
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
Hybrid
-2.83
0.23
0/1
-51783.79
-43437.06
-35090.33
standard error
Axsen
Mountain
2009
REE
RP & SP
2006
MNL
Hybrid
-6.43
1.15
0/1
-211963.47
.180394.44
-148825.41
standard error
Axsen
Mountain
2009
REE
RP & SP
2006
MNL
Hybrid
-2.37
-0.23
0/1
-147825.23
-125808.72
-103792.19
standard error
Axsen
Mountain
2009
REE
SP survey
2006
MNL
Hybrid
-2.57
0.28
0/1
-16598.21
-14126.14
-11654.06
standard error
Axsen
Mountain
2009
REE
SP survey
2006
MNL
Hybrid
0.30
0.05
0/1
1959.12
2374.69
2790.26
standard error
Axsen
Mountain
2009
REE
SP survey
2006
MNL
Hybrid
0.17
0.06
0/1
3314.61
3770.62
4226.62
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
flex fuel
0.11
0.14
0/1
-798.18
3547.46
7893.10
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
flex fuel
0.28
0.21
0/1
2136.93
8681.29
15225.64
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
EV
sport
0.51
0.19
0/1
-31927.98
-20378.31
-8828.64
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
EV
college
-0.19
0.10
0/1
-16227.85
-5442.05
5343.76
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
EV
sport
-0.41
0.38
0/1
-33619.55
-21246.51
-8873.47
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
EV
truck
-0.30
0.14
0/1
-29431.94
-19953.45
-10474.97
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
EV
truck
-0.26
0.23
0/1
-27910.74
-19008.01
-10105.28
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
EV
college
0.92
0.35
0/1
-17497.83
-5516.26
6465.30
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
methanol
0.64
0.17
0.84
0.44
0/1
3973.87
12587.18
21200.49
random coef.
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
methanol
1.18
0.32
1.33
0.92
0/1
-1850.62
13962.67
29775.96
random coef.
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
natural gas
0.24
0.15
2.07
0.49
0/1
-5018.84
4619.87
14258.57
random coef.
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
natural gas
0.42
0.26
3.66
0.98
0/1
.38447.77
5006.16
48460.08
random coef.
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
EV
short commute
0.36
0.16
0/1
-23513.54
-11391.65
730.23
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
EV
college
0.77
0.22
0/1
-16086.01
-2816.62
10452.78
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
methanol
0.48
0.15
0/1
8488.87
12796.83
17104.79
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
methanol
college
0.34
0.13
0/1
4318.81
6989.38
9659.95
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
natural gas
0.62
0.15
0.97
0.41
0/1
-7302.34
12956.44
33215.22
random coef.
Daziano
2013
REE
SP survey
1999
MXL
EV
-0.21
0/1
-16694.51
-2419.02
10766.28
random coef.
Daziano
2013
REE
SP survey
1999
MXL
Hybrid
1.05
0/1
-92.68
12143.78
24076.70
random coef.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
diesel
0.00
0.04
0/1
-517.55
-53.08
411.39
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
Hybrid
-3.27
0.34
1.22
0.39
0/1
-55137.40
-40155.75
-25174.09
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
Hybrid
-0.42
0.19
0/1
-612.66
-425.23
-237.81
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
Hybrid
-1.18
1.16
0/1
-2353.92
-1196.35
-38.78
standard error
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
diesel
0.21
0.13
0/1
2364.35
6304.92
10245.50
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
EV
-2.64
0.71
0/1
-98745.38
-77780.33
-56815.28
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
flex fuel
0.31
0.10
0/1
6447.77
9251.15
12054.52
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
Hybrid
0.17
0.10
0/1
2239.13
5038.04
7836.96
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
natural gas
-1.90
0.90
0/1
-82634.58
-55978.27
-29321.95
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
PHEV
0.45
0.08
0/1
10722.59
13110.70
15498.80
varied income
Hess
Tram
2006
TR-B
SP survey
1999
MXL
EV
-1.98
0.21
1.28
0.13
0/1
-59106.41
-35920.68
-12734.95
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
Hybrid
0.79
0.10
1.14
0.10
0/1
-6345.04
14351.07
35047.18
random coef.
Hidrue
Parsons
2011
REE
SP survey
2009
Other
EV
0.54
0.13
0/1
-476.33
2538.76
5553.85
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
EV
-7.46
-1.52
0/1
-19445.11
-24306.39
-29167.66
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
EV
-1.05
0.53
4.02
4.40
0/1
-75577.19
-15628.43
44320.32
random coef.
Kavalec
1999
Energy Journal
SP survey
1993
MXL
flex fuel
0.20
0.31
0/1
-1660.89
2952.70
7566.29
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
methanol
0.63
0.36
1.61
3.01
0/1
-14718.76
9335.30
33389.37
random coef.
Kavalec
1999
Energy Journal
SP survey
1993
MXL
natural gas
0.22
0.25
2.10
2.88
0/1
-28080.47
3295.69
34671.85
random coef.
Liu
2014
Energy Economics
RP survey
2009
MXL
Hybrid
income (<25k)
0/1
1020.50
1063.47
1106.45
random coef.
Liu
2014
Energy Economics
RP survey
2009
MXL
Hybrid
income (>100k)
0/1
1672.82
1897.60
2122.38
random coef.
Liu
2014
Energy Economics
RP survey
2009
MXL
Hybrid
income (25-50k)
0/1
1452.36
1615.38
1778.39
random coef.
Liu
2014
Energy Economics
RP survey
2009
MXL
Hybrid
income (76-99k)
0/1
1634.11
1847.08
2060.05
random coef.
Liu
2014
Energy Economics
RP survey
2009
MXL
Hybrid
income (50-76k)
0/1
1651.61
1869.84
2088.07
random coef.
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
EV
-1.57
0.58
0/1
-35812.02
-26284.78
-16757.54
standard error
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
EV
short commute
0.48
0.22
0/1
4336.81
8007.56
11678.31
standard error
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
EV
education
1.05
0.31
0/1
12500.48
17598.92
22697.37
standard error
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
Hybrid
multiple
1.01
0.22
0/1
11969.59
14703.78
17437.96
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
Hybrid
multiple
1.01
0.22
0/1
8721.43
11455.61
14189.80
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
PHEV
multiple
2.59
0.54
0/1
17072.58
17988.12
18903.65
varied interaction
B-5
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WIT
Range Dcsc.
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
PHEV
multiple
2.59
0.54
0/1
13824.42
14739.95
15655.49
varied interaction
Nixon
Saphores
2011
Grey
SP survey
2010
MXL
CGV
early adopter
0.23
0/1
17095.11
4685.74
1284.35
varied income
Nixon
Saphores
2011
Grey
SP survey
2010
MXL
EV
early adopter
0.49
0/1
36783.33
10082.25
2763.53
varied income
Nixon
Saphores
2011
Grey
SP survey
2010
MXL
fuel cell
early adopter
0.24
0/1
18271.75
5008.26
1372.76
varied income
Nixon
Saphores
2011
Grey
SP survey
2010
MXL
Hybrid
early adopter
0.16
0/1
11736.49
3216.95
881.76
varied income
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
EV
0/1
6407.65
10351.81
14295.96
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
EV
EV-oriented
2.50
0.14
0/1
28743.96
30650.97
32557.98
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
EV
GV-oriented
-2.07
0.99
0/1
-20568.57
-14164.10
-7759.62
standard error
Segal
1995
Energy Journal
SP survey
1994
MNL
CGV
0.81
0/1
6059.34
Segal
1995
Energy Journal
SP survey
1994
MNL
EV
0.39
0/1
2883.98
Shiau
Michalek
2009
TR-A
market data
2007
MXL
Hybrid
0.99
0.01
0/1
12347.45
12421.23
12495.00
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
Hybrid
-1.27
0.61
0.23
0.16
0/1
-8914.99
-7548.02
-6181.06
random coef.
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
CGV
6.58
3.85
0/1
9642.31
22635.05
35627.80
random coef.
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
EV
5.77
1.58
0/1
14526.99
19854.82
25182.65
random coef.
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
PHEV
6.92
0.10
0/1
23466.09
23809.76
24153.43
random coef.
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
EV
0.34
0.15
0/1
6515.46
11812.59
17109.71
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
flex fuel
0.32
0.25
0/1
2400.85
10975.31
19549.77
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
flex fuel
-0.13
0.27
0/1
-13597.23
-4409.91
4777.41
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
natural gas
-0.41
0.13
0/1
-18849.53
-14296.50
-9743.48
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
natural gas
0.15
0.10
0/1
1640.77
5059.04
8477.31
standard error
Train
Weeks
2005
Book
SP survey
2000
MXL
EV
-2.54
1.41
0/1
-53887.19
.34894.48
-15901.77
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
EV
-1.95
1.28
0/1
-96054.99
-43983.86
8087.27
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
Hybrid
0.83
1.19
0/1
-18990.48
18860.06
56710.59
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
Hybrid
0.87
1.46
0/1
-7604.20
12026.50
31657.21
random coef.
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
CGV
1.34
1.34
1.57
0/1
-1532.10
10255.29
22042.67
random coef.
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
EV
-2.20
-2.20
1.84
0/1
-30646.25
-16837.04
-3027.82
random coef.
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
Hybrid
0.52
0.52
0.93
0/1
-3016.00
3979.66
10975.32
random coef.
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
PHEV
-1.04
-1.04
1.78
0/1
-21291.86
-7959.33
5373.20
random coef.
Incentives
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Axsen
Mountain
2009
REE
RP survey
2006
MNL
purchase subsidy
0 00
0.00
$ $CAN
0.71
0.83
0.95
standard error
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
purchase subsidy
0 72
0.47
0.00
0.00
0 1
2747.43
7512.34
12277.25
random coef.
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
purchase subsidy
0.52
0.58
0/1
-404.83
4564.98
9534.78
random coef.
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
no gas guzzler tax ($1000)
-2526.10
394.09
0/1
2854.00
3381.54
3909.08
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
no gas guzzler tax ($1300)
-1488.80
285.21
0/1
1611.17
1992.97
2374.76
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
no gas guzzler tax ($1700)
185.72
309.53
0/1
-662.97
-248.61
165.74
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
no gas guzzler tax ($2100)
-3477.60
434.16
0/1
4074.08
4655.26
5236.44
standard error
Gallagher
Muehlegger
2011
JEEM
market data
2011
Other
HOV access
-0.06
0.06
0/1
-333.07
-170.73
-8.38
standard error
Gallagher
Muehlegger
2011
JEEM
market data
2011
Other
HOV access
0.65
0.23
0/1
9269.72
14114.01
18958.30
standard error
Gallagher
Muehlegger
2011
JEEM
market data
2011
Other
tax credit
0.02
0.02
$/$1000s
0.00
0.07
0.14
standard error
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
free parking
0.03
0.05
0/1
-680.11
901.54
2483.20
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
HOV access
0.05
0.06
0/1
-245.40
1390.62
3026.64
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
purchase subsidy
0.06
0.06
0/1
16.36
1652.83
3289.30
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
tax credit
0.16
0.05
0/1
3108.99
4625.57
6142.15
varied income
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
cashback
GV-oriented
0.30
0.05
$/1000s
1.65
1.95
2.24
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
cashback
EV-oriented
0.19
0.02
$/1000s
2.37
2.62
2.88
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
cashback
$/1000s
2.04
2.32
2.59
standard error
B-6
-------
Model Availability
hirst Author
Second Author
Pub Vear
Journal
Data Tvpe
Dollar Veai*
Stat Model
Attribute
Interaction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central WIT
Nigh WIT
Range Dcsc.
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
number of models
0.69
0.08
$/MakeModel
5009.03
5670.91
6332.79
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
number of models
0.72
0.08
$/MakeModel
6113.66
6841.48
7569.29
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
1.38
0.05
$/MakeModel
0.72
0.75
0.78
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
1.38
0.05
$/MakeModel
2.16
2.23
2.31
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
1.38
0.05
$/MakeModel
21.36
22.12
22.89
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
0.80
0.02
$/MakeModel
0.55
0.56
0.58
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
0.80
0.02
$/MakeModel
1.18
1.21
1.24
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
0.80
0.02
$/MakeModel
1.89
1.94
1.98
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
0.82
0.02
$/MakeModel
0.53
0.54
0.55
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
0.82
0.02
$/MakeModel
0.93
0.96
0.98
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
0.82
0.02
$/MakeModel
3.07
3.15
3.23
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
1.18
0.07
$/MakeModel
0.67
7.09
7.52
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
1.18
0.07
$/MakeModel
1.04
11.11
11.78
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
number of models
income
1.18
0.07
$/MakeModel
1.81
19.26
20.41
standard error
Non-fuel Operating Costs
h irst Author
Second Author
Pub Year
Journal
Data I'vpe
Dollar Year
Stat Model
Attribute
Interaction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central \\ IT
Nigh WIT
Range Dcsc.
Dasgupta
Siddarth
2007
J. Marketing Research
market data
2000
MXL
maintenance cost reduction
-7.47
2.48
$/C$/yr)
4.19
6.27
8.35
standard error
Greene
2001
Grey
Lit. Review
1990
Other
battery cost reduction
0.00
$/$
0.93
Greene
2001
Grey
Lit. Review
1990
Other
maintenance cost reduction
0.00
$/($/yr)
1.00
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
battery cost reduction
EV
0.00
0.00
$/yr
23.12
28.69
34.26
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
maintenance cost reduction
0.00
0.00
$/($/yr)
12.38
16.21
20.04
standard error
Performance
First Author
Second Author
Pub Year
Journal
Data Tvpe
Dollar Yeai*
Stat Model
Attribute
1 nlcraclion
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Dcsc.
Axsen
Mountain
2009
REE
RP & SP
2006
MNL
horsepower
0.01
0.00
$/s
878.12
938.64
999.16
standard error
Axsen
Mountain
2009
REE
SP survey
2006
MNL
horsepower
-0.03
$/s
934.73
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
horsepower
-0.03
0.00
$/s
1596.09
1706.09
1816.09
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
horsepower
-0.03
0.00
$/s
998.17
1059.61
1121.06
standard error
Axsen
Mountain
2009
REE
RP & SP
2006
MNL
horsepower
-0.03
0.01
$/s
412.19
677.94
943.69
standard error
Axsen
Mountain
2009
REE
RP & SP
2006
MNL
horsepower
0.00
0.00
$/s
43.31
46.30
49.28
standard error
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
horsepower
0.00
0.00
5.73
0.34
$/hp
-58630.89
0.00
58630.89
random coef.
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
horsepower
0.00
0.00
0.00
7.57
$/hp
-64778.51
0.00
64778.51
random coef.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
acceleration (0-60) s faster
2.88
2.02
2.88
4.63
$/s
-9.04
15.77
40.58
random coef.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
acceleration (0-60) s faster
2.19
0.90
2.19
1.59
$/s
6.33
21.93
37.53
random coef.
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
acceleration (0-60) s faster
-0.04
0.02
$/s
267.44
690.17
1112.90
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
acceleration (0-60) s faster
income
-0.08
0.02
$/s
1077.34
1513.61
1949.89
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
acceleration (0-60) s faster
income
0.08
-0.05
$/s
-2494.35
-1546.88
-599.42
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
top speed
0.00
0.00
$/mph
-26.51
27.62
81.75
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
top speed
0.00
0.00
$/mph
29.02
74.88
120.75
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
acceleration (0-60) s faster
-0.15
0.04
$/s
776.74
1066.25
1355.76
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
acceleration (0-60) s faster
-0.33
0.38
$/s
-519.61
3287.55
7094.71
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
acceleration (0-60) s faster
-0.09
0.02
$/s
772.73
1048.70
1324.67
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
top speed
0.63
0.24
$/mph
46.03
74.97
103.92
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
top speed
1.25
2.63
$/mph
-231.67
209.68
651.02
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
top speed
0.39
0.14
$/mph
48.45
75.37
102.28
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
acceleration (0-60) s faster
-1.05
0.83
$/s
265.02
1299.02
2333.02
random coef.
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
top speed
0.36
0.83
$/mph
-98.39
75.32
249.03
random coef.
Daziano
2013
REE
SP survey
1999
MXL
high perf
0.21
0/1
-2722.56
2405.12
7779.57
random coef.
Daziano
2013
REE
SP survey
1999
MXL
low perf
-0.60
0/1
-11488.04
-6896.77
-1743.60
random coef.
B-7
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coeir.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
acceleration (0-60) s faster
0.27
0.10
$/s
13.59
21.56
29.54
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
acceleration (0-60) s faster
-1643.50
90.60
$/s
2078.77
2200.06
2321.34
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
braking distance
-115.88
10.90
$/ft
-169.71
-155.12
-140.53
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
turning circle
-901.00
50.17
$/ft
-1273.27
-1206.12
-1138.96
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
acceleration (0-60) s faster
van
0.26
0.12
$/s
516.87
961.20
1405.52
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
acceleration (0-60) s faster
SUV
0.20
0.13
$/s
766.53
1681.36
2596.20
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
acceleration (0-60) s faster
pickup
0.23
0.13
$/s
569.28
1078.82
1588.37
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
acceleration (0-60) s faster
0.61
0.04
$/s
1835.44
1976.24
2117.05
standard error
Feng
Fullerton
2013
J. Regulatory Economics
market data
2000
Hedonic
cylinders
1993.56
411.23
$/# of cylinders
-56.20
1277.26
2610.72
standard error
Feng
Fullerton
2013
J. Regulatory Economics
market data
2000
Hedonic
cylinders
3150.55
288.44
$/# of cylinders
36.11
1004.75
1973.38
standard error
Fifer
Bunn
2009
Grey
market data
2002
Hedonic
displacement
3954.07
406.95
$/mA3
4673.74
5209.94
5746.14
standard error
Fifer
Bunn
2009
Grey
market data
2002
Hedonic
horsepower
29.91
6.06
$/hp
31.42
39.41
47.40
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
acceleration (0-60) s faster
0.61
0.13
0.02
0.35
$/s
12.15
12.58
13.02
random coef.
Greene
2001
Grey
Lit. Review
1990
Other
acceleration (0-60) s faster
-0.35
$/s
460.84
Greene
2001
Grey
Lit. Review
1990
Other
acceleration (0-60) s faster
-0.24
$/s
532.66
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
horsepower
midsize
$/hp
13.84
Greene
Duleep
2004
Grey
Lit. Review
2002
Other
horsepower
smallSUV
$/hp
13.81
Goldberg
1995
Econometrica
RP survey
1982
NMNL
horsepower/cid
3.58
0.86
hp/cid
1509.64
2651.14
5005.93
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
horsepower/cid
-0.02
0.59
hp/cid
-529.54
-12.46
497.46
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
horsepower/cid
0.17
1.16
hp/cid
-4581.32
587.78
6233.86
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
horsepower/cid
young
0.51
2.18
hp/cid
-7685.62
1736.31
12567.25
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
horsepower/cid
young
-0.20
0.90
hp/cid
-969.49
-140.50
607.70
standard error
Goldberg
1995
Econometrica
RP survey
1982
NMNL
horsepower/cid
young
0.28
1.76
hp/cid
-1639.50
199.88
2261.46
standard error
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
acceleration (0-60) s faster
9.90
$/s
1760.21
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
acceleration (0-60) s faster
13.60
$/s
651.02
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
acceleration (0-60) s faster
13.70
$/s
2325.13
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
acceleration (0-60) s faster
-1.27
5.77
$/s
-4457.51
1290.96
7039.43
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
acceleration (0-60) s faster
-1.17
0.26
$/s
938.06
1192.28
1446.51
standard error
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
acceleration (0-60) s faster
-0.04
0.01
$/s
1241.66
1051.80
861.95
varied income
Hess
Tram
2006
TR-B
SP survey
1999
MXL
high perf
0.18
0.06
0.61
0.09
0/1
-7711.40
3332.01
14375.42
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
low perf
-0.49
0.06
0.55
0.10
0/1
-18925.83
-8926.68
1072.46
random coef.
Hidrue
Parsons
2011
REE
SP survey
2009
Other
acc. (1% faster)
2.20
0.88
$/%
2462.26
4049.78
5637.29
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
acc. (1% faster)
0.59
0.06
$/%
4469.79
4977.96
5486.13
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
acc. (1% faster)
1.97
0.82
$/%
8631.22
14587.98
20544.74
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
acc. (1% faster)
0.33
0.06
$/%
9129.39
11200.41
13271.43
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
acc. (1% faster)
0.15
0.05
$/%
-3309.21
-5091.10
-6872.98
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
acc. (1% faster)
1.10
0.79
$/%
-2443.67
-8145.57
-13847.47
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
acceleration (0-60) s faster
age
$/s
674.62
1140.64
1606.66
varied interaction
Kavalec
1999
Energy Journal
SP survey
1993
MXL
top speed
age
$/mph
4.03
68.90
133.77
varied interaction
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
acceleration (0-60) s faster
8.25
5.36
$/s
-115.39
34.00
183.39
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
acceleration (0-60) s faster
38.75
9.51
$/s
-130.85
198.40
527.65
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
acceleration (0-60) s faster
42.18
11.84
$/s
-166.62
185.52
537.67
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
acceleration (0-60) s faster
47.20
10.70
$/s
-92.09
172.81
437.71
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
acceleration (0-60) s faster
38.75
19.41
$/s
-473.60
198.40
870.40
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
acceleration (0-60) s faster
9.53
9.13
$/s
-183.95
33.58
251.10
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
acceleration (0-60) s faster
0.01
0.00
$/s
-229.55
5.30
240.14
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
displacement
0.00
0.00
$/inA3
0.00
0.00
0.00
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
displacement
0.00
0.00
$/inA3
-0.01
0.00
0.01
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
horsepower
0.01
0.00
$/hp
-967.30
9.18
985.65
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
horsepower
0.00
0.00
$/hp
-109.09
1.24
111.58
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
horsepower
0.01
0.00
$/hp
-360.17
8.31
376.79
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
horsepower
0.00
0.00
$/hp
-117.03
1.33
119.70
standard error
Lave
Tram
1979
TR-A
market data
1976
MNL
acceleration (0-60) s faster
age
-0.02
-0.01
$/s
157.70
250.93
343.81
varied interaction
Liu
2014
Energy Economics
RP survey
2009
MXL
acceleration (0-60) s faster
0.14
0.01
$/s
140.92
156.35
171.78
random coef.
McCarthy
1996
RE Stat
RP survey
1989
MNL
horsepower
0.01
0.00
$/hp
297.02
355.01
412.99
standard error
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
acceleration (0-60) s faster
-0.13
0.03
$/s
1009.62
1261.54
1514.43
standard error
McManus
2007
Business Economics
market data
2002
Hedonic
acceleration (0-60) s faster
630.61
23.67
$/s
4760.20
4945.84
5131.49
standard error
Petrin
2002
J. Political Economy
market data
1983
BLP
acceleration (0-60) s faster
3.40
0.10
$/s
1032.53
1063.17
1093.82
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
acceleration (0-60) s faster
-2.83
4.43
$/s
-901.90
-355.91
190.08
random coef.
B-8
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coelf.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Petrin
2002
J. Political Economy
market data
1983
BLP
acceleration (0-60) s faster
3.40
0.10
$/s
1534.47
1580.01
1625.56
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
acceleration (0-60) s faster
3.40
0.10
$/s
377.71
388.92
400.13
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
acceleration (0-60) s faster
-2.83
4.43
$/s
-856.51
-338.00
180.51
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
acceleration (0-60) s faster
-2.83
4.43
$/s
-2180.39
-860.43
459.52
random coef.
Shiau
Michalek
2009
TR-A
market data
2007
MXL
acceleration (0-60) s faster
0.24
0.00
$/s
1778.05
1807.32
1836.60
random coef.
Skerlos
Raichur
2013
Grey
market data
2008
MXL
acceleration (0-60) s faster
-19.99
11.71
0.41
1.24
$/s
533.97
1289.14
2044.31
random coef.
Skerlos
Raichur
2013
Grey
market data
2008
MXL
horsepower
2.49
1.00
0.08
0.05
$/hp
143.23
147.99
152.74
random coef.
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
acceleration (0-60) s faster
-0.06
0.01
$/s
1025.67
1275.23
1524.78
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
top speed
0.00
0.00
$/mph
58.17
115.19
172.22
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
top speed
0.01
0.00
$/mph
140.00
178.99
217.99
standard error
Train
Weeks
2005
Book
SP survey
2000
MXL
high perf
0.55
0.96
0/1
-17169.44
12392.61
41954.66
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
high perf
0.60
1.95
0/1
-18035.85
8322.76
34681.37
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
low perf
0.36
0.76
0/1
-5268.29
4932.82
15133.93
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
low perf
0.25
1.18
0/1
-31511.12
5731.10
42973.32
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
low perf
0.36
0.76
0/1
-5268.29
4932.82
15133.93
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
low perf
0.25
1.18
0/1
-31511.12
5731.10
42973.32
random coef.
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
acceleration (0-60) s faster
0.03
0.00
$/s
5542.56
5543.53
5544.51
random coef.
Walls
1996
RE Stat
market data
1990
Hedonic
acceleration (0-60) s faster
8.20
5.55
$/s
326.01
1173.02
2020.03
standard error
Whitefoot
Fowlie
2011
Grey
market data
2006
BLP
acceleration (0-60) s faster
1.13
0.40
$/s
31.18
34.64
38.71
standard error
Pollution
h irst Author
Second Author
Pub Vear
Journal
Data Tvpc
Dollar Veai*
Stat Model
Attribute
Interaction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central \\ TP
High WTP
Range Desc.
Brownstone
Bunch
1996
Transportation Econ.
SP survev
1993
NMNL
emissions reduction
children
-0 25
0 22
$/10%
8356.52
76601.44
144846 36
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
emissions reduction
children
-0.46
0.14
$/10%
103065.41
144802.64
186539.88
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
emissions reduction
children
-0.54
0.30
$/10%
76777.36
168535.66
260293.97
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
emissions reduction
children
-0.03
0.26
$/10%
-72273.29
8212.87
88699.03
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
emissions reduction
-0.69
0.25
$/10%
51603.76
81735.61
111867.46
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
emissions reduction
0.40
0.10
$/10%
-83601.63
-66982.03
-50362.43
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
emissions reduction
-0.39
0.14
$/10%
47822.73
75953.74
104084.76
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
emissions reduction
-0.70
0.83
$/10%
-28708.63
145003.63
318715.88
random coef.
Hidrue
Parsons
2011
REE
SP survey
2009
Other
emissions reduction
0.75
0.47
$/10%
189.38
488.73
788.08
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
emissions reduction
0.12
0.06
$/10%
173.55
358.41
543.28
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
emissions reduction
0.19
0.06
$/10%
262.46
378.32
494.19
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
emissions reduction
0.90
0.36
$/10%
237.72
390.99
544.25
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
emissions reduction
1.20
0.39
$/10%
281.46
411.57
541.67
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
emissions reduction
0.37
0.06
$/10%
489.70
581.64
673.57
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
emissions reduction
-0.95
0.27
$/10%
102201.19
141968.19
181735.19
standard error
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
emissions reduction
-0.79
0.20
$/10%
100047.31
132461.39
164891.88
standard error
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
emissions reduction
0.01
0.00
$/10%
266.26
297.28
328.30
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
emissions reduction
0.00
0.00
$/10%
1158.43
1491.32
1824.20
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
emissions reduction
0.00
0.00
$/10%
-975.74
-543.99
-112.25
standard error
Prestige
First Author
Second Author
Pub Year
Journal
Data Tvpe
Dollar Year
Stat Model
Attribute
1 nlcraction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central \\ I P
High WTP
Range Desc.
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
foreign
-0.29
0.11
0/1
-7806.36
-5696.53
-3586.70
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
foreign
-0.26
0.13
0/1
-6547.12
-4381.53
-2215 Q5
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
new
1.07
0.25
0/1
13799.31
18012.96
22226.62
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
new
0.77
0.23
0/1
10612.34
15034.14
19455.95
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
used
0.23
0.23
0/1
0.00
4463.26
8926.52
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
used
0.47
0.25
0/1
3626.09
7822.96
12019.83
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
American Motors
-0.12
0.09
0/1
-2800.09
-1645.55
-491.01
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
Chrysler
0.06
0.01
0/1
544.09
729.88
915.67
standard error
B-9
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coeir.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
German
0.40
0.03
0/1
4936.64
5308.22
5679.79
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
GM
0.02
0.01
0/1
159.25
291.95
424.66
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
Japanese
0.24
0.01
0/1
3012.41
3198.20
3383.99
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
resale value retained
0.12
0.04
$/%
1141.27
1605.74
2070.21
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
luxury
0.24
0.03
0/1
7519.90
8386.90
9253.90
standard error
Feng
Fullerton
2013
J. Regulatory Economics
market data
2000
Hedonic
foreign
2371.11
894.32
0/1
-809.84
2219.46
5248.76
standard error
Feng
Fullerton
2013
J. Regulatory Economics
market data
2000
Hedonic
foreign
1417.36
1584.27
0/1
1289.31
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
Chrysler
0.11
0.04
0/1
810.48
1350.80
1891.13
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
European
-0.25
0.06
0/1
-3782.25
-3070.01
-2357.77
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
GM
-0.35
0.04
0/1
.4764.66
-4298.02
-3831.37
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
Japanese
0.19
0.04
0/1
1866.57
2333.21
2799.85
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
Korean
-0.51
0.06
0/1
-6987.35
-6262.82
-5538.30
standard error
Gramlich
2008
Grey
market data
2007
NMNL
Asian
-0.06
0.01
0/1
-2424.55
-2078.19
-1731.82
standard error
Gramlich
2008
Grey
market data
2007
NMNL
European
0.14
0.03
0/1
3810.01
4849.11
5888.20
standard error
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
1 -2 yrs
-0.17
0.08
0/1
-7365.05
-5126.43
-2887.81
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
3+yrs
-0.38
0.06
0/1
-13112.15
-11313.50
-9514.85
varied income
Sexton
Sexton
2014
JEEM
market data
2010
Other
prius
0/1
199.19
1122.69
2046.19
standard error
Sexton
Sexton
2014
JEEM
market data
2010
Other
prius
0/1
1759.10
3658.92
5558.74
standard error
Sexton
Sexton
2014
JEEM
market data
2010
Other
prius
0/1
732.95
1524.54
2316.13
standard error
Sexton
Sexton
2014
JEEM
market data
2010
Other
prius
0/1
248.99
1403.37
2557.75
standard error
Sexton
Sexton
2014
JEEM
market data
2010
Other
prius
0/1
2198.86
4573.64
6948.41
standard error
Sexton
Sexton
2014
JEEM
market data
2010
Other
prius
0/1
83.00
467.79
852.58
standard error
Shiau
Michalek
2009
TR-A
market data
2007
MXL
foreign
-0.83
0.00
0/1
-10426.05
-10413.76
-10401.46
standard error
Walls
1996
RE Stat
market data
1990
Hedonic
European
0.26
0.04
0/1
1473.47
5301.73
9129.98
standard error
Range
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
range
0.01
0.00
$/mi
-138.95
60.95
260.85
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
range
0.01
0.00
$/mi
-99.08
127.16
353.40
standard error
Brownstone
Bunch
2000
TR-B
RP & SP
1995
MXL
range
1.00
0.24
$/mi
-7.32
79.09
165.50
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
range
1.78
0.50
$/mi
-31.20
84.35
199.89
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
range
2.48
1.61
$/mi
-289.92
160.82
611.57
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
range
0.59
0.83
$/mi
-51.03
122.68
296.39
random coef.
Daziano
2013
REE
SP survey
1999
MXL
range
1.17
$/mi
37.45
104.32
164.91
random coef.
Greene
2001
Grey
Lit. Review
1990
Other
range
-233.90
$/mi
3.23
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
range BEV
-18.95
1.90
$/mi
-211.67
-192.75
-173.83
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
range BEV
-19.50
1.98
$/mi
-218.11
-198.33
-178.55
standard error
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
range BEV
-12.73
10.49
$/mi
-156.01
-86.31
-16.62
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
range BEV
-13.69
1.96
$/mi
-105.87
-92.85
-79.83
standard error
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
range BEV
-18.45
4.18
$/mi
-305.80
-250.30
-194.80
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
range BEV
-20.14
1.98
$/mi
-299.43
-273.14
-246.85
standard error
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
range PHEV
0.82
2.20
$/mi
-135.41
83.62
302.65
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
range PHEV
0.03
1.78
$/mi
-174.91
2.75
180.40
standard error
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
range PHEV
3.21
8.66
$/mi
-268.76
163.12
595.01
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
range PHEV
1.70
1.75
$/mi
-1.07
86.22
173.50
standard error
Helveston
Liu
2015
TR-A
SP survey
2013
MXL
range PHEV
3.30
7.14
$/mi
-93.95
84.03
262.01
random coef.
Helveston
Liu
2015
TR-A
SP survey
2013
MNL
range PHEV
2.65
1.77
$/mi
23.18
67.40
111.61
standard error
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
range
0.24
0.17
$/mi
16.76
53.21
89.65
varied income
Hess
Tram
2006
TR-B
SP survey
1999
MXL
range
0.56
0.07
$/mi
92.98
100.96
109.63
random coef.
Hidrue
Parsons
2011
REE
SP survey
2009
Other
range
1.32
0.73
$/mi
13.06
28.67
44.28
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
range
0.53
0.06
$/mi
47.02
52.77
58.51
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
range
1.94
0.72
$/mi
20.13
31.60
43.08
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
range
0.92
0.06
$/mi
64.46
68.70
72.93
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
range
1.28
0.07
$/mi
60.47
63.72
66.97
standard error
Hidrue
Parsons
2011
REE
SP survey
2009
Other
range
2.60
0.70
S/mi
20.76
28.24
35.72
standard error
B-10
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coeff.
SE
mu
signia
Standard Units
Low WTP
Central WTP
High WTP
Range Dcsc.
Kavalec
1999
Energy Journal
SP survey
1993
MXL
range
age
$/mi
124.77
162.05
181.44
varied interaction
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
range
0.01
0.00
$/mi
96.78
113.19
127.95
standard error
Nixon
Saphores
2011
Grey
SP survey
2010
MXL
range
0.03
$/mi
16.07
62.12
240.19
random coef.
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
GV-oriented
0.26
0.19
$/mi
3.86
13.51
23.17
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
EV-oriented
-0.17
0.07
$/mi
-26.77
-18.78
-10.79
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
$/mi
-12.89
-4.15
4.59
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
EV-oriented
-0.79
0.08
$/mi
-483.21
-436.29
-389.38
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
$/mi
-432.36
-371.69
-311.03
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
GV-oriented
-1.13
0.30
$/mi
-370.96
-293.68
-216.39
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
GV-oriented
-0.37
0.19
$/mi
-48.84
-32.05
-15.27
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
EV-oriented
-0.44
0.07
$/mi
-94.43
-81.00
-67.57
standard error
Parsons
Hidrue
2014
Energy Economics
SP survey
2009
Other
range
$/mi
-73.78
-58.83
-43.88
standard error
Segal
1995
Energy Journal
SP survey
1994
MNL
range
0.01
$/mi
59.77
Tanaka
Ida
2014
TR-A
SP survey
2012
MXL
range
0.00
0.00
$/mi
1.93
2.20
2.47
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
range
non-large *multifuel
0.00
0.00
$/mi
34.10
54.64
75.18
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
range
non-large *onefuel
0.00
0.01
$/mi
-411.70
78.42
568.54
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
range
large*onefuel
0.00
0.00
$/mi
96.35
123.49
150.63
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
range
large*multifuel
0.00
0.00
$/mi
75.64
97.57
119.49
standard error
Train
Weeks
2005
Book
SP survey
2000
MXL
range
0.57
0.36
$/mi
-20.51
128.89
278.30
random coef.
Train
Weeks
2005
Book
SP survey
2000
MXL
range
0.76
0.43
$/mi
47.68
105.10
162.52
random coef.
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
range
0.01
0.01
0.00
S/mi
78.67
90.31
101.95
random coef.
Reliability
First Author
Second Author
Pub Year
Journal
Data Tvpe
Dollar Year
Stat Model
Attribute
1 nlcraction
Coeff.
SE
mu
sigma
Standard Units
Low WTP
Central \\ I P
High WTP
Range Dcsc.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
reliability index (->=2)
0.03
0.01
0/1
278.68
451.20
623.72
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
reliability index
103.94
76.99
$/scale [1,51
36.07
139.14
242.20
standard error
McCarthy
1996
RE Stat
RP survey
1989
MNL
reliability index
0.01
0.00
$/scale [1,51
228.94
317.97
407.01
standard error
McCarthy
Tay
1998
TR-E
RP survey
1989
NMNL
reliability index
0.01
0.00
$/scale [1,51
108.67
178.27
247.87
standard error
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
reliability index
female
0.39
0.06
$/scale [1,51
4788.26
5606.34
6424.42
standard error
Walls
1996
RE Stat
market data
1990
Hedonic
reliability index
0.04
3.56
$/scale [1,51
195.24
702.49
1209.75
standard error
Safety
hirst Author
Second Author
Pub Year
Journal
Data Tvpc
Dollar Year
Stat Model
Attribute
Interaction
Coeff
SE
mu
sigma
Standard Units
Low WTP
Central \\ I P
High WTP
Range Dcsc.
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
crash test rating ("front + side)
191.25
70.31
$/scale [1,101
161.89
256.02
350.14
standard error
Fifer
Bunn
2009
Grey
market data
2002
Hedonic
airbags
138.27
447.02
0/1
-406.82
182.18
771.19
standard error
McCarthy
1996
RE Stat
RP survey
1989
MNL
safety index
0.24
0.08
0/1
6067.90
9011.73
11955.56
standard error
McCarthy
Tay
1998
TR-E
RP survey
1989
NMNL
airbags
0.22
0.07
0/1
4171.75
6100.20
8028.65
standard error
Size
First Author
Second Author
Pub Year
Journal
Data Tvpe
Dollar Year
Stat Model
Attribute
Interaction
Coeff.
SE
mu
sigma
Sta 1 1 I s
Low WTP
Central WTP
High WTP
Range Dcsc.
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
footprint
3.26
8.77
$/ft
4285.86
4411.90
4537.95
random coef.
Beresteanu
Li
2011
Int. Econ. Rev.
market data
2006
BLP
footprint
2.68
0.00
9.36
$/ftA2
2827.38
3010.67
3193.96
random coef.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
footprint
2.60
0.29
2.60
1.51
$/ftA2
27.29
63.22
99.15
random coef.
Berry
Levinsohn
1995
Econometrica
market data
1983
BLP
footprint
3.46
0.61
3.46
2.06
$/ftA2
19.12
45.79
72.45
random coef.
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
luggage space (%)
0.49
0.35
$/%
4587.65
15292.18
25996.71
standard error
Brownstone
Bunch
1996
Transportation Econ.
SP survey
1993
NMNL
luggage space (%)
0.62
0.35
$/%
8885.52
19504.79
30124.07
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
luggage space (%)
1.56
0.46
5.38
1.29
$/%
-79637.24
32610.17
144857.58
random coef.
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
size (index)
1.54
0.53
6.81
2.07
0 to 0.3 scale
-109889.80
32151.16
174192.13
random coef.
B-ll
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
CoelT.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
luggage space
-0.04
0.01
$/ftA3
-663.53
-530.82
-398.12
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
weight
18.50
0.61
$/lb
23.95
24.76
25.58
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
weight
pickup
-0.87
0.34
$/lb
3.66
8.24
12.82
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
weight
1.81
0.11
$/lb
17.25
18.35
19.45
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
weight
van
2.05
0.33
$/lb
12.03
14.29
16.56
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
weight
SUV
-0.58
0.34
$/lb
5.63
10.23
14.82
standard error
Fifer
Bunn
2009
Grey
market data
2002
Hedonic
weight
7.61
0.51
$/lb
9.35
10.03
10.70
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
footprint
$/ftA2
5975.44
6301.47
6627.49
random coef.
Goldberg
1995
Econometrica
RP survey
1982
NMNL
footprint
-1.34
1.71
$/ftA2
-15537.91
-4247.82
3595.18
standard error
Greene
2001
Grey
Lit. Review
1990
Other
luggage space
0.12
$/ftA3
272.08
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
footprint
0.05
$/ftA2
1430.79
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
footprint
0.05
$/ftA2
1554.98
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
footprint
0.05
$/ftA2
390.04
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
footprint
0.05
$/ftA2
1539.68
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
size (index)
0.11
$/100ft
2417.67
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
size (index)
0.10
$/100ft
2985.83
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MXL
size (index)
9.62
0.00
0.10
0.00
$/100ft
1978.49
1978.49
1978.49
random coef.
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
width
0.99
$/ft
21050.94
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MNL
width
0.72
$/ft
22581.92
Haaf
Michalek
2014
J. Mechanical Design
market data
2004
MXL
width
0.95
0.19
0.95
0.19
$/ft
15708.63
19538.09
23367.56
random coef.
Kavalec
1999
Energy Journal
SP survey
1993
MXL
luggage space (%)
1.26
0.85
$/%
6179.23
18789.91
31400.58
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
length
0.42
0.12
$/ft?
-1.18
0.39
1.95
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
length
0.18
0.01
$/ft?
0.02
0.06
0.09
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
-0.25
0.27
$/lb
-0.57
-0.14
0.29
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
1.62
0.49
$/lb
0.07
0.48
0.89
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
1.47
0.29
$/lb
0.22
0.51
0.79
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
1.28
0.17
$/lb
0.27
0.43
0.59
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
1.14
1.20
$/lb
-0.98
0.49
1.96
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
-0.99
0.33
$/lb
-1.77
-0.91
-0.05
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
1.28
0.18
$/lb
0.24
0.40
0.56
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
-0.18
0.29
$/lb
-0.52
-0.09
0.34
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
0.82
0.80
$/lb
-0.54
0.30
1.14
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
1.14
0.53
$/lb
-0.16
0.49
1.14
standard error
Klier
Linn
2012
Rand J. Econ.
market data
2008
MNL
weight
1.47
0.68
$/lb
-0.14
0.45
1.05
standard error
Lave
Tram
1979
TR-A
market data
1976
MNL
weight
age
0.69
0.42
$/lb
38.90
87.36
129.74
varied income
Liu
2014
Energy Economics
RP survey
2009
MXL
volume
0.00
0.00
$/ftA3
1.18
1.18
1.19
random coef.
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.22
0.03
$/ftA3
1.01
1.17
1.33
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.22
0.03
$/ftA3
3.03
3.50
3.97
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.22
0.03
$/ftA3
29.96
34.61
39.26
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.16
0.03
$/ftA3
0.91
1.10
1.29
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.16
0.03
$/ftA3
1.97
2.36
2.76
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.16
0.03
$/ftA3
3.14
3.78
4.42
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.10
0.02
$/ftA3
0.49
0.62
0.75
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.10
0.02
$/ftA3
0.77
0.97
1.17
standard error
Liu
Tremblay
2014
TR-A
RP survey
2009
MNL
luggage space
income
0.10
0.02
$/ftA3
1.33
1.68
2.03
standard error
McCarthy
1996
RE Stat
RP survey
1989
MNL
length
0.02
0.00
$/ft
42.36
51.75
61.14
standard error
McCarthy
Tay
1998
TR-E
RP survey
1989
NMNL
length
0.05
0.00
$/ft
93.56
104.92
116.28
standard error
McCarthy
Tay
1998
TR-E
RP survey
1989
NMNL
luggage space
0.04
0.01
$/ftA3
741.89
974.92
1207.95
standard error
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
luggage space (%)
2.26
7.62
$/%
-86983.26
37809.32
162601.89
random coef.
McFadden
Tram
2000
J. Applied Econometrics
SP survey
1993
MXL
size (index)
5.78
26.93
0 to 0.3 scale
-344278.22
96627.10
537530.81
random coef.
McManus
2007
Business Economics
market data
2002
Hedonic
weight
10.50
15.00
$/lb
-5.93
13.83
33.60
standard error
Petrin
2002
J. Political Economy
market data
1983
BLP
footprint
4.80
0.46
$/ftA2
3490.65
3852.46
4214.27
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
footprint
4.80
0.46
$/ftA2
8886.03
9807.08
10728.13
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
footprint
4.60
0.14
$/ftA2
13936.66
14365.11
14793.57
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
footprint
4.60
0.14
$/ftA2
9377.81
9666.11
9954.42
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
footprint
4.60
0.14
$/ftA2
3430.50
3535.96
3641.43
random coef.
Petrin
2002
J. Political Economy
market data
1983
BLP
footprint
4.80
0.46
$/ftA2
3675.65
4056.64
4437.62
random coef.
Shiau
Michalek
2009
TR-A
market data
2007
MXL
footprint
0.04
0.00
$/ftA2
476.90
477.15
477.40
random coef.
Skerlos
Raichur
2013
Grey
market data
2008
MXL
footprint
0.79
2.00
0.31
0.21
$/ftA2
410802.69
676112.75
941422.81
random coef.
B-12
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
CoelT.
SE
mu
sigma
Standard Units
Low WTP
Central WTP
High WTP
Range Desc.
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
length-WB
0.03
0.01
$/ft
3584.08
4736.07
5888.06
standard error
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
wheelbase
0.05
0.01
$/ft
6670.36
8790.69
10911.02
standard error
Walls
1996
RE Stat
market data
1990
Hedonic
volume
0.01
0.00
$/ftA3
33.93
122.09
210.24
standard error
Whitefoot
Fowlie
2011
Grey
market data
2006
BLP
footprint
2.45
0.75
$/ftA2
633.86
905.08
1176.29
standard erro
Vehicle Class
First Author
Second Author
Pub Year
Journal
Data Tvpe
Dollar Year
Stat Model
Attribute
1 ntcraclion
Coeff.
SE
mu
sigma
Standard linils
Low WTP
Central W IT
High WTP
Range Desc.
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
large
0.00
0.00
0/1
-30755.80
-27322 92
-23890.04
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
large
-1.24
0.17
0/1
.39464.71
-34793.76
-30122.82
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
large
0.01
0.00
0/1
-16413.96
-14622.68
-12831.41
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
large
0.02
0.00
0/1
-9924.20
-8749.60
-7574.99
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
large
0.01
0/1
-12516.59
-11017.05
-9517.51
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
SUV
-1.38
0.19
0/1
-11023.30
-9778.73
-8534.17
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
SUV
-1.12
0.15
0/1
-8011.78
-7192.20
-6372.62
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
SUV
-1.45
0.18
0/1
-14361.96
-12937.39
-11512.83
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
SUV
-1.13
0.13
0/1
.35414.14
-31791.40
-28168.66
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
SUV
-1.33
-0.17
0/1
-27391.92
-24754.98
-22118.04
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
truck
-2.18
0.24
0/1
-67623.48
-61113.56
-54603.64
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
truck
-2.68
0.77
0/1
-22220.96
-20081.81
-17942.66
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
truck
-1.17
0.15
0/1
-30409.66
-28165.57
-25921.48
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
truck
-1.04
0.14
0/1
-24466.40
-22633.29
-20800.18
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
truck
-0.87
-0.18
0/1
-53388.63
-48708.76
-44028.90
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
van
-0.92
0.11
0/1
-26856.72
-20981.81
-15106.91
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
van
-0.68
0.20
0/1
-24530.72
-19164.63
-13798.53
standard error
Axsen
Mountain
2009
REE
RP survey
2006
MNL
van
-1.22
0.16
0/1
-9466.87
-8332.69
-7198.52
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
van
-1.21
-0.13
0/1
-21521.43
-17934.52
-14347.62
standard error
Axsen
Mountain
2009
REE
RP&SP
2006
MNL
van
-1.28
0.14
0/1
-13284.06
-11766.77
-10249.48
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
luxury
-0.28
0.21
0/1
-8175.50
-4700.49
-1225.49
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
luxury
-0.24
0.17
0/1
-8121.12
-4737.32
-1353.52
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
small
-0.47
0.15
0/1
-12101.46
-9220.16
-6338.86
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
small
-0.07
0.07
0/1
-1648.95
-782.95
83.04
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
small
-0.45
0.15
0/1
-10173.21
-7621.51
-5069.82
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
sport
hhsize>3
0.87
0.30
0/1
11158.15
17030.87
22903.58
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
sport
hhsize>3
0.85
0.31
0/1
9048.45
5136.97
19322.38
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
sport
-0.73
0.27
0/1
-14773.48
-9042.57
-3311.67
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
sport
hhsize>3
-1.07
0.39
0/1
-17331.75
-12717.06
-8102.38
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
station wagon
-1.53
0.07
0/1
-18921.38
-18114.70
-17308.02
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
station wagon
-0.94
0.25
0/1
-23299.85
-18342.44
-13385.02
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
SUV
-1.40
0.64
0/1
-39977.15
-27484.29
-14991.43
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
SUV
-1.18
0.79
0/1
-33172.05
-19876.37
-6580.69
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
SUV
0.37
0.42
0/1
-593.15
4353.70
9300.54
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
SUV
0.39
0.39
0/1
0.00
6547.12
13094.23
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
SUV
0.25
0.18
0/1
1415.04
4952.65
8490.26
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
SUV
0.94
0.15
0/1
9395.44
11198.61
13001.78
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
van
0/1
2366.87
3409.37
4451.87
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
van
0/1
2900.37
3227.79
3555.20
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
van
hhsize>3
0.88
0.27
0/1
10290.72
14823.34
19355.96
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
van
hhsize>3
1.05
0.27
0/1
15298.67
20574.07
25849.47
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
van
0/1
3518.69
4732.04
5945.38
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
van
hhsize>3
1.18
0.12
0/1
12610.30
14033.85
15457.40
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
truck
-0.39
0.41
0/1
-13396.41
-6580.69
235.02
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
truck
-0.38
0.16
0/1
-10538.26
-7438.77
-4339.28
standard error
Brownstone
Bunch
2000
TR-B
RP survey
1995
MNL
van
-0.36
0.44
0/1
-13312.47
-6009.92
1292.64
standard error
Brownstone
Bunch
2000
TR-B
RP&SP
1995
MXL
van
-0.45
0.20
0/1
-12670.82
-8711.19
-4751.56
standard error
Brownstone
Bunch
2000
TR-B
SP survey
1993
MXL
van
-1.21
0.08
0/1
-15243.87
-14342.29
-13440.70
standard error
B-13
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
Interaction
Coeff.
SE
mu
signia
Standard Units
Low WTP
Central WTP
High WIT
Range Dcsc.
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
non-compact
hh>2
0.25
0.83
0/1
-12238.72
5132.50
22503.73
random coef.
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
sport
0.70
0.16
0/1
11162.15
14562.95
17963.76
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
station wagon
-1.51
0.07
0/1
-32860.53
-31462.66
-30064.78
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
SUV
0.90
0.15
0/1
15606.15
18714.86
21823.57
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
truck
-1.09
0.06
0/1
-23993.40
-22825.03
-21656.66
standard error
Brownstone
Tram
1999
J. Econometrics
SP survey
1993
MXL
van
-0.82
0.06
0/1
-18255.85
-17087.48
-15919.10
standard error
Daziano
2013
REE
SP survey
1999
MXL
large
-0.48
0/1
-19574.63
-5565.61
7685.73
random coef.
Daziano
2013
REE
SP survey
1999
MXL
small
-3.65
0/1
-69405.58
-42316.69
-15225.48
random coef.
Daziano
2013
REE
SP survey
1999
MXL
small
-1.69
0/1
-33585.96
-19565.36
-7086.77
random coef.
Daziano
2013
REE
SP survey
1999
MXL
SUV
-0.16
0/1
-21256.82
-1862.93
16553.16
random coef.
Daziano
2013
REE
SP survey
1999
MXL
SUV
0.46
0/1
-7068.23
5303.78
15651.82
random coef.
Daziano
2013
REE
SP survey
1999
MXL
SUV
-0.97
0/1
-25124.01
-11182.19
907.13
random coef.
Daziano
2013
REE
SP survey
1999
MXL
truck
-1.53
0/1
-32976.57
-17707.07
-2631.04
random coef.
Daziano
2013
REE
SP survey
1999
MXL
truck
-0.84
0/1
-26540.90
-9674.93
5833.23
random coef.
Daziano
2013
REE
SP survey
1999
MXL
van
-0.56
0/1
-24797.31
-6507.50
11934.08
random coef.
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
convertible
0.35
0.06
0/1
3861.73
4618.15
5374.57
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
luxury
0.21
0.01
0/1
2654.11
2826.63
2999.14
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
station wagon
0.05
0.03
0/1
278.68
703.34
1128.00
standard error
Dreyfus
Viscusi
1995
J. Law and Economics
RP survey
1988
Hedonic
two-seater
-0.07
0.07
0/1
-1924.23
-955.48
13.27
standard error
Espey
Nair
2005
Contemp. Econ. Policy
market data
2001
Hedonic
luxury
15853.00
593.08
0/1
20427.55
21221.47
22015.39
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
large
-0.21
-0.03
0/1
-8145.06
-7087.96
-6030.86
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
midsize
-0.11
-0.02
0/1
-4394.39
-3677.12
-2959.84
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
SUV
8.19
3.67
0/1
156041.16
278377.44
400713.69
standard error
Fan
Rubin
2010
TRR
RP survey
2007
Hedonic
truck
11.45
3.66
0/1
267391.03
389271.59
511152.16
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
small
-0.06
0.27
1.32
0.26
0/1
-16983.30
-773.64
15436.02
random coef.
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
SUV
children
0.86
0.13
0/1
122.80
3070.01
6017.22
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
truck
rural
2.26
0.21
0/1
4052.41
8227.63
12402.84
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
van
children
2.08
0.27
0/1
-43102.96
-32664.92
-22226.88
standard error
Frischknecht
Whitefoot
2010
J. Mechanical Design
market data
2006
MXL
van
-4.49
0.50
2.65
0.40
0/1
-87679.52
-55137.40
-22595.28
random coef.
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
large
-0.16
0.18
0/1
-10076.09
-4772.88
530.32
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
midsize
0.24
0.11
0/1
3672.96
7012.01
10351.07
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
small
-0.11
0.11
0/1
-6540.62
-3270.31
0.00
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
sport
0.00
0.14
0/1
-3937.63
39.77
4017.18
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
SUV
0.39
0.17
0/1
6434.14
11460.81
16487.49
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
SUV
0.36
0.21
0/1
4440.53
10694.80
16949.06
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
SUV
0.26
0.15
0/1
3231.25
7719.11
12206.96
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
SUV
0.10
0.18
0/1
-2381.62
2910.87
8203.36
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
SUV
0.40
0.12
0/1
8194.71
11667.05
15139.39
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
SUV
0.27
0.16
0/1
3351.80
8072.66
12793.51
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
truck
0.03
0.16
0/1
-3905.71
857.35
5620.41
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
truck
-0.10
0.18
0/1
-8199.13
-2943.28
2312.58
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
van
0.00
-0.28
0/1
-8421.19
-83.38
8254.44
varied income
Hess
Fowler
2012
Transportmetrica
RP & SP
2009
NMNL
van
-0.55
0.28
0/1
-24391.33
-16233.70
-8076.06
varied income
Hess
Tram
2006
TR-B
SP survey
1999
MXL
large
-0.46
0.17
1.18
0.27
0/1
-29862.45
-8399.89
13062.68
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
small
-2.98
0.23
1.94
0.31
0/1
-89256.65
-54128.45
-19000.25
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
small
-1.33
0.17
1.12
0.29
0/1
.44374.10
-24057.33
-3740.56
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
SUV
-0.80
0.16
0.76
0.28
0/1
-28197.03
-14443.62
-690.21
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
SUV
-0.16
0.24
1.58
0.41
0/1
-31574.35
-2898.01
25778.32
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
SUV
0.33
0.15
0.78
0.33
0/1
-8102.59
6010.79
20124.18
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
truck
-1.29
0.18
1.04
0.28
0/1
-42243.60
-23404.23
-4564.87
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
truck
-0.77
0.19
1.59
0.31
0/1
-42823.14
-13995.21
14832.72
random coef.
Hess
Tram
2006
TR-B
SP survey
1999
MXL
van
-0.48
0.19
1.50
0.25
0/1
-35915.82
-8688.60
18538.62
random coef.
Kavalec
1999
Energy Journal
SP survey
1993
MXL
large
0.35
0.15
0/1
3097.64
5279.07
7460.50
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
midsize
0.31
0.10
0/1
3056.49
4608.00
6159.52
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
small
-0.42
0.27
0/1
-10301.49
-6293.13
-2284.77
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
small
-0.28
0.16
0/1
-6504.82
-4130.80
-1756.78
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
sport
0.75
0.16
0/1
8680.63
11139.73
13598.83
standard error
Kavalec
1999
Energy Journal
SP survey
1993
MXL
station wagon
age
0/1
-20594.34
-9465.79
-6725.59
varied interaction
Kavalec
1999
Energy Journal
SP survey
1993
MXL
SUV
0.97
0.15
0/1
12197.37
14435.42
16673.47
standard error
B-14
-------
First Author
Second Author
Pub Year
Journal
Data Type
Dollar Year
Stat Model
Attribute
1 nteraclion
Coeff.
SE
mu
signia
Standard Units
Low WTP
Central WTP
High WIT
Range Desc.
Kavalec
1999
Energy Journal
SP survey
1993
MXL
truck
age
0/1
-21101.37
-13048.55
-10364.27
varied interaction
Kavalec
1999
Energy Journal
SP survey
1993
MXL
van
age
0/1
-38489.48
-20146.96
-15225.79
varied interaction
McCarthy
1996
RE Stat
RP survey
1989
MNL
luxury
-0.49
0.13
0/1
-23393.44
-18494.81
-13596.18
standard error
McCarthy
1996
RE Stat
RP survey
1989
MNL
sport
-1.28
0.27
0/1
-57731.49
-47770.78
-37810.06
standard error
McCarthy
1996
RE Stat
RP survey
1989
MNL
truck
1.45
0.30
0/1
43019.11
54055.42
65091.74
standard error
McCarthy
Tay
1998
TR-E
RP survey
1989
NMNL
SUV
2.73
0.33
0/1
66959.95
76099.33
85238.71
standard error
McCarthy
Tay
1998
TR-E
RP survey
1989
NMNL
truck
1.79
0.32
0/1
41164.92
49915.81
58666.71
standard error
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
large
multiple
0.00
0.00
0/1
1830.95
2629.16
2852.19
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
large
multiple
0.00
0.00
0/1
-1417.22
-619.01
-395.98
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
luxury
multiple
2.18
0.44
0/1
12114.57
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
luxury
multiple
2.18
0.44
0/1
8866.40
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
small
multiple
0.00
0.00
0/1
5039.77
8572.17
11529.38
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
small
multiple
0.00
0.00
0/1
1791.60
5324.00
8281.21
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
small
multiple
-2.14
0.22
0/1
-10121.01
-9322.80
-9099.77
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
small
multiple
-2.14
0.22
0/1
-6872.84
-6074.63
-5851.60
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
small
multiple
-1.96
0.22
0/1
-5383.20
-1850.80
1106.41
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
small
multiple
-1.96
0.22
0/1
-2135.03
1397.37
4354.58
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
SUV
multiple
-1.38
0.21
0/1
1652.26
4386.44
7120.63
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
SUV
multiple
-1.38
0.21
0/1
-1595.91
1138.27
3872.46
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
truck
multiple
0.00
0.00
0/1
-7242.00
-4634.23
-2026.45
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
truck
multiple
0.00
0.00
0/1
-3993.83
-1386.06
1221.71
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
van
multiple
0.00
0.00
0/1
5317.16
6005.98
6694.81
varied interaction
Musti
Kockelman
2011
TR-A
SP survey
2009
MNL
van
multiple
0.00
0.00
0/1
2068.99
2757.82
3446.64
varied interaction
Shiau
Michalek
2009
TR-A
market data
2007
MXL
large
0.10
0.00
0/1
1204.73
1217.03
1229.33
standard error
Shiau
Michalek
2009
TR-A
market data
2007
MXL
luxury
0.56
0.00
0/1
6963.92
6988.51
7013.10
standard error
Shiau
Michalek
2009
TR-A
market data
2007
MXL
small
0.03
0.00
0/1
301.37
313.67
325.96
standard error
Shiau
Michalek
2009
TR-A
market data
2007
MXL
small
-0.12
0.00
0/1
-1580.38
-1555.79
-1531.20
standard error
Shiau
Michalek
2009
TR-A
market data
2007
MXL
sport
0.11
0.00
0/1
1355.80
1392.68
1429.57
standard error
Shiau
Michalek
2009
TR-A
market data
2007
MXL
two-seater
-0.77
0.00
0/1
.9647.40
-9598.22
.9549.04
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
sport
-0.30
0.38
0/1
-4041.46
-1783.00
475.47
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
SUV
nochildren
-0.05
0.34
0/1
-2317.90
-297.17
1723.56
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
SUV
children
0.44
0.05
0/1
0.00
2317.90
4635.79
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
truck
rural
1.11
0.06
0/1
-4338.63
-416.03
3506.56
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
truck
urban
-1.18
0.60
0/1
-10579.12
-7013.12
-3447.13
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
van
children
0.89
0.14
0/1
-12718.72
-8796.12
-4873.53
standard error
Skerlos
Raichur
2013
Grey
market data
2008
MXL
van
nochildren
-2.37
0.52
0/1
-17176.21
-14085.68
-10995.15
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
large
0.38
0.10
0/1
9596.64
13099.06
16601.48
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
large
-0.17
0.14
0/1
-10558.57
-5865.87
-1173.17
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
luxury
0.56
0.06
0/1
17344.93
19395.15
21445.38
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
small
-0.11
0.06
0/1
-5870.70
-3781.49
-1692.27
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
small
-0.64
0.10
0/1
-25454.82
-22123.04
-18791.26
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
SUV
0.12
0.12
0/1
-127.76
4130.97
8389.70
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
SUV
0.44
0.10
0/1
11652.64
15271.48
18890.31
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
SUV
0.33
0.09
0/1
8011.08
11241.36
14471.63
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
SUV
0.74
0.09
0/1
22216.62
25404.08
28591.54
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
van
-0.34
0.11
0/1
-15443.37
-11676.69
-7910.02
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
van
-0.61
0.13
0/1
-25459.88
-21077.80
-16695.72
standard error
Tompkins
Bunch
1998
UC ITS
SP survey
1995
MNL
truck
-0.64
0.07
0/1
-24691.18
-22202.15
-19713.12
standard error
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
luxury
leased
0.61
0.75
0/1
6813.61
7674.82
21900.77
standard error
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
SUV
children
2.80
0.90
0/1
40585.50
50384.63
76537.37
standard error
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
truck
0.07
6.85
0/1
-94249.92
1060.50
96370.93
random coef.
Train
Winston
2007
Int. Econ. Rev.
RP survey
2000
MXL
van
children
2.11
0.88
0/1
6965.98
14022.23
37639.06
standard error
Whitefoot
Fowlie
2011
Grey
market data
2006
BLP
sport
-0.47
0.31
0/1
-1977.30
-1191.95
-406.60
varied income
Whitefoot
Fowlie
2011
Grey
market data
2006
BLP
SUV
1.09
0.25
0/1
1176.92
1708.02
2239.11
varied income
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
sedan
0.55
0.55
0.92
0/1
-2705.52
4209.26
11124.04
random coef.
Zhang
Gensler
2011
J. Product Innov. Mgmt.
SP survey
2010
MXL
SUV
0.05
0.05
2.14
0/1
-15650.79
382.66
16416.12
random coef.
Whitefoot
Fowlie
2011
Grey
market data
2006
BLP
truck
0.04
0.31
0/1
-492.70
112.69
718.08
varied income
Whitefoot
Fowlie
2011
Grey
market data
2006
BLP
van
1.28
0.30
0/1
-9203.53
-14034.57
-13273.56
varied income
B-15
-------
APPENDIX C:
DISCUSSION OF l lll POTENTIAL BIAS FROM ESTIMATING WTP FROM RATIOS
OF ATTRIBUTE AND PRICE DERIVATIVES
Our general method for estimating the willingness to pay for an attribute is to divide the
derivative of the utility function with respect to the quantity of the attribute by the derivative
with respect to the price of a vehicle, and reversing the sign. Typically, the derivatives are linear
functions of the parameter estimates, so that the expected values of the derivative functions are
functions of the expected values of the estimated coefficients. In the case of mixed logit models
(MXL), the coefficients are assumed to be random variables with the variance representing
heterogeneity of preferences across the population. In the case of fixed coefficient models like
multinomial logit (MNL) or nested multinomial logit (NMNL), the coefficient estimates are
random variables with the variance representing the uncertainty of estimation from a sample. In
either case, our method requires calculating the ratio of two random variables or functions of
random variables. It is important to know whether that ratio is a good or a poor estimate of the
central tendency of preferences in the population.
It is also useful to describe the uncertainty associated with WTP estimates. In the case of
MNL or NMNL models, the standard errors of the coefficient estimates provide a basis for
characterizing their uncertainty. In the case of MXL models, the standard deviations of random
parameter estimates are intended to describe the heterogeneity of preferences in the sample
population. In either case, estimating a confidence interval for a WTP estimate would require
estimating the variance of a ratio of random variables. In general, published articles do not
provide sufficient information to calculate valid estimates of the variance of WTP estimates.
Rather than providing no information on uncertainty or heterogeneity, we provide ranges of
uncertainty based on the variance of the estimated attribute coefficient conditional on specific
values of the price derivative. While this method is less than ideal, until authors routinely
provide the covariances of coefficient estimates or simulated distributions of WTP estimates, we
believe it is preferable to no description of uncertainty.
In general, the expected value of the ratio of two random variables, is not equal to the
ratio of their expected values, E(a/p) ^ p = E(a)/E(P) = Since a/p is undefined at P = 0,
E(a/p) is also undefined if there is probability density > 0 at P = 0. Although many methods of
C-l
-------
estimating price coefficients allow finite probability density at P = 0, it can be neglected for
practical purposes.17
The ratio of two random variables is a non-linear function. A widely used approach to
estimate non-linear functions of random variables is the delta method. The delta method
approximates a non-linear function of random variables by means of Taylor series expansions. It
can be shown that the ratio of expected values, p, is a first order Taylor Series expansion
estimate of p. However, p is a biased estimate of E(a/p) in general even if the covariance of a
and P is zero. The second order Taylor Series expansion of E(a/p) is usually preferred because it
includes an estimate of the bias of p.
Define VAR(a) = oa2, VAR(P) = op2 and Cov(a,P) = oap, then
-------
errors will be half and more frequently less than half of the value of the coefficient. This implies
that Oa < a/2 and op< |P|/2. The covariance of two coefficient estimates is a function of their
correlation, 0 < c < 1.
Cov(a,(l) = capaa(jp (C-2)
Substituting these relationships into Equation C-l, we get the following approximation of the
maximum bias for statistically significant coefficients.
EC'//?)-S I1+K1-c«f)]
Equation C-3 implies that for uncorrelated coefficient estimates the bias will be smaller than one
fourth of the ratio of the coefficients and that the bias will disappear as the correlation between
coefficient estimates increases. For two coefficients with t-statistics of about 3.3, with a
correlation coefficient of 0.5, the bias would be about 5%.
For mixed logit models the bias is less easily bounded and could be important. In many
mixed logit models, the price coefficient is not assumed to be randomly distributed. In such
models, the uncertainty in the price derivative arises from estimation uncertainty while the
uncertainty in the attribute derivative arises from preference heterogeneity, as well as estimation
uncertainty. Assuming that the price coefficient is statistically significant, we again have op
< |P|/2. In that case, Equation C-l becomes approximately the following.
E(a/R)~ — + l—-CaB^ (C-4)
\ !p) lip 4 lip 2\P\
As Equation C-4 shows, it is not possible to make meaningful statements about the
importance of the bias term without knowing the correlation, or covariance, of the attribute and
price coefficients. Clearly, when the coefficients are uncorrelated, the bias will be approximately
one fourth or less of the ratio of the expected values. But when the coefficients are correlated one
cannot even know whether the bias is greater or less than that amount without knowing the
covariance. A partial correction excluding the term that includes the covariance could increase or
decrease the bias. When the price coefficient is itself randomly distributed across the population,
it is even more difficult to make statements about the size of bias. For many MXL models, the
only way to obtain valid estimates of the expected value of WTP for attributes is via a simulation
using the data set on which the model was estimated. The lack of availability of variance-
covariance matrices for all models from all the authors make performing such calculations
infeasible. Because of this, we calculate the ratios of the attribute and price derivatives using
mean values but caution that the resulting WTP estimates contain an unknown bias.
C-3
-------
One of the studies in our main sample, Nixon and Saphores, 2011, provided a variance-
covariance matrix for the four random parameters of its mixed logit model of alternative fuel
vehicle choice. Unfortunately, the variances and covariances are for the logarithms of the
lognormally distributed coefficients and not for the coefficients themselves. Recovering the
means, medians and variances of the lognormal coefficients is straightforward but untangling the
covariances, unfortunately, is not. Fortunately, Nixon and Saphores report the results of a model
simulation consisting of 500,000 draws repeated 100 times, by which they estimated trade-offs
between vehicle price and three other vehicle attributes. They state, "We chose to report the
median trade-off because it is less sensitive than the mean to large values in the tail of a
lognormal distribution" (Nixon and Saphores, 2011, p. 32). They estimated that a $1,000
increase in the price difference between an AFV and a conventional vehicle corresponded to a
$300 increase in annual fuel savings, a 17.5 mile increase in vehicle range, and 7.8 minute
reduction in refueling time. Converting their reported coefficients to the median of the
corresponding lognormal distribution (median = exp(coefficient)) and taking simple ratios of the
resulting values produced estimates of $295 per year in fuel costs, 17.5 miles of range and 7.5
minutes of refueling time. The implied biases are 1.7%, 0% and 3.8% of the simulated values,
respectively. While this is only one example, it gives us some confidence that our use of the ratio
of medians in MXL models with lognormal coefficients may produce useful indicators of the
central tendency of WTP for vehicle attributes in these models.
The uncertainty or heterogeneity of WTP estimates depends on the variance of the ratio
of the derivatives with respect to the attribute in question and vehicle price, both of which are
random variables. Let a and b be estimates of the population parameters a and (3. The second
order approximation to the variance of a/b is given by Equation C-5.
Unfortunately, in general, Equation C-5 cannot be calculated because the covariance of a
and b is almost never provided. Omitting the term in the numerator involving the covariance
would be as likely to increase bias as to decrease it. Instead, we provide an uncertainty or
heterogeneity interval conditional on the value of the price derivative. This is not a confidence
interval for WTP. The confidence interval could be larger or smaller, depending in large part on
whether the coefficient estimates in the cases of MNL or NMNL models, or population
preference distributions in the case of MXL models, are correlated positively or negatively. We
acknowledge that such conditional uncertainty intervals are less than ideal, yet we believe they
are preferable to providing no indication of uncertainty. In the future, we encourage researchers
to routinely calculate WTP measures for vehicle choice models and to provide accurate
confidence intervals for the WTP measures.
(C-5)
C-4
-------
APPENDIX D:
HISTOGRAMS OF UNTRIMMED CENTRAL WTP ESTIMATES BY ATTRIBUTE
Comfort
-5000.00
WTP for Automatic Transmission
0.00 500000
2015 dollars
WTP for All Wheel Drive
10000.01
10000.00 20000.00 30000.00 40000.00 50000.00 60000.00
2015 dollars
WTP for Air Conditioning
-20000.00 -10000.00
0.00
2015 dollars
WTP for Shoulder Room
10000.00 20000.0i 0.00 1.00
2.00 3.00
2015 dollars
4.00 5.00
Fuel Cost
WTP for Reduced Cents per Mile: Untrimmed
-1000000.00 -800000.00 -600000.00 -400000.00 -200000.00 0.00
2015 dollars
D-l
-------
Fuel Type
WTP for EVs
WTP for Hybrids: Untrimmed
-80000.00 -60000.00 -40000.00 -20000.00
2015 dollars
-200000.00 -150000.00 -100000.00 -50000.00
2015 dollars
WTP for Natural Gas
WTP for Flex Fuel
40000.00
-20000.00
2015 do ars
5000.00
2015 dollars
WTP for PHEV
10000.00
2015 dollars
D-2
-------
Performance
WTP for a 1 Second Decrease 0-60 mph
WTP for Horsepower
2000.00
2015 dollars
200.00
2015 dollars
Range
WTP for Range
-400.00
-200.00 0.00
2015 dollars
200.00
Size
WTP for Footprint Increase: Untrimmed
WTP for Luggage Space
400000.00
2015 dollars
D-3
-------
APPENDIX E:
UNTRIMMED DISTRIBUTIONS OF CENTRAL WTP ESTIMATES BY ATTRIBUTE
Comfort
WTP for Air Conditioning:
Range is +/-1 Standard Deviation
Preference Variation
$40,000
WTP for Air Conditioning:
Range is +/-1 Standard Error
Estimation Error
$30,000
$20,000
$10,000 d -
-$40,000
1 1
$0
Low Central High
WTP for Automatic Transmission:
Range is +/-1 Standard Error
Estimation Error
$15,000
WTP for Rear-wheel Drive:
Range is +/-1 Standard Deviation
Preference Variation
$100,000
$50,000
Low Central High
$0
Low Central High
WTP for an Additional Inch of Shoulder
Room:
Range is +/-1 Standard Error
Estimation Error
$10
Low Central High
E-l
-------
Fuel Costs
WTP for $0.01/Mile Reduction in Fuel Cost:
Range is +/-1 Standard Deviation
Preference Variation
$100,000
WTP for $0.01/Mile Reduction in Fuel Cost:
Range is +/-1 Standard Error
Estimation Error
$40,000
$20,000
High
-$50,000 j
-$100,000
Low "" Central High
-$20,000
E-2
-------
Fuel Type
WTP for Electric Vehicle:
Range is +/-1 Standard Deviation
Preference Variation
$100,000
$50,000
$0
-$50,000
-$100,000
-$150,000
High
WTP for Electric Vehicle:
Range is +/-1 Standard Error
Estimation Error
$50,000
$0
-$50,000
-$100,000
-$150,000
WTP for Hybrid Vehicle:
Range is +/-1 Standard Deviation
Preference Variation
$100,000
$50,000
$0
-$50,000
-$100,000
WTP for Hybrid Vehicle:
Range is +/-1 Standard Error
Estimation Error
$100,000
$0
-$100,000
-$200,000
-$300,000
WTP for Natural Gas Vehicle:
Range is +/-1 Standard Deviation
Preference Variation
$100,000
$50,000
$0
-$50,000
-$100,000
Central
-------
Performance
WTP for 1 Second Reduction in 0-60 mph
Acceleration Time:
Range is +/-1 Standard Deviation
Preference Variation
$10,000
WTP for 1 Second Reduction in 0-60 mph
Acceleration Time:
Range is +/-1 Standard Error
Estimation Error
$10,000
-$5,000 Li*r^ Central High
-$10,000
$0
L^W """"CeSralHigh
-$5,000
Horsepower: Range is +/-1 Standard
Deviation
Preference Variation
$100,000
Horsepower: Range is +/-1 Standard Error
Estimation Error
$15,000
$50,000
$10,000
55,000 ^
Low^^^^Central High
-$50,000
-$100,000
*c Low Central High
-^j;UUU
-$10,000
Range
WTP for One Addtional Mile of Range:
Range is +/-1 Standard Deviation
Preference Variation
$1,000
$500
WTP for One Additional Mile of Range:
Range is +/-1 Standard Error
Estimation Error
$1,000
$500
Central High
-$500
-C
—1
o o
in- o
LO
-co-
E-4
-------
Size
WTP for an Increase in Vehicle Footprint
(ft2):
Range: Range is +/- 1 Standard Deviation
Preference Variation
$1,000,000
$500,000
$0
Low
Central
High
$150
$100
$50
$0
-$50
WTP for an Increase in Vehilce Weight (lb):
Range is +/-1 Standard Error
Estimation Error
Low
Central
High
WTP for Luggage space:
Range is +/-1 Standard Error
Estimation Error
$2,000
-$1,000
E-5
-------
APPENDIX F:
AUTHOR FEEDBACK RECEIVED AND RESPONSE TO COMMENTS
There are a number of steps required to calculate WTP estimates from many of the papers
in the literature that do not report those values directly. In many cases, it is necessary to make
assumptions regarding the details of the calculations made by the authors where they are not
fully specified in the literature, which is common given limitations on journal paper length. In
this project, we made an effort to contact authors of each of the papers included in our sample
via email. We contacted the corresponding author where possible, but contact information on
some of the publications was out of date. In that case, we attempted to find updated contact
information for each corresponding author. In cases where we could not find their current contact
information, we reached out to the other authors of the paper for multi-authored studies. We
asked each of the authors contacted to review our WTP calculations for their publication(s)
(some authors were involved in multiple papers included within our main sample). There were
cases where neither the corresponding author nor coauthors responded to the initial or follow-up
requests for feedback. Table F-l summarizes the outcome of our request, comments received,
and actions that we took in response.
F-l
-------
Table F-l. Summary of Author Feedback Received and Response to Comments
r1
to
Pii per
Contacted
Responded
Provided
Comments
Comments
Response
Allcott and Wozny (2014)
Yes
Yes
Yes
No adjustments suggested
NA
Axsen, Mountain, and Jaccard (2009)
Yes
Yes
Yes
No adjustments suggested, provided more recent papers with WTP coefficients
NA
Beresteanu and Li (2011)
Yes
Yes
No
Indicated they would try to provide feedback, but we did not receive any
NA
Berry, Levinsohn, and Pakes (1995)
Yes
No
No
NA
NA
Brownstone and Train (1998)
Yes
Yes
Yes
Notes that this paper was primarily methodological rather than focused on
parameter estimation and recommends Brownstone, Bunch and Train as
preferred source of WTP estimates among their papers
NA
Brownstone, Bunch, and Train (2000)
Yes
Yes
Yes
No adjustments suggested; identified this paper as preferred source of estimates
among their papers
NA
Brownstone et al. (1996)
Yes
Yes
Yes
No adjustments suggested
NA
Busse, Knittel, and Zettelmeyer
(2013)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Dasgupta, Siddarth, and Silva-Risso
(2007)
Yes
No
No
NA
NA
Daziano (2013)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Dreyfus and Viscusi (1995)
Yes
Yes
Yes
Verified that the values used from their study were correct, but did not find the
WTP calculations sufficiently transparent to check
NA
Espey and Nair (2005)
Yes
Yes
Yes
No adjustments suggested
NA
Fan and Rubin (2010)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Feng, Fullerton, and Gan (2013)
Yes
Yes
No
Indicated they did not have time to complete the review
NA
Fifer and Bunn (2009)
No contact
information
identified
NA
No
NA
NA
Frischknecht, Whitefoot, and
Papalambros (2010)
Yes
Yes
Yes
Provided numerous comments suggesting modifications to our calculations as
well as suggesting that we use Monte Carlo simulations to generate distributions
around our WTP estimates
Adjusted calculations to the extent possible,
though some were not feasible due to lack of
data and/or project resources, e.g., conducting
Monte Carlo simulations for parameter data from
all papers
Gallagher and Muehlegger (2011)
Yes
Yes
Yes
Provided some caveats to the calculations, but agreed they were correct overall
NA
Goldberg (1995)
Yes
Yes
Yes
Suggested that we review units and calculations
Reviewed calculations and determined that they
were consistent with the descriptions in the 1995
paper so we made no adjustments
Gramlich (2008)
Yes
Yes
Yes
Raised questions about the sign of the WTP estimates we were calculating and
suggested we review units
Adjusted calculations
Greene (2001)
Yes
Yes
NA
NA, author is involved in this project
NA
Greene, Duleep, and McManus
(2004)
Yes
Yes
NA
NA, author is involved in this project
NA
Haafetal. (2014)
Yes
Yes
Yes
Provided a corrected supplement to their paper and suggested modifications to
our spreadsheet calculations
Adjusted calculations
Helveston et al. (2015)
Yes
Yes
Yes
Noted sign of a parameter was incorrect in their published paper so suggested
adjustment for that as well as updating our assumption regarding gasoline price
to align with their assumption; suggested more discussion of uncertainty and
heterogeneity and more information on our methods and interpretation
Adjusted calculations; added more discussion in
the report as suggested
(continued)
-------
Table F-l. Summary of Author Feedback Received and Response to Comments (continued)
r1
U>
Pii per
Contacted
Responded
Provided
Comments
Comments
Response
Hess, Train, and Polak (2006)
Yes
Yes
Yes
Asked about methods for calculating WTP, but no adjustments suggested
NA
Hess et al. (2011)
Yes
Yes
Yes
Asked about methods for calculating WTP, but no adjustments suggested
NA
Hidrue et al. (2011)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Kavalec (1999)
No contact
information
identified
NA
No
NA
NA
Klier and Linn (2012)
Yes
Yes
Yes
Suggested that we use delta method for deriving standard errors and asked us to
focus on their main instrumental variables (IV) estimate
Added discussion of the rationale and potential
implications of our using the ratio of random
variables to estimate WTP in both the main body
of the report and Appendix C; continued using
the range of results reported for consistency with
other papers and to show the importance of
specification
Lave and Train (1979)
Yes
Yes
Yes
Provided suggested adjustments to our spreadsheet calculations
Adjusted calculations
Liu, Tremblay, and Cirillo (2014)
Yes
Yes
Yes
Indicated that WTP should be adjusted because of the income scaling used in
their model
Adjusted calculations
Liu (2014)
No contact
information
identified
NA
No
NA
NA
McFadden and Train (2000)
Yes
Yes
Yes
Provided suggested adjustments to our spreadsheet calculations
Adjusted calculations
McCarthy (1996)
Yes
Yes
Yes
Requested clarification of spreadsheet calculations; no adjustments suggested
Provided clarification; no changes made to
spreadsheets
McCarthy and Tay (1998)
Yes
Yes
Yes
Requested clarification of spreadsheet calculations; no adjustments suggested
Provided clarification; no changes made to
spreadsheet
McManus (2007)
Yes
Yes
Yes
No adjustments suggested
NA
Musti and Kockelman (2011)
Yes
Yes
No
Requested additional clarification regarding the review request
NA
Nixon and Saphores (2011)
Yes
Yes
No
Indicated they would try to provide feedback, but we did not receive any
NA
Parsons et al. (2014)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Petrin (2002)
Yes
Yes
No
Requested additional clarification regarding the review request, which was
provided but we did not receive review comments
NA
Sallee, West, and Fan (2016)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Segal (1995)
No contact
information
identified
NA
No
NA
NA
Sexton and Sexton (2014)
Yes
No
No
NA
NA
Shiau, Michalek, and Hendrickson
(2009)
Yes
Yes
Yes
Suggested more discussion of uncertainty and heterogeneity, our use of the ratio
of random variables to estimate WTP, and the interpretation of these values
Added discussion in the main body of the report
and Appendix B of the rationale and potential
implications of our using the ratio of random
variables to estimate WTP
Skerlos and Raichur (2013)
Yes
No
No
NA
NA
Tanaka et al. (2014)
Yes
No
No
NA
NA
Tompkins et al. (1998)
Yes
Yes
Yes
Cautioned against using WTP values from SP models in general
NA
Train and Sonnier (1995)
Yes
Yes
Yes
Suggested dropping this paper from the analysis because Train and Weeks (2005)
reports the same information but with the authors' calculation of WTP
NA
(continued)
-------
Table F-l. Summary of Author Feedback Received and Response to Comments (continued)
Pii per
Contacted
Responded
Provided
Comments
Comments
Response
Train and Weeks (2005)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Train and Winston (2007)
Yes
Yes
Yes
Provided suggested modifications to our spreadsheet calculations
Adjusted calculations
Walls (1996)
Yes
No
No
NA
NA
Whitefoot, Fowlie, and Skerlos
(2011)
Yes
Yes
Yes
Provided several adjustments and expressed concern regarding the endogeneity
of the attributes and the effect on WTP
Adjusted calculations
Zhang, Gensler, and Garcia (2011)
Yes
Yes
No
Requested additional clarification regarding the review request
NA
~n
------- |