Are consumers willing to pay to let cars drive for
them? Analyzing response to autonomous vehicles
Ricardo A. Daziano
∗
, Mauricio Sarrias
†
and Benjamin Leard
‡
2016
Abstract
Autonomous vehicles use sensing and communication technologies to navigate
safely and efficiently with little or no input from the driver. These driverless
technologies will create an unprecedented revolution in how people move, and
policymakers will need appropriate tools to plan for and analyze the large impacts
of novel navigation systems. In this paper we derive semiparametric estimates of
the willingness to pay for automation. We use data from a nationwide online panel
of 1,260 individuals who answered a vehicle-purchase discrete choice experiment
focused on energy efficiency and autonomous features. Several models were estimated
with the choice microdata, including a conditional logit with deterministic consumer
heterogeneity, a parametric random parameter logit, and a semiparametric random
parameter logit. We draw three key results from our analysis. First, we find
that the average household is willing to pay a significant amount for automation:
about $3,500 for partial automation and $4,900 for full automation. Second, we
estimate substantial heterogeneity in preferences for automation, where a significant
share of the sample is willing to pay above $10,000 for full automation technology
while many are not willing to pay any positive amount for the technology. Third,
our semiparametric random parameter logit estimates suggest that the demand for
automation is split approximately evenly between high, modest and no demand,
highlighting the importance of modeling flexible preferences for emerging vehicle
technology.
JEL classification: C25, D12, Q42
Key words: willingness to pay, autonomous vehicle technology, discrete choice models,
semiparametric heterogeneity
∗
School of Civil and Environmental Engineering, Cornell University, Ithaca, NY 14853; Email:
daziano@cornell.edu
†
Department of Economics, Universidad Catolica del Norte, Chile
‡
Resources for the Future, Washington D.C.
1
© 2017. This manuscript version is made available under the Elsevier user license
http://www.elsevier.com/open-access/userlicense/1.0/
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
1 Introduction
Personal mobility is about to experience an unprecedented revolution motivated
by technological change in the automotive industry (National Highway Traffic
Safety Administration, 2013; Fagnant and Kockelman, 2014). The introduction of
automated vehicles –in which at least some (and potentially all) control functions
occur without direct input from the driver– will completely change how people move.
The adoption of automated navigation systems has the potential to dramatically
reduce traffic congestion and accidents, while creating substantial improvements in
the overall trip experience as well as providing enhanced accessibility opportunities
to people with reduced mobility (Fagnant and Kockelman, 2015).
Automated vehicles use sensing and communication technologies to navigate
safely and efficiently with little or no human input. Automated navigation technology
comprises any combination of (1) self-driving navigation systems informed by on-
board sensors (autonomous vehicles) vehicle-to-vehicle (V2V) and (2) vehicle-to-
infrastructure (V2I) communication systems that inform navigation and collision
avoidance applications (connected vehicles). The National Highway Traffic Safety
Administration (NHTSA) has suggested five levels of automated navigation: level
0 (no automation), where the driver is in complete control of safety-critical
functions; level 1 (function-specific automation), where the driver cedes limited
control of certain functions to the vehicle especially in crash-imminent situations
(adaptive cruise control, electronic stability control ESC, automatic braking); level
2 (combined-function automation), which enables hands-off-wheel and foot-off-pedal
operations, but the driver is expected to be available at all times to resume control of
the vehicle (adaptive cruise control and lane centering); level 3 (limited self-driving
or conditional automation), where the vehicle potentially controls all safety functions
under certain traffic and environmental conditions, but some conditions require
transition to driver control; and level 4 (driverless or full self-driving automation),
where the vehicle controls all safety functions and monitors conditions for the whole
trip.
1
1
A six level categorization is proposed by the Society of Automotive Engineers, which further
distinguishes levels within NHTSA level 4.
2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Imminent commercialization of automated cars is best exemplified by the recent
announcement (October 2016) that all new Tesla vehicles will have full self-driving
hardware.
2
Several semi-autonomous features are already available in the automotive
market, mostly in the form of in-vehicle crash avoidance upgrades with preventive
warnings or limited automated control of safety functions, such as braking when
danger is detected. Self-parking assist systems are another example of a more
advanced upgrade that is currently available in select makes and models. These
entry-level automation packages are possible as a result of vehicles being equipped
with radar, cameras, and other sensors. Even though technology is still evolving, full
automation is possible with the current stage of development. The Google car and
its more than 2 million miles of driverless driving is the most publicized effort.
3
The literature on vehicle-to-vehicle, vehicle-to-infrastructure, and control systems
for safe navigation is extensive. Regulation, insurance, and liability are other areas
where there is strong debate. However, little attention has been devoted to the
analysis of automated vehicles as marketable products. Consumer acceptance is
critical to forecast adoption rates, especially if one considers that there may be
strong barriers to entry (potential high costs, concerns that technology may fail).
Our work contributes to two strands of literature on the demand for new
technology. The first area is the recent development in understanding the demand,
penetration, and policy implications of autonomous vehicle technology. Several
recent studies attempt to understand how consumer preferences for attributes such
as safety, travel time, and performance shape the demand for driverless cars.
Kyriakidis et al. (2015) conducted an international public opinion questionnaire of
5,000 respondents from 109 countries. Responses were diverse: 22 percent of the
respondents did not want to pay any additional price for a fully automated navigation
system, whereas 5 percent indicated they would be willing to pay more than $30,000.
Payre et al. (2014) conducted a similar survey of 421 French drivers with questions
eliciting the acceptance of fully automated driving. Among those surveyed, 68.1
percent accepted fully automated driving unconditionally, with higher acceptance
2
Source: https://www.tesla.com/blog/all-tesla-cars-being-produced-now-have-
full-self-driving-hardware
3
Source: https://www.google.com/selfdrivingcar/faq/
3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
conditional on the type of driving, including usage of highway driving, in the presence
of traffic congestion, and for automated parking. Similar results were obtained in
a survey of Berkeley, California, residents conducted by Howard and Dai (2013).
Individuals in this survey were most attracted to the potential safety, parking, and
multi-tasking benefits. Schoettle and Sivak (2014) conducted a much larger and
more internationally based survey of residents from China, India, Japan, the United
States, the United Kingdom, and Australia. The authors found that respondents
expressed high levels of concern about riding in self-driving vehicles, with the most
pressing issues involving those related to equipment or system failure. While most
expressed a desire to own an autonomous vehicle, many respondents stated that they
were unwilling to pay extra for the technology.
A paper related to our own is that by Bansal et al. (2016), which estimates
willingness to pay for different levels of automation. They find that for their sample
of 347 residents of Austin, Texas, willingness to pay (WTP) for full automation
is $7,253, which is substantially higher than our own estimate. The authors also
estimate WTP for partial automation of $3,300, which is similar to our estimate.
Our demand estimates contribute to the assessment of the social costs and benefits
of autonomous vehicles. Fagnant and Kockelman (2015) estimate the external net
benefits from autonomous vehicle penetration. They find that the social net benefits
including crash savings, travel time reduction from less congestion, fuel efficiency
savings, and parking benefits total between $2,000 and $4,000 per vehicle. These
estimates, however, greatly depend on how the presence of autonomous vehicles will
impact both vehicle ownership and utilization. For example if autonomous vehicles
make owning a vehicle more desirable, then the stock and use of vehicles may increase,
reducing the external net benefits.
We designed a web-based survey with a discrete choice experiment to determine
early-market empirical estimates of the structural parameters that characterize
current preferences for autonomous and semi-autonomous electric vehicles. The
discrete choice experiment contained as experimental attributes three levels of
automation: no automation, some or partial automation (“automated crash
avoidance”), and full automation (“Google car”). Automation was allowed for
alternative powertrains (hybrid electric, plug-in hybrid and full battery electric).
4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Based on the results from this experiment, we estimate WTP for automation. Our
estimates of WTP for privately owned autonomous vehicles take a first step to
understanding the demand for this technology, which is critical for understanding
how aggregate demand for vehicles and vehicle miles traveled will respond to the
technology over time.
4
In addition to the discrete choice experiment of vehicle purchase, the survey
also contained an experiment to elucidate the subjective discount rate of potential
vehicle buyers. Expanding on the work of Newell and Siikam¨aki (2013), we used
the individual-level experimental discount rate to determine the present value of fuel
costs for each alternative.
To derive flexible estimates of the heterogeneity distribution of the willingness to
pay for automation, we implemented the maximum simulated likelihood estimator
of a logit-based model with discrete continuous heterogeneity distributions, in
which the parameters (mean and standard deviation) of continuous heterogeneity
distributions have associated discrete, unknown probabilities. The approach adopted
to unobserved preference heterogeneity in this paper thus takes into consideration
a mixed-mixed logit model (Bujosa et al., 2010; Greene and Hensher, 2013; Keane
and Wasi, 2013), where the random willingness-to-pay parameters are distributed
according to a Gaussian mixture. The weights of the Gaussian mixture can
include individual-specific covariates that allow us to identify clusters with differing
willingness to pay for automation. The estimator was implemented with analytical
expressions of the score for computation efficiency.
Methodologically, we highlight the importance of allowing for flexible distribu-
tions of preferences for vehicle attributes such as automation by comparing estimates
from a standard mixed logit specification with a more flexible mixed-mixed logit spec-
ification. We find richer heterogeneity estimates with the more flexible specification,
4
We do not explore demand for autonomous commercial vehicles or for autonomous public
transportation. Initial work in this area includes a study by Greenblatt and Saxena (2015)
which simulates the greenhouse gas impact of autonomous vehicle taxis and finds that they can
dramatically reduce greenhouse gas emissions relative to conventional taxis. A promising area of
future research involves incorporating our survey and econometric methods for eliciting WTP to
determine how households tradeoff cost savings, travel time, safety, and other desirable attributes
with alternative travel modes with and without a human driver.
5