Transportation Research Part D xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Transportation Research Part D

journal homepage: www.elsevier.com/locate/trd

Electric vehicles, air pollution, and the motorcycle city: A stated

preference survey of consumers’ willingness to adopt electric
motorcycles in Solo, Indonesia
Erick Guerra
Department of City and Regional Planning, University of Pennsylvania, School of Design, 127 Meyerson Hall, 210 S. 34th St, Philadelphia, PA 19104,
United States

AR TI CLE I NF O AB S T R A CT

Keywords: This paper presents the results of a choice experiment to evaluate the extent to which electric
Motorcycles motorcycles are a potential replacement for gasoline-powered motorcycles in Solo, Indonesia.
Electric vehicles Survey respondents faced five choice scenarios where they selected between a conventional
Stated preference motorcycle, an electric motorcycle, and no motorcycle based on price, speed, range, and charge
Vehicle ownership
time. Approximately 1200 individuals completed all five choice scenarios and fully responded to
Mixed logit
socioeconomic and preference questions. As in much of Southeast Asia, motorcycles dominate the
transportation system and provide inexpensive, relatively safe, and convenient point-to-point
travel. However, motorcycles also produce substantial, harmful local emissions. To estimate the
probability of selecting different motorcycle types, the survey data are fit with a mixed logit
model with random coefficients. This specification allows correlation across choices over time,
flexible substitution across the choice alternatives, and variation in consumer preferences. The
results of the survey and analysis indicate that there is almost certainly a market for electric
motorcycles but their price and performance will have to be competitive with low-cost, gas-
powered ones. Speed, range, charge time, and price all mattered substantially with respondents
willing to pay a 7–13% premium for motorcycles with 10 km longer range, 10 km/h faster speed,
or an hour shorter charge time. Charge time was particularly important, suggesting that im-
provements in battery charging technology and charging infrastructure could substantially im-
pact consumers’ willingness to adopt electric motorcycles. Younger non-smoking respondents
with concerns about the environment and favorable views of e-bikes were most likely to choose
electric motorcycles.

1. Introduction

Residents of a number of cities, particularly in Southeast Asia, have adopted the motorcycle as the primary means of urban
transportation. Relatively high but flatly distributed population and job densities, narrow streets, and a uniformly distributed road
network make the motorcycle a convenient choice for point-to-point travel in many cities. Low-cost Chinese motorcycles, credit, and
inexpensive fuel make it an affordable choice as well. Motorcycles not only use space efficiently due to their size and maneuverability
(Yeung et al., 2015; Fan, 1990; Adnan, 2014), but their efficiency likely increases with the proportion of traffic comprised of mo-
torcycles (Rongviriyapanich and Suppattrakul, 2005). In motorcycle cities like Hanoi and Solo, narrow arterial roads often carry more
people per lane than superhighways in American ones (Cao and Sano (2012) and author’s estimates from traffic counts in Solo). Even

E-mail address: erickg@upenn.edu.

http://dx.doi.org/10.1016/j.trd.2017.07.027
Received 20 December 2016; Received in revised form 9 June 2017; Accepted 26 July 2017
1361-9209/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Guerra, E., Transportation Research Part D (2017), http://dx.doi.org/10.1016/j.trd.2017.07.027
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

in highly congested conditions, streets filled with motorcycles tend to maintain vehicle flow instead of coming to the bumper-to-
bumper halt that is familiar to most car drivers and bus passengers. Finally, despite the perceptions of motorcycling as a high-risk
activity, low speeds and a high proportion of motorcyclists and bicyclists on the road in motorcycle cities make motorcycling sub-
stantially less dangerous than in the United States and Europe.1
Despite these advantages, a heavy reliance on motorcycles has disadvantages. Most notably, inexpensive motorcycles produce
high levels of concentrated local pollution, which reduces life expectancy and causes harmful respiratory diseases, particularly for the
young and elderly. The transportation sector contributes substantially to poor air quality, which has been associated with billions of
years of lost life in Northern China and India (Chen et al., 2013; Greenstone et al., 2015). Even 4-stroke engines, which produce
substantially lower emissions than 2-stroke ones, degrade local air quality. Four-stroke-engine motorcycles comprise 95–99% of
Hanoi and Ho Chi Minh City’s motorcycle fleets (Oanh et al., 2012; Hieu et al., 2013), but both cities have elevated levels of local
pollution. Hieu et al. (2013) estimate that the number of premature deaths attributable to pollution from motorcycles is substantially
higher than deaths attributable to traffic collisions in Hanoi and Hai Phong.
Shifting from gas-powered motorcycles to electric motorcycles would improve local air quality and reduce one of the primary
harmful impacts associated with motorcycles (Cherry et al., 2009). Although the effect on global pollution and fossil fuel con-
sumption would depend on power generation and vehicle substitution (Ma et al., 2012; Cherry et al., 2009), a shift to electric
motorcycles could also support efforts to move away from fossil fuels. For example, the Indonesian Ministry of Energy and Mineral
Resources and Energy Conservation (2012) set a 2025 target to reduce the share of oil-based transportation energy by 20%, but the
transportation sector, which relies on gasoline, consumes and is predicted to continue to consume roughly 40% of all energy
(CAREPI, 2009). In many countries, reliance on gasoline for personal travel makes it difficult to reduce or remove expensive and
economically distorting fuel subsidies and switch to less carbon intensive fuels. Indonesia’s recently elected President Joko Widodo
ended longstanding public fuel subsidies in 2015, more than a decade after Indonesia became a net oil importer.
In order for electric motorcycles to reduce local emissions and potentially facilitate a transition away from fossil fuels, consumers
would need to accept them. This paper presents the results of a 2015 stated choice experiment to estimate consumers’ willingness to
adopt electric motorcycles in Solo, Indonesia, as a function of price and electric motorcycle range, speed, and charge times.
Researchers have used stated preference surveys and qualitative choice models to analyze markets for alternative-fuel cars for the
past four decades (Beggs and Cardell, 1980; Brownstone et al., 2000; Brownstone and Train, 1998; Calfee, 1985; Daziano, 2013).
Consumers’ willingness to adopt electric cars varies by price, range, speed, charging infrastructure, and personal preferences. More
recently, a body of work has begun to examine the potential for e-bikes and e-scooters, particularly in China (Cherry and Cervero,
2007; Cherry et al., 2009; Fishman and Cherry, 2015). These e-bikes have substantially lower speeds and pick-up than a car or
motorcycle and tend to be a substitute for transit and traditional bicycles, not cars or motorcycles. Although vendors sell e-bikes
throughout cities in Indonesia and Vietnam, market penetration is low. Two studies have examined consumer preferences for higher
powered electric motorcycles in Hanoi and Taipei (Jones et al., 2013; Chiu and Tzeng, 1999).
This study contributes to the literature on consumer preferences for electric motorcycles in five primary ways. First, it adds an
additional and substantially different case location to the literature. Solo is smaller and has higher motorcycle mode share than cities
in existing studies in Vietnam, China, and Taiwan. This size and relative motorcycle dependence make it a better case for small-to-
medium-sized fast-growing motorcycle cities with limited transit options and shorter average trip distances. Second, the survey offers
an electric alternative with a realistic home- and work-based charging infrastructure. This contrasts with most of the scenarios
presented in the two existing studies of consumer preference for higher powered electric motorcycles (Jones et al., 2013; Chiu and
Tzeng, 1999). While rapid charging and battery swaps likely make electric vehicles more attractive, these recharging systems require
expensive investments in infrastructure that may take years to materialize. Third, the survey design presents respondents with
monthly credit payments instead of a single up-front purchase price. This payment mechanism is the primary way to purchase new
motorcycles in Indonesia and many other parts of Southeast Asia. Fourth, the survey includes more questions on consumers’ pre-
ference for electric motorcycles and concerns about the local and global environment. These provide for estimates of how much of a
price premium environmentally conscious respondents are willing to put on electric motorcycles. Finally, with over 1200 respondents
this study triples the sample size of existing studies.

2. Solo, motorcycle city

Solo (also known as Surakarta) is home to 500,000–600,000 inhabitants and is at the center of a metropolitan area of 3 million in
the populous province of Central Java. Long a walking and biking city, Solo has transformed rapidly into a city dominated by
motorcycles and motorcyclists with nearly one motorcycle per man, woman, and child. Between 2009 and 2013, the motorcycle fleet
more than doubled from 208,000 to 424,000 (Fig. 1). Motorcycles dominate traffic and account for around two-thirds of the peak-
hour, peak-direction trips on major arterials. In surveys for this choice experiment, 81% of respondents indicated that the motorcycle
was their primary mode of transportation. This section provides a brief background on why motorcycles have become so popular in
Solo and why it is a good location to investigate consumers’ willingness to adopt electric motorcycles.

1
Countries with high rates of motorcycle use have similar fatality rates to peer countries with low rates of motorcycle use. Moreover, the difference in traffic
fatalities per 100,000 residents between a country like Indonesia (15) or Vietnam (25) and the United States (11) is proportionally much smaller than the difference
between the United States and Sweden (3) or Great Britain (3) (World Health Organization, 2015). Despite high rates of motorcycle use in Southeast Asia, 34% of
traffic fatalities were on a motorized 2- or 3-wheeler, compared to 20% in the Americas (World Health Organization, 2015). In the United States, the number of
fatalities per person trip is 60 times higher by motorcycle than by car (Beck et al., 2007).

2
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

450000

400000

Number of registered motorcycles

350000

300000

250000

200000

150000

100000

50000

0
2008 2009 2010 2011 2012 2013

Fig. 1. Number of registered motorcycles in Solo city (2012 data unavailable) (Dinas Pendapatan Dan Pengelolaan Aset Daerah Provinsi Jawa Tengah, 2015).

Motorcycles provide convenient, inexpensive, point-to-point travel in Solo. Purchasing a motorcycle requires a $40 (500,000
rupiah) down-payment, and $40 monthly payments for three years. A liter of gasoline, which currently costs around fifty cents (6700
rupiah) supports 30–40 km of travel on an average motorcycle—enough for a day’s travel. This makes the marginal cost of motorcycle
travel less expensive than transit, which costs between 2000 and 5000 rupiah per segment and often requires several paid transfers
(Kota Kita, 2015; Rifai et al., 2011). Motorcycles also work well with Solo’s existing urban form. Fairly uniform population densities,
dispersed employment centers, and a gridded road network of two-to-four-lane two-way arterial roads connected to narrow re-
sidential side-streets create a flat distribution of traffic patterns well-suited to private forms of transportation. Narrow residential
streets are much better suited to pedestrians, bicycles, and motorcycles, than to cars or buses. Even during peak time periods on major
corridors less than 10% of travelers use public transit, and 80% of these are on large regional buses, not local ones. After 5 pm, nearly
all transit service stops (Kota Kita, 2015).
Several other factors beyond high motorcycle use make Solo an attractive place to examine the potential for electric motorcycles.
Short daily trips, good access to electricity at home and work, and relatively slow traffic speeds make lower cost electric motorcycles
potentially more attractive. While electric motorcycles can go as fast and accelerate as quickly as the highest performance gas-
powered ones, the price of electric motorcycles rises rapidly with increased speed and range. For example, an electric-powered Zero
motorcycle with a top speed of 140 km per hour and range of 320 km costs over $10,000 (Zero Motorcycles, 2015). Lower cost, lower
performance e-bikes and electric scooters exist in Solo but have had very little success in winning over consumers.
Electric motorcycles are also likely to become relatively more attractive as fuel prices rise. President Joko Widodo removed
longstanding fuel subsidies at the beginning of 2015, but a weak global economy and the glut in global oil production due to fracking
and tar-sands extraction has kept prices low. As prices rise, Indonesian consumers will no longer be insulated from global prices and
alternative fuel vehicles may begin to look more attractive.
Solo is also well-known for progressive urban policies (Bunnell et al., 2013; The Economist, 2009) and Dimitriou (2006) identified
Solo as institutionally well-suited to developing and implementing policies to encourage sustainable transportation. Nevertheless,
efforts to improve public transportation in Solo and other similarly sized Indonesian cities have fared poorly. With support from the
central Ministry of Transportation, the local government has fully developed two of nine planned bus rapid transit routes in an effort
to improve public transportation and reduce emissions from older bus fleets. Ridership has been extremely low. Of 19 BRT systems
that opened worldwide in 2010 and 2011, Solo and two other Indonesian cities had an average of only 39 weekday riders per
kilometer of the system, compared to over 7000 on average for the remaining 16 systems (Hidalgo and Gutiérrez, 2013). If im-
provements to transit are unlikely to attract residents off of their motorcycles, it is all the more important to work to reduce the harm
caused by motorcycles.

3. Survey design

3.1. Choice and attributes

Survey respondents faced five choice scenarios where they selected between a conventional motorcycle, an electric motorcycle,
and no motorcycle based on price, speed, range, and charge time. Table 1 summarizes the levels of the six randomly varying features
in the choice experiment. With the exception of fuel price, which varies periodically, the levels for conventional gas-powered mo-
torcycles remained fixed. In Solo as in much of Indonesia, consumers buy motorcycles on credit. The most common purchasing
mechanism requires three years of monthly payments of 500,000 rupiah (roughly $40) and a down payment of 500,000 rupiah.
Consumers have an intimate knowledge of the features of gas-powered motorcycle, while electric motorcycles are less well under-
stood.
Focus groups with motorcycle-owning staff from a local NGO helped determine the relevant features and their levels. For ex-
ample, because participants indicated that vehicle acceleration was not an important consideration for motorcycle purchases, I

3
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

Table 1
Summary of choice sets and feature values.

Gas Motorcycle Electric Motorcycle

Cash down (thousands) 500 400, 500, 600

Monthly payment (3 years in thousands) 500 400, 500, 600
Fuel price (thousands of rupiah per liter and liter equivalent) 7, 8, 9 4, 5, 6, 7, 8
Maximum range on single charge (km) NA 40, 60, 80, 100
How long to charge (hours) NA 2, 3, 4, 5
Max speed (km/h) 100 60, 70, 80, 90, 100

dropped this feature from the survey. Where possible, I set the electric vehicle feature levels relative to a typical gas-powered
motorcycle. Instead of adjusting immediately around these levels, however, I constructed the survey to reflect electric motorcycles’
slower speeds and lower fuel costs. Charge time and range, by contrast, do not have an equivalent gas-powered attribute and were set
to reflect existing battery technology. The minimum range offered is more than sufficient for a typical day of travel. Focus group
participants expressed particular concerns about maximum range and where and how they could charge an electric vehicle.
For all non-varying features, such as acceleration or vehicle body type, surveyors told respondents that the features were
equivalent for the two motorcycle alternatives under consideration. Surveyors also asked respondents to consider the purchase as
occurring at some point in the next several years at a time when an existing personal motorcycle would need to be replaced. After
answering the five choice scenarios, respondents provided information on household income, age, gender, travel habits, current
motorcycle ownership, residential location, smoking habits, environmental concerns, and experiences with existing models of electric
scooters. Smoking habits reflect respondents’ tolerance for low local air quality, but are also correlated with other attributes like
gender and income.

3.2. Survey design

Unlike previous e-bike and electric scooter choice experiments (Chiu and Tzeng, 1999; Cherry and Cervero, 2007; Jones et al.,
2013), this survey uses randomly varying choice attribute levels to create choice scenarios instead of fractional factorial design. This
approach maintains uncorrelated attribute levels, but allows for the estimation of a full range of interaction effects, non-linear
treatment effects, and the inclusion of dominant scenarios—i.e., where the features of electric or conventional motorcycle are
substantially better than the alternative across most or all dimensions. These dominant scenarios are particularly important when
little is known about preferences for an unknown technology that may not be well-received by consumers.
The random variation in feature levels presents two primary problems. First, the design requires a larger sample to estimate main
effects, since these are not given special priority as in a fractional factorial design. Second, the survey collection requires computer
assistance and would be costly and difficult to implement with pen and paper surveys—though pen and paper surveys may have other
advantages in the developing world (Campbell et al., 2014).

3.3. Survey process

A total of 1307 people responded to the survey from August 4 to September 1, 2015. Of these, 1208 completed all five choice
scenarios and fully responded to other socioeconomic and preference questions. Local surveyors consisted of members of Kota Kita, a
local non-governmental organization focused on urban planning issues that has experience conducting interviews using tablets, cell
phones, and other methods (Kota Kita, 2015, 2012). After an afternoon of training in English and Bahasa, 10 surveyors conducted the
surveys in Bahasa over the course of four weeks.
Surveys were designed and collected using the Flocktracker application on Android tablets (MIT Department of Urban Studies,
and Planning, Mobility Futures Collaborative, and Urban Launchpad/MX, 2016). This application has been used to map informal
transit routes and conduct onboard passenger surveys in Dhaka, Mexico City, and Solo (Zegras et al., 2015b,a; Kota Kita, 2015).
In addition to providing a platform that could handle random variation in feature levels and facilitate the collection of a large
sample, the Flocktracker application provides geolocations, start and end times, and surveyor information for all survey responses.
This allowed for the daily checking of data and monitoring of adherence to the survey protocol. As a result of this daily checking, two
women surveyors were hired for an additional two weeks. This increased the survey sample and also the proportion of respondents
who were women, since male surveyors tended to under-survey from women, likely for cultural reasons. Again likely for cultural
reasons, response rates were high at around 95%. Jones et al. (2013) had similar response rates from Hanoi. Respondents also had the
option to provide unlinked contact information to be entered into a lottery for a small prize (T-shirts with the Kota Kita logo).
Surveyors targeted 20 pre-specified locations where large groups of people of varying socioeconomic status congregate in Solo.
These included five markets, a hospital, five universities, three commercial corridors, three mosques, two regional rail stations, and
the main recreational stadium. Fig. 2 maps the location of each survey. At the end of each day, surveyors reported back on any
qualitative exchanges that they had with survey respondents. This gave additional insight into gender differences and the importance
of charging and range for certain respondents.
Although the survey method allowed for the rapid collection of 1300 surveys, the sample is not representative of the population at
large. Specifically, it includes more men, younger respondents, and fewer from the city’s poorest households. Nevertheless, these are

4
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

Fig. 2. Location of field surveys.

the types of respondents most likely to be making choices about motorcycle purchases in the future. A number of women respondents
indicated that male relatives make decisions about motorcycle purchases for them. The larger sample also likely increases the
precision of individual parameter estimates but warrants extra caution when making inferences about Solo’s population from the
model results.

4. Data and modeling framework

4.1. Data sample

Table 2 presents the summary statistics for the 1208 respondents who completed all five choice experiments and fully answered
all demographic and behavioral questions. Comparing the sample to Solo’s resident population, the mean age for people 16 and over
is roughly correct, but the data oversamples from 19-to-25 year olds and substantially under samples from those 65 and older. It also
includes nearly twelve percentage points fewer women. This is likely a result of the sampling method and the men surveyors’ higher
propensity to survey other men. Seventy-two percent of households earn between 500,000 and 2,500,000 rupiah per month. The
Indonesian Bureau of Statistics groups households into five categories of prosperity based on 21 different social indicators. In Solo,
10% of households fall into the most vulnerable group; 15%, into the most prosperous group (Bappeda Kota Surakarta, 2014). Most
survey respondents fall in the second and third income groups and have one to three motorcycles in the household (Fig. 3). Across
income groups, the typical household has two motorcycles with 95% of respondents indicating access to a personal motorcycle.
The majority of the sample indicated that they do not smoke and think a lot about the quality of the natural environment. Both
variables suggest that respondents might prefer low-polluting vehicles such as electric vehicles. Despite low e-bike market pene-
tration, approximately a fifth of the sample has ridden an e-bike and more than a quarter thinks that e-bikes are high quality. Another
fifth indicate that they think e-bikes are low quality. For those who have tried an e-bike, 31% think the quality is low and 48% think
the quality is high. Quality, however, is not defined and the quality variables are best interpreted as indicators of whether re-
spondents have relatively more or less favorable attitudes toward electric vehicles in general.

4.2. Mixed logit specification with random coefficients

To estimate how the probability of selecting different motorcycle types varies with vehicle types and across different consumers,

5
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

Table 2
Descriptive sample statistics (N = 1208).

Statistic Mean St. Dev. Min Max

Woman 0.39 0.49 0 1

Age 32.0 12.6 16 80
Household Motorcycles 2.11 1.07 0 8
Has own motorcycle 0.95 0.21 0 1
Motorcycle: principal mode 0.81 0.39 0 1

Monthly income (100,000’s of Rupiah)

< 500 0.07 0.25 0 1
500–1000 0.31 0.46 0 1
1000–2500 0.41 0.49 0 1
2500–5000 0.17 0.38 0 1
> 5000 0.04 0.19 0 1

Thinks about the natural environment

A lot 0.52 0.50 0 1
Sometimes 0.32 0.47 0 1
A little 0.10 0.30 0 1
Rarely 0.06 0.23 0 1

Smoking habits
Smokes 0.33 0.47 0 1
Smokes sometimes 0.07 0.26 0 1

E-bike exposure and perception

Has used an e-bike 0.22 0.41 0 1
Thinks quality good 0.27 0.44 0 1
Thinks quality poor 0.19 0.39 0 1

Fig. 3. Bar plot of household motorcycles by income group (the second and fourth income groups fall between reported values in the first, third, and fifth groups).

the data are fit with a mixed logit model with random coefficients. Train (2009, chap. 6) describes the estimation procedure and its
primary advantages. Mixed logit models can approximate any random utility model and overcome the three primary limitations of a
multinomial logit by allowing correlation across choices over time, flexible substitution across the choice alternatives, and variation
in consumer preferences. All three are necessary for this survey design: the five choices of each individual are correlated as in any
panel; the electric and gas-powered motorcycles choices are correlated and more likely substitutes for each other than the no mo-
torcycle alternative; and there is almost certainly substantial variation in how much consumers care about electric motorcycles’
range, speed, and charge times.
The final models presented rely on 4000 quasi-random Halton draws to estimate variation in preferences, as indicated by the
standard deviation of the estimated coefficients. The larger the standard deviation coefficient, the greater the variation in individuals’
preferences. A single coefficient draw is used for each individual, which implies stability in preferences across the five survey
responses. The final models draw from normal distributions to establish preference variations. I also tested triangular distributions,
truncated normal, and log-normal distributions for features related to motorcycle performance. Price coefficients are held at the mean
(i.e., with fixed preferences). This facilitates estimation of willingness-to-pay for different motorcycle features and is a common
practice in random parameter mixed logit models. In the final models, I drop statistically insignificant coefficient estimates. A further
test of model fits using random 50% data samples indicated that a number of the statistically significant random coefficients were
overfitting the data. In the final presented models, only random coefficients that are stable across the 50% data samples are included

6
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

(Appendix A). While the motorcycle feature estimates are all within a standard deviation of the original estimates, the socioeconomic
and preference attributes are less stable and dropped from the final specifications.
I also tested interaction terms for motorcycle features, such as range and charge time, but these interactions complicated in-
terpretations without statistically improving model fits according to likelihood ratio tests. It is unclear whether the sample is too
small to capture these interaction effects or respondents did not consider them. Income enters the estimation as a single dummy
variable for electric- and gas-powered motorcycles. A series of full- and reduced-model likelihood ratio tests with a range of income
factors indicated that these single dummy variables fit the data best.

5. Results and discussion

Across the choice experiments, respondents selected gas-powered motorcycles 4230 times (70%), electric motorcycles 1564 times
(26%), and no motorcycle 246 times (4%). Table 3 summarizes the final model results with random coefficient estimates presented
below the other coefficients. Model 1 includes only information about the price and performance of the motorcycles under con-
sideration. Model 2 adds further information about respondents’ sociodemographic attributes, such as age and income. Model 3
includes preference variables, such as whether a respondent already has a motorcycle or thinks a lot about the environment.
Parameter estimates for motorcycle features are generally stable across specifications, while the inclusion of preference variables
tends to weaken the relationship between socioeconomic attributes and motorcycle choice. Neither demographic nor preference
variables increase the model fits substantially. However, these variables are perfectly correlated with each individual within the panel
and thus partially accounted for in Model 1.
Across models, the probability of selecting an electric motorcycle increases with higher speeds, longer ranges, faster charging
times, lower fuel costs, and lower monthly payments.2 This is intuitive and conforms to existing findings about the relationship
between price, quality, and preference for alternative fuel vehicles (Beggs and Cardell, 1980; Brownstone et al., 2000; Brownstone
and Train, 1998; Calfee, 1985; Daziano, 2013; Jones et al., 2013; Cherry and Cervero, 2007). There is also substantial variation in
respondents’ preferences for speed, range, and charge times. In each case, the standard deviation of the parameter estimate is larger
than the point parameter estimate itself. Follow-up discussions with surveyors about qualitative conversations with respondents
support this finding. For example, some respondents asked a lot of questions about range because they frequently travel by motorbike
to visit relatives in the countryside.
Demographics also appear to play an important role in consumers’ willingness to adopt electric motorcycles. Respondents were
decreasingly likely to select electric motorcycles with age. Those in their 30s and 40s were least likely of all, perhaps because these
age groups tend to travel more than younger or older people. In several model iterations, including Model 2, women were less likely
to choose electric motorcycles. Although the relationship is not statistically different from zero when including preference variables
(Model 3), two of the three random data samples also identified a statistically significant relationship at the 90% confidence level
(Appendix A). These findings add some additional uncertainty to the existing literature and suggest that markets for electric mo-
torcycles almost certainly vary by city and cultural norms. In Taipei, Chiu and Tzeng (1999) find that women are more likely to
choose electric motorcycles than men, while Jones et al. (2013) find no relationship between gender and motorcycle choice in Hanoi.
In Solo, women appear less likely to choose electric motorcycles—particularly when excluding the preference variables, many of
which are correlated with gender.
Once a household attains a baseline income of 500,000 rupiah per month, respondents have a statistically equivalent preference
for choosing a gas-powered motorcycle over no motorcycle (confirmed with a series of likelihood ratio tests across income groups).
This finding matches the relationship plotted in Fig. 3, where most households own two or more motorcycles regardless of income
group. By contrast, only respondents from households in the highest two income groups were more likely to select electric motor-
cycles than those in other income groups. They were not more likely to choose an electric motorcycle than a gas-powered one,
however. Including preference variables weakens the relationship between income and motorcycle choice, but only slightly.
In addition to motorcycle features and respondents’ socioeconomic attributes, preferences also matter. This is apparent in the
large variation in preferences for different motorcycle features and also in the directly estimated preference variables. Smokers—a
proxy for lack of concern about local air quality or personal safety—were less likely to choose electric motorcycles, as were those that
rarely think about the quality of the natural environment. Perceptions of e-bike quality also appear to influence the probability of
choosing an electric motorcycle. Respondents with a favorable view of the quality of the low-powered e-bikes currently available are
more likely to consider a higher powered electric motorcycle in the future. Respondents with a negative view of e-bike quality,
however, displayed a strong preference for gas-powered motorcycles.

5.1. Willingness to pay for electric features

Using the coefficient estimates from the mixed logit model, Table 4 presents consumers’ estimated marginal willingness to pay for
electric motorcycle range, speed, and charge times in relationship to both monthly payments and fuel prices. Across the sample, the
willingness to pay for shorter charge time, faster speeds, and greater range is economically important, as well as statistically sig-
nificant. On average, respondents were willing to pay an extra 35,000–40,000 rupiah in purchase price for a motorcycle with 10 km/

2
The down payment did not enter the model with statistical significance. This is unsurprising since the down payment is roughly equivalent to each of the 36
monthly payments and represents a small fraction of total costs.

7
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

Table 3
Mixed logit estimations of motorcycle choice.

Models

(1) (2) (3)

Motorcycle attributes
Monthly payment (Rp. millions) −2.933*** −2.917*** −2.674***
(0.869) (0.857) (0.840)
Fuel price (Rp. thousands) −0.700*** −0.685*** −0.666***
(0.052) (0.051) (0.050)
Charge time −0.622*** −0.563*** −0.520***
(0.073) (0.070) (0.069)
Max speed (100s of km/h) 3.176*** 3.220*** 3.094***
(0.539) (0.528) (0.513)
Max range (100s of km) 2.795*** 2.862*** 2.929***
(0.342) (0.339) (0.332)

Sociodemographic attributes
Female 1.315*** −0.119
(Electric specific) (0.155) (0.169)
Monthly income > Rp. 2.5 million 1.843*** 1.408***
(Electric-specific) (0.190) (0.177)
Monthly income > Rp. 500,000 2.167*** 1.437***
(Gas-specific) (0.271) (0.258)
Age −0.128*** −0.068*
(Electric-specific) (0.036) (0.036)
Age squared 0.002*** 0.001*
(Electric-specific) (0.001) (0.001)

Preference attributes
Smokes −1.584***
(Electric-specific) (0.201)
Has motorcycle 2.866***
(Electric-specific) (0.700)
Thinks e-bike quality is high 3.280***
(Electric-specific) (0.216)
Thinks e-bike quality is low −1.719***
(Electric-specific) (0.240)
Thinks a lot about environment 0.694***
(Electric-specific) (0.144)
Rarely thinks about environment −1.739***
(Electric-specific) (0.490)
Has motorcycle 6.446***
(Gas-specific) (0.822)

Intercepts
Electric intercept 9.051*** 9.606*** 5.713***
(0.908) (1.058) (1.162)
Gas Intercept 13.798*** 11.395*** 5.224***
(1.015) (0.967) (1.131)
Random coefficients (normally distributed)
Charge time 0.851*** 0.728*** 0.711***
(0.059) (0.057) (0.055)
Max speed (100s of km/h) 5.330*** 5.166*** 4.607***
(0.429) (0.426) (0.376)
Max range (100s of km) 4.377*** 4.202*** 3.646***
(0.302) (0.291) (0.265)
Electric intercept 2.573*** 2.957*** 2.058***
(0.311) (0.294) (0.334)
Gas Intercept 4.110*** 3.963*** 3.881***
(0.287) (0.288) (0.265)
Observations 6040 6040 6040
McFadden R2 0.390 0.394 0.411
Log Likelihood −2687 −2671 −2595

Note: ***p < 0.01; **p < 0.05; *p < 0.1.

h faster speed or 10 km longer range. Charge time appears particularly important to consumers with respondents equating an hour of
charge time with over 13% of the total purchase price and 12% of fuel costs. This reinforces the importance of including realistic
charging alternatives in electric motorcycle choice experiments. It also suggests that developing adequate charging infrastructure
could do much to encourage the adoption of electric motorcycles. I choose to report adjusted monthly willingness-to-pay figures as
described and justified below.

8
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

Table 4
Marginal willingness to pay for electric motorcycle features.

Willingness to pay

Monthly (Rp.)a Fuel (Rp.)

An hour shorter charge time 65,508 781

10 km/h faster 38,960 465
10 km more range 36,885 440

Average price for comparison 500,000 7000

1USD = Rp. 13,500

a
Monthly willingness to pay estimates adjusted by factor of 0.34 (see Appendix B).

Comparing the monthly payment and fuel price coefficients, respondents are willing to pay 250,000 rupiah more (50% of the
average purchase price) to reduce fuel prices by 1000 rupiah. This suggests that higher fuel prices could go a long way to encouraging
the adoption of moderately priced electric motorcycles, so long as the quality is competitive with lower cost gas-powered motor-
cycles. However, given typical fuel consumption of one liter per day, this relationship indicates that a typical respondent expects a
motorcycle to last nearly 30 years and has a discount rate of zero. This is economically unrealistic and likely stems from a weakness in
the stated preference survey. Many respondents—despite being asked to assume that motorcycles only vary across the dimensions
included in the survey—likely believe that a more expensive motorcycle is better along some undisclosed dimension such as quality,
size, or brand. That is, all things being equal, respondents may assume that a 600,000 rupiah per month motorbike is 100,000 rupiah
better than a 500,000 rupiah per month motorbike.
I therefore adjust the reported monthly payment estimates in Table 4 to reflect a 10% discount rate and a 12-year average
motorcycle life. The 10% discount rate is in line with a previous study from Java (Basu and Wong, 2015) and is in the 8–15% range
commonly applied in developing countries (Dhaliwal et al., 2013; Zhuang et al., 2007). Motorcycle lifespans depend substantially on
the quality of the motorcycle, maintenance, and use. A 12-year lifespan and 1-liter fuel consumption assumes approximately
150,000 km of total use. A 10- to 15-year lifespan with a 10% discount rate produces a range of adjustment factors from 0.29 to 0.36,
and is far less important to the final calculation than assumptions about the discount rate or typical fuel consumption.
Appendix B reports both the original willingness-to-pay estimates and adjusted estimates based on a series of assumptions about
fuel consumption (0.5, 1, and 2 liters per day), discount rates (5%, 10%, and 15%), and motorcycle lifespans (10, 12, and 15 years).
The assumptions result in willingness-to-pay estimates that range from 13–93% of the original estimates.

5.2. Potential for electric motorcycles

Taken together, these findings give several reasons for optimism and several reasons for pessimism about the potential for electric
motorcycles to reduce local pollution in Solo and other small- and medium-sized East Asian motorcycle cities. The first reason for
optimism is that nearly a quarter of respondents chose electric motorcycles under a realistic, immediately available, and not par-
ticularly convenient charging system. Replacing a quarter of conventional motorcycles with electric motorcycles could substantially
reduce local emissions—though almost certainly by less than a quarter since wealthier people with favorable environmental views are
less likely to drive the highest polluting vehicles and more likely to consider electric motorcycles. Given the relatively high will-
ingness to pay for faster charge times, a more convenient charging system would almost certainly increase adoption and reduce local
pollution as well.
In addition to charge time, a subset of respondents indicated a substantial willingness to pay for better quality electric motor-
cycles. This provides another reason for optimism since higher quality electric motorcycles are likely to be more expensive than
otherwise equivalent gas-powered ones. As gasoline prices increase and electric batteries become more efficient, furthermore, the
results suggest that consumers will be willing to accept the higher up-front cost of an electric motorcycle for the lower cost to power
it.
There are also some clear submarkets of consumers—such as the environmentally conscious—who likely favor electric motor-
cycles and may respond well to private or public campaigns to increase adoption. On average, respondents who think a lot about the
environment value electric motorcycles about 17% above the average purchase price more than other respondents do (Table 5). By
contrast, the 6% of respondents who rarely think about the environment would need around a 50% discount on average to choose an
electric motorcycle. Young people also appear more likely to consider electric motorcycles. This not only presents a potential target
market but suggests that interest in electric motorcycles could increase over time if the younger generation maintains its preferences
as it ages. Finally, the subset of the population with more favorable views of e-bike quality are substantially more likely to choose
electric motorcycles than those with unfavorable views. This suggests that greater exposure to higher performance electric vehicles
might influence overall perceptions of quality and change the opinions of some of respondents who were least likely to choose electric
motorcycles.
Despite these reasons for optimism, there are several compelling reasons to be pessimistic about the potential for electric mo-
torcycles to reduce local pollution in Solo. Just over half of respondents opted for gas-powered motorcycles in all five choice ex-
periments. This is despite presenting respondents with high electric speeds and ranges relative to the purchase prices presented.

9
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

Table 5
Implied monetary value of individual characteristics and preferences for electric motorcycle.

Monthly Paymenta Fuel

Thinks a lot about environment 87,375 1042

Rarely thinks about environment −218,938 −2611
Smokes −199,424 −2378
Monthly income > Rp. 2.5 million 177,307 2114
Thinks e-bike quality is high 413,007 4925
Thinks e-bike quality is low −216,448 −2581

Average price for comparison 500,000 7000

$1 USD = Rp. 13,500

a
Monthly willingness to pay estimates adjusted by factor of 0.34 (see Appendix B).

Given the survey design, it remains unclear whether and at what price points or quality levels these consumers might be willing to
consider electric motorcycles. Finally, the existing energy grid would likely have a hard time handling the increased energy needs
from a high penetration of high-powered electric motorcycles. An overnight 100% shift to high-powered electric motorcycles would
require an estimated 63% of current energy consumption to maintain average daily travel of 30 km per motorcycle (Guerra and
Artavia, 2016). Even a modest shift to lower performance electric motorcycles could create energy distribution challenges, parti-
cularly in the rainy season when power outages are common.

5.3. Limitations

As with any stated preference survey about a new technology, the results must be interpreted with caution. Although many
respondents have had exposure to e-bikes, the technology presented is unfamiliar and respondents may have a difficult time un-
derstanding the range of choices. The relative convenience of charge times may be particularly problematic, since consumers have
had little-to-no experience charging electric bikes or motorcycles. Furthermore, respondents may assume differences across the
alternatives even when they are asked to assume that these features are equivalent. For example, the surveys did not provide varying
acceleration speed. Focus group participants had earlier indicated that it was not an important consideration when choosing a
motorcycle and electric motorcycles tend to accelerate quickly. However, some respondents may care a great deal about acceleration
and assume that electric motorcycles accelerate more slowly or more quickly than gas-powered ones. As discussed in the previous
section, respondents probably associate a higher purchase price of motorcycles with a higher quality motorcycle.
The research also does not address a number of important public policy issues related to electric motorcycles. For example,
switching more of the local motorcycle fleet to electric vehicles would help reduce local emissions and potentially help reduce
dependence on fossil fuels, but would not prevent the physical degradation of public space due to motorcycles. The quieter engines of
electric motorbikes could even introduce safety issues, if pedestrians are less aware of their presence. Depending on sources of power,
furthermore, the switch to electric vehicles may concentrate local pollution around specific power plants, do little to reduce
greenhouse gas emissions, and even introduce new forms of environmental degradation through improper battery disposal.

6. Conclusion

This paper used 1208 complete stated preference surveys to examine the potential for electric motorcycles in Solo, Indonesia. As
in many rapidly growing cities throughout Asia, Latin America, and Africa (Law et al., 2015), motorcycle ownership and use have
increased rapidly over the past two decades. Motorcycles provide inexpensive, relatively safe, and convenient point-to-point travel
and dominate the transportation system. However, they also degrade the quality of the urban environment and create substantial and
harmful amounts of local pollution. Electric motorcycles could help reduce one of the most harmful effects of motorcycles, but have
barely penetrated the market in Solo or the rest of Indonesia.
The results of the survey and analysis indicate that there is almost certainly a market for electric motorcycles but their price and
performance will have to be competitive with low-cost gas-powered ones. Speed, range, charge time, and price all mattered sub-
stantially. Residents were willing to pay a 7–13% premium for motorcycles with 10 km longer range, 10 km/h faster speed, or an
hour shorter charge time. Charge time was particularly important, suggesting that improvements in battery charging technology and
charging infrastructure could have a substantial impact on consumers’ willingness to adopt electric motorcycles. Younger non-
smoking respondents with concerns about the environment and favorable views of e-bikes were most likely to choose electric mo-
torcycles.
As battery technologies improve and gasoline prices rise, the potential for electric motorcycles to replace higher polluting gas
motorcycles will increase. However, just over half of respondents never chose electric motorcycles regardless of price or quality.
Furthermore, the electricity grid may not be able to handle a substantial shift to electric motorcycles due to distribution and pro-
duction limitations. Finally, electric motorcycles only address emissions while ignoring other negative impacts of motorcycles and
may create additional environmental problems.
In summary, there appears to be an opportunity to lower emissions by attracting a subset of consumers with higher-powered
electric vehicles provided they are not too much more expensive than gas-powered ones. However, electric motorcycles are not a

10
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

silver bullet for sustainable transportation planning or even for local emissions more specifically. This suggests a need for additional
policies that might address the problem such as lowering tailpipe emissions from gas-powered motorcycles, taxing or banning highest
emitters, increasing fuel prices, and encouraging a shift to less-polluting modes.

Partners and funding support

The Kleinman Center for Energy Policy at the University of Pennsylvania, the Technologies for Safe and Efficient Transportation
University Transportation Center, and Kota Kita helped support this project. John Taylor, Rifai Ahmad, Fuad Jamil, Icha Thamrin,
Rizqa Hidayani, and Bima Pratama from Kota Kita helped organize, conduct, and document focus groups and stated preference
surveys. Lucia Artavia of the University of Pennsylvania provided research support throughout the project. Daniel Palencia and
Arturo Cadena from Urban Launchpad wrote the code to conduct the surveys using the Flocktracker application. Ricardo Dazanio,
Cornell University, and Megan Ryerson, University of Pennsylvania, provided feedback on model specifications, internal validity
tests, and willingness-to-pay estimations.

Appendix A

Mixed logit estimations of motorcycle choice: final specification with full dataset and three random 50% data samples.
Models

(1) (2) (3) (4)

Motorcycle attributes
Monthly payment (Rp. millions) −2.674⁎⁎⁎ −1.41 −2.991⁎⁎ −3.215⁎⁎⁎
(0.840) (1.172) (1.230) (1.219)
Fuel price (Rp. thousands) −0.666⁎⁎⁎ −0.649⁎⁎⁎ −0.558⁎⁎⁎ −0.673⁎⁎⁎
(0.050) (0.071) (0.070) (0.072)
Charge time −0.520⁎⁎⁎ −0.711⁎⁎⁎ −0.565⁎⁎⁎ −0.605⁎⁎⁎
(0.069) (0.102) (0.104) (0.106)
Max speed (100s of km/h) 3.094⁎⁎⁎ 2.413⁎⁎⁎ 3.835⁎⁎⁎ 3.569⁎⁎⁎
(0.513) (0.723) (0.798) (0.757)
Max range (100s of km) 2.929⁎⁎⁎ 2.398⁎⁎⁎ 3.152⁎⁎⁎ 2.225⁎⁎⁎
(0.332) (0.478) (0.488) (0.475)
Sociodemographic attributes
Female −0.119 −0.570⁎⁎ 0.279 −0.412⁎
(Electric specific) (0.169) (0.264) (0.245) (0.244)
Monthly income > Rp. 2.5 million 1.408⁎⁎⁎ 1.853⁎⁎⁎ 1.435⁎⁎⁎ 1.686⁎⁎⁎
(Electric-specific) (0.177) (0.267) (0.247) (0.254)
Monthly income > Rp. 500,000 1.437⁎⁎⁎ 1.846⁎⁎⁎ 0.957⁎⁎ 2.236⁎⁎⁎
(Gas-specific) (0.258) (0.399) (0.376) (0.390)
Age −0.068⁎ −0.127⁎⁎⁎ −0.279⁎⁎⁎ −0.045
(Electric-specific) (0.036) (0.049) (0.059) (0.053)
Age squared 0.001⁎ 0.002⁎⁎⁎ 0.004⁎⁎⁎ 0.001
(Electric-specific) (0.001) (0.001) (0.001) (0.001)
Preference attributes
Smokes −1.584⁎⁎⁎ −1.636⁎⁎⁎ −0.810⁎⁎⁎ −2.049⁎⁎⁎
(Electric-specific) (0.201) (0.301) (0.266) (0.305)
Has motorcycle 2.866⁎⁎⁎ 3.255⁎⁎⁎ 2.443⁎⁎ 3.603⁎⁎⁎
(Electric-specific) (0.700) (0.799) (0.982) (0.916)
Thinks e-bike quality is high 3.280⁎⁎⁎ 3.332⁎⁎⁎ 4.049⁎⁎⁎ 2.064⁎⁎⁎
(Electric-specific) (0.216) (0.302) (0.390) (0.279)
Thinks e-bike quality is low −1.719⁎⁎⁎ −1.499⁎⁎⁎ −0.600⁎⁎ −1.482⁎⁎⁎
(Electric-specific) (0.240) (0.339) (0.305) (0.307)
Thinks a lot about environment 0.694⁎⁎⁎ −0.051 0.885⁎⁎⁎ 0.538⁎⁎
(Electric-specific) (0.144) (0.208) (0.214) (0.214)
Rarely thinks about environment −1.739⁎⁎⁎ −2.517⁎⁎⁎ −5.044⁎⁎⁎ −1.431⁎
(Electric-specific) (0.490) (0.681) (0.781) (0.756)
Has motorcycle 6.446⁎⁎⁎ 6.422⁎⁎⁎ 4.363⁎⁎⁎ 6.446⁎⁎⁎
(Gas-specific) (0.822) (0.961) (1.195) (1.066)
Intercepts
Electric intercept 5.713⁎⁎⁎ 7.962⁎⁎⁎ 7.880⁎⁎⁎ 5.371⁎⁎⁎

11
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

(1.162) (1.578) (1.847) (1.580)

Gas Intercept 5.224⁎⁎⁎ 5.525⁎⁎⁎ 6.981⁎⁎⁎ 4.582⁎⁎⁎
(1.131) (1.460) (1.753) (1.544)
Random coefficients (normally distributed)
Charge time 0.711⁎⁎⁎ 0.806⁎⁎⁎ 0.730⁎⁎⁎ 0.695⁎⁎⁎
(0.055) (0.079) (0.083) (0.084)
Max speed (100s of km/h) 4.607⁎⁎⁎ 5.505⁎⁎⁎ 5.494⁎⁎⁎ 4.911⁎⁎⁎
(0.376) (0.562) (0.663) (0.545)
Max range (100s of km) 3.646⁎⁎⁎ 4.203⁎⁎⁎ 3.766⁎⁎⁎ 4.204⁎⁎⁎
(0.265) (0.413) (0.429) (0.400)
Electric intercept 2.058⁎⁎⁎ 0.169 3.241⁎⁎⁎ 1.244⁎⁎
(0.334) (2.007) (0.419) (0.538)
Gas Intercept 3.881⁎⁎⁎ 3.854⁎⁎⁎ 2.672⁎⁎⁎ 3.776⁎⁎⁎
(0.265) (0.327) (0.494) (0.369)
Observations 6040 3020 3020 3020
McFadden R2 0.411 0.410 0.404 0.405
Log Likelihood −2595 −1340 −1281 −1290
Note: ⁎⁎⁎p < 0.01; ⁎⁎p < 0.05; ⁎p < 0.1.

Appendix B

Willingness to pay for electric motorcycle features based on assumptions about discount rate, average, motorcycle life, and fuel consumption. Preferred estimates are
bolded.
Discount Estimated motorcycle Liters of fuel per An hour shorter charge 10 km/h 10 km more Adjustment
rate life day time faster range factor

0% 25 1 194,543 115,700 109,540 (unadjusted)

5% 10 2 180,157 107,144 101,440 0.93
5% 12 2 157,313 93,558 88,577 0.81
5% 15 2 134,320 79,884 75,631 0.69
10% 10 1 71,367 42,444 40,184 0.37
10% 12 1 65,508 38,960 36,885 0.34
10% 15 1 58,033 34,514 32,676 0.30
15% 10 0.5 29,434 17,505 16,573 0.15
15% 12 0.5 27,999 16,652 15,765 0.14
15% 10 0.5 25,534 15,186 14,377 0.13

References

Adnan, Muhammad, 2014. Passenger car equivalent factors in heterogenous traffic environment-are we using the right numbers? In: Procedia Engineering, Fourth
International Symposium on Infrastructure Engineering in Developing Countries, IEDC 2013, vol. 77, pp. 106–113. http://dx.doi.org/10.1016/j.proeng.2014.07.
004.
Bappeda Kota Surakarta, 2014. Surakarta in Figures (2008–2013). Surakarta.
Basu, Karna, Wong, Maisy, 2015. Evaluating seasonal food storage and credit programs in east Indonesia. J. Dev. Econ. 115 (July), 200–216. http://dx.doi.org/10.
1016/j.jdeveco.2015.02.001.
Beck, Laurie F., Dellinger, Ann M., O’Neil, Mary E., 2007. Motor vehicle crash injury rates by mode of travel, United States: using exposure-based methods to quantify
differences. Am. J. Epidemiol. 166 (2), 212–218. http://dx.doi.org/10.1093/aje/kwm064.
Beggs, Steven D., Cardell, N. Scott, 1980. Choice of smallest car by multi-vehicle households and the demand for electric vehicles. Transp. Res. Part A: Gen. 14 (5–6),
389–404. http://dx.doi.org/10.1016/0191-2607(80)90057-6.
Brownstone, David, Bunch, David S., Train, Kenneth, 2000. Joint mixed logit models of stated and revealed preferences for alternative-fuel vehicles. Transp. Res. Part
B: Methodol. 34 (5), 315–338. http://dx.doi.org/10.1016/S0191-2615(99)00031-4.
Brownstone, David, Train, Kenneth, 1998. Forecasting new product penetration with flexible substitution patterns. J. Econom. 89 (1–2), 109–129. http://dx.doi.org/
10.1016/S0304-4076(98)00057-8.
Bunnell, Tim, Miller, Michelle Ann, Phelps, Nicholas A., Taylor, John, 2013. Urban development in a decentralized Indonesia: two success stories? Pacific Affairs 86
(4), 857–876. http://dx.doi.org/10.5509/2013864857.
Calfee, John E., 1985. Special issue economic models of automobile demand estimating the demand for electric automobiles using fully disaggregated probabilistic
choice analysis. Transp. Res. Part B: Methodol. 19 (4), 287–301. http://dx.doi.org/10.1016/0191-2615(85)90037-2.
Campbell, Andrew A., Cherry, Christopher, Ryerson, Megan, Jones, Luke, 2014. Better pen-and-paper surveys for transportation research in developing countries: a
modified, stated preference, pivoting approach. Transp. Res. Rec.: J. Transp. Res. Board 2405 (1), 42–48. http://dx.doi.org/10.3141/2405-06.
Cao, Nguyen, Sano, Kazushi, 2012. Estimating capacity and motorcycle equivalent units on urban roads in Hanoi, Vietnam. J. Transp. Eng. 138 (6), 776–785. http://
dx.doi.org/10.1061/(ASCE)TE.1943-5436.0000382.
CAREPI, 2009. Regional Energy Baseline Central Java Province. Pusat Studi Energi Regional Universitas Muhammadiyah Yogyakarta & Departemen energi dan Sumber
Daya Mineral Provinsi Jawa Tengah, Yogyakarta. <https://ec.europa.eu/energy/intelligent/projects/sites/iee-projects/files/projects/documents/carepi_
regional_energy_situation_central_java.pdf>.
Chen, Yuyu, Ebenstein, Avraham, Greenstone, Michael, Li, Hongbin, 2013. Evidence on the impact of sustained exposure to air pollution on life expectancy from

12
E. Guerra Transportation Research Part D xxx (xxxx) xxx–xxx

China’s Huai River Policy. Proc. Natl. Acad. Sci. 110 (32), 12936–12941. http://dx.doi.org/10.1073/pnas.1300018110.
Cherry, Christopher, Cervero, Robert, 2007. Use characteristics and mode choice behavior of electric bike users in China. Transp. Policy 14 (3), 247–257. http://dx.
doi.org/10.1016/j.tranpol.2007.02.005.
Cherry, Christopher, Weinert, Jonathan, Xinmiao, Yang, 2009. Comparative environmental impacts of electric bikes in China. Transp. Res. Part D: Transp. Environ. 14
(5), 281–290. http://dx.doi.org/10.1016/j.trd.2008.11.003.
Chiu, Yi-Chang, Tzeng, Gwo-Hshiung, 1999. The market acceptance of electric motorcycles in Taiwan experience through a stated preference analysis. Transp. Res.
Part D 4 (2), 127–146.
Daziano, Ricardo A., 2013. Conditional-logit Bayes estimators for consumer valuation of electric vehicle driving range. Resour. Energy Econ. 35 (3), 429–450. http://
dx.doi.org/10.1016/j.reseneeco.2013.05.001.
Dhaliwal, Iqbal, Duflo, Esther, Glennerster, Rachel, Tulloch, Caitlin, 2013. Comparative cost-effectiveness analysis to inform policy in developing countries: a general
framework with applications for education. In: Glewwe, Paul (Ed.), Education Policy in Developing Countries. University of Chicago Press, NY.
Dimitriou, Harry T., 2006. Towards a generic sustainable urban transport strategy for middle-sized cities in Asia: lessons from Ningbo, Kanpur and Solo. Habitat Int. 30
(4), 1082–1099. http://dx.doi.org/10.1016/j.habitatint.2006.02.001.
Dinas Pendapatan Dan Pengelolaan Aset Daerah Provinsi Jawa Tengah, 2015. < http://dppad.jatengprov.go.id > .
Fan, Henry S.L., 1990. Passenger car equivalents for vehicles on Singapore expressways. Transp. Res. Part A: Gen. 24 (5), 391–396. http://dx.doi.org/10.1016/0191-
2607(90)90051-7.
Fishman, Elliot, Cherry, Christopher, 2015. E-bikes in the mainstream: reviewing a decade of research. Transp. Rev. 1–20. http://dx.doi.org/10.1080/01441647.2015.
1069907.
Greenstone, Michael, Nilekani, Janhavi, Pande, Rohini, Ryan, Nicholas, Sudarshan, Anant, Sugathan, Anishi, 2015. Lower pollution, longer lives life expectancy gains
if India reduced particulate matter pollution. Econ. Polit. Weekly 8, 40–46.
Guerra, Erick, Artavia, Lucia, 2016. Energy Production and the Potential for Electric Motorcycles in Solo and Central Java, Indonesia. Policy Digest. Kleinman Center
for Energy Policy. <http://kleinmanenergy.upenn.edu/policy-digests/energy-production-and-potential-electric-motorcycles-solo-and-central-java-indonesia>.
Hidalgo, Dario, Gutiérrez, Luis, 2013. BRT and BHLS around the world: explosive growth, large positive impacts and many issues outstanding. Res. Transp. Econ. 39
(1), 8–13. http://dx.doi.org/10.1016/j.retrec.2012.05.018.
Hieu, Vu Van, Quynh, Le Xuan, Ho, Pham Ngoc, Hens, Luc, 2013. Health risk assessment of mobility-related air pollution in Ha Noi, Vietnam. J. Environ. Prot. 4,
1165–1172. http://dx.doi.org/10.4236/jep.2013.410133.
Jones, Luke R., Cherry, Christopher R., Vu, Tuan A., Nguyen, Quang N., 2013. The effect of incentives and technology on the adoption of electric motorcycles: a stated
choice experiment in Vietnam. Transp. Res. Part A: Policy Pract. 57 (November), 1–11. http://dx.doi.org/10.1016/j.tra.2013.09.003.
Kota Kita. 2012. “Solo Kota Kita Population Estimates.” <http://solokotakita.org/>.
Kota Kita, 2015. The Angkots of Solo. Kota Kita, Solo. <http://www.kotakita.org/project-angkot-study.html>.
Law, Teik Hua, Hamid, Hussain, Goh, Chia Ning, 2015. The motorcycle to passenger car ownership ratio and economic growth: a cross-country analysis. J. Transp.
Geogr. 46 (June), 122–128. http://dx.doi.org/10.1016/j.jtrangeo.2015.06.007.
Ma, Hongrui, Balthasar, Felix, Tait, Nigel, Riera-Palou, Xavier, Harrison, Andrew, 2012. A new comparison between the life cycle greenhouse gas emissions of battery
electric vehicles and internal combustion vehicles. Energy Policy 44 (May), 160–173. http://dx.doi.org/10.1016/j.enpol.2012.01.034.
Ministry of Energy and Mineral Resources and Energy Conservation, 2012. Development of New and Renewable Energy and Energy Conservation in Indonesia.
Ministry of Energy and Mineral Resources and Energy Conservation. <http://www.usea.org/sites/default/files/event-/Indonesia%20Country%20Presentation_0.
pdf>.
MIT Department of Urban Studies, and Planning, Mobility Futures Collaborative, and Urban Launchpad/MX. 2016. “Flocktracker.” <http://www.flocktracker.org/>.
Oanh, Nguyen Thi, Kim, Mai Tai, Phuong, Thuy, Permadi, Didin Agustian, 2012. Analysis of motorcycle fleet in Hanoi for estimation of air pollution emission and
climate mitigation co-benefit of technology implementation. Atmos. Environ. 59, 438–448. http://dx.doi.org/10.1016/j.atmosenv.2012.04.057.
Rifai, Ahmad, Guerra, Erick, Haggery, Michael, Taylor, John, 2011. Informal Transportation Networks in Three Indonesian Cities. Cities Development Initiative for
Asia. <http://cdia.asia/2012/05/02/informal-public-transportation-networks-in-three-indonesian-cities/>.
Rongviriyapanich, Terdsak, Suppattrakul, Charong, 2005. Effects of motorcycles on traffic operations on arterial streets. J. Eastern Asia Soc. Transp. Stud. 6, 137–146.
http://dx.doi.org/10.11175/easts.6.137.
The Economist, 2009. Things Do Not Fall Apart, September 10. < http://www.economist.com/node/14391454 > .
Train, Kenneth, 2009. Discrete Choice Methods with Simulation, second ed. Cambridge University Press, New York. <http://elsa.berkeley.edu/books/choice2.html>.
World Health Organization, 2015. Global Status Report on Road Safety. World Health Organization. <http://www.who.int/violence_injury_prevention/road_safety_
status/2015/en/>.
Yeung, Jian Sheng, Wong, Yiik Diew, Secadiningrat, Julius Raditya, 2015. Lane-harmonised passenger car equivalents for heterogeneous expressway traffic. Transp.
Res. Part A: Policy Pract. 78 (August), 361–370. http://dx.doi.org/10.1016/j.tra.2015.06.001.
Zegras, Christopher, Butts, Kuan, Cadena, Arturo, Palencia, Daniel, 2015a. Spatiotemporal dynamics in public transport personal security perceptions: digital evidence
from Mexico City’s periphery. In: Rietveld, Piet, Givoni, Moshe, Thomopoulos, Nikolas (Eds.), ICT for Transport: Opportunities and Threats. Edward Elgar
Publishing, NY.
Zegras, Christopher, Eros, Emily, Butts, Kuan, Resor, Elizabeth, Kennedy, Stephen, Ching, Albert, Mamun, Muntasir, 2015b. Tracing a path to knowledge? Indicative
user impacts of introducing a public transport map in Dhaka, Bangladesh. Cambridge J. Regions Econ. Soc. 8 (1), 113–129. http://dx.doi.org/10.1093/cjres/
rsu028.
Zero Motorcycles. 2015. Zero Motorcycles – The Electric Motorcycle Company. < http://www.zeromotorcycles.com/ > .
Zhuang, Juzhong, Zhihong, Liang, Tun, Lin, Franklin, De Guzman, 2007. Theory and Practice in the Choice of Social Discount Rate for Cost-Benefit Analysis: A Survey.
< https://think-asia.org/handle/11540/1853 > .

13