A dynamic marketing model for hybrid electric vehicles: A case study of Taiwan

(1)

A dynamic marketing model for hybrid electric vehicles: A case

study of Taiwan

Chaug-Ing Hsu

a,⇑

, Hui-Chieh Li

b

, Shan-Mei Lu

a

Department of Transportation Technology and Management, National Chiao Tung University, Hsinchu 300, Taiwan, ROC

b

Tourism Management, Ta Hwa University of Science and Technology, Hsinchu County 307, Taiwan, ROC

a r t i c l e

i n f o

Keywords:

Hybrid electric vehicles Small world network Word of mouth

a b s t r a c t

This study uses an integrated model utilizing a small-world network and choice-based con-joint adoption model to examine the dynamics of consumer choice and diffusion in the hybrid electric vehicles market. It specifically compares the effectiveness of hybrid diffu-sion through the traditional word of mouth and via social media. The results show that without the advantage of increased gasoline prices, the growth of the hybrid vehicles mar-ket is insignificant, and that the Internet has a significant influence on the word of mouth effect in the purchasing process. Hybrid electric vehicles market shares decrease dramati-cally as a result of negative word of mouth communication via social media. The use of a higher fuel taxes is more effective than providing a subsidy for disposing of old vehicles and purchasing a hybrid.

1. Introduction

The advantage of hybrid electric vehicles (HEVs) over conventional automobiles lies in their relatively lower environmen-tal impact resulting from better fuel efﬁciency. There is a strong evidence of the rate of the adoption of HEVs and gasoline prices, the status given to different vehicles types, and the driving patters and the socioeconomic features of potential buy-ers, but we know less about the impacts of ofﬁcial incentive polices. But beyond this we have, for example, little idea of how individuals obtain information about the relative attributes of HEVs, how they assess this information or how they pass it onto other. Here the focus is to learn more about the role of word of mouth (WOM) and other communications channels in the HEV market.

2. Dynamic choice model

In the study of the role of word of mouth’s influence on car buying behavior, the consumer choice probability is described as a time-varying process based on the socioeconomic and spatial variables relevant to the dynamic characteristics exhib-ited. Specifically, the consumer’s decision to adopt is the result of these variables at the time at the time of the decision, but what is relevant, and how, is influenced by prior WOM communications.

WOM effects are diffused via consumers’ social networks. Each consumer, and the social relationship between consumers, is represented by a node and by links. A strong tie refers to the social relationship in which consumers frequently interact with each other, while a weak tie is less inﬂuential and generally involves less interaction. The signiﬁcance of weak ties is

that they are far more likely to be a bridge than strong ones, and more ﬂexible; hence we followGoldenberg et al. (2007), and

⇑Corresponding author.

E-mail address:cihsu@mail.nctu.edu.tw(C.-I. Hsu).

Contents lists available atSciVerse ScienceDirect

Transportation Research Part D

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / t r d

(2)

treat them as random effects. For example, vehicle usage experiences and purchase information may be randomly generated via consumers occasionally chatting with a taxi driver, or to clients during regular interactions. Consumers can also receive product information via social media where one or more individuals with similar interests share reviews, often at a very ra-pid pace. A particular social media is denoted as a node where information from the social media that impacts on WOM is discussed.

An experienced consumer has already owned more than one vehicle, and his or her choice behavior differs from that of an

inexperienced consumer because of this prior knowledge (Bettman and Park, 1980). We use an experienced consumer’s

util-ity function based onRoorda et al. (2009)where j and I denote a speciﬁc consumer and choice alternative. An experienced

consumer can choice: to maintain the status quo; scrap an existing vehicle without replacement; sell an existing vehicle and replace it with a HEV, or with a gasoline vehicle; purchase an additional HEV or an additional gasoline vehicle; or scrap an

existing vehicle with a replacement of HEV or gasoline vehicle. The utility Ut

ijof the experienced consumer I who chooses

alternative j at time t can is;

Utij¼

a

x t jþ by t iþ h

m

t iþ

c

z t jþ dw t ijþ

e

ð1Þ

where

a

, b, h,

c

and d are the vectors of the parameters to be estimated; xt

jis a vector of the characteristics of alternative j at

time t; yt

iand

m

tiare vectors of socioeconomic variables and the vehicle ﬂeet of experienced consumer i at time t; ztjis a vector

of government incentives with respect to alternative j at time t; wt

ijis the inﬂuence of WOM on buying alternative j on

expe-rienced consumer i at time t; and e is the error term. The ‘‘costs’’ associated with the current vehicle is seen in terms of vehi-cle utilization, increases in family size and vehivehi-cle maintenance costs. The alternatives for an inexperienced consumer are to

purchase a HEV or a gasoline vehicle. The utility of inexperienced consumer l who chooses alternative k at time t, Ut

lk, is thus Ut lk¼

a

0x t kþ b 0_yt lþ

c

0z t kþ d 0_wt lkþ

e

0 ð2Þ

A WOM function is used to investigate the inﬂuences of social relationship, memory, negative and positive communications

as well as the probability of having received information on consumer choice.Goldenberg et al. (2007)classify potential

sumers into three categories; the positive who adopt the new product and can be expected to influence other potential con-sumers through future positive WOM; the disappointed who adopt the new product but are not satisfied, and thus potentially to negative WOM influences on potential consumers; and rejecters who are ex-potential adopters who received negative WOM and who may spread negative information to other potential consumers. There are also consumers who do not adopt the product but may still spread positive WOM influence on potential consumers, although here we classify con-sumers only as A, B, C or D corresponding to positive, ex-potential and positive concon-sumers, disappointed concon-sumers and

rejecters. Thus, the WOM function, wt

ijis thus wt ij¼ Rp Ps Xt tmþ1 Qtps m þ Pw Xt tmþ1 Qtpw m þ Pi Xt tmþ1 Qtpi m 0 B B B B @ 1 C C C C A Rn Ns Xt tmþ1 Qtns m þ Nw Xt tmþ1 Qtnw m þ Ni Xt tmþ1 Qtni m 0 B B B B @ 1 C C C C A ð3Þ

where Rpand Rnare the probabilities of receiving positive and negative WOM, Psand Nsare the inﬂuential levels of positive

and negative information from strong ties, Pwand Nware the inﬂuences of positive and negative information from weak ties,

Piand Niare the inﬂuence levels of positive and negative information from social media, Qtps;Q

t

pware the numbers of A and B

consumers in the strong and weak tie networks, Qt

nsand Q

t

nware the number of C and D consumers in the strong and weak tie

networks, and Qt

piand Q

t

ni are the number of users spreading information via social media. Following,Dodds and Watts

(2005), m is the time that consumers retain memory of WOM information received from (t m + 1) to t.

3. Questionnaire and results

Two questionnaires are used to illicit information on experienced and inexperienced consumers, and to obtained data regarding consumer characteristics and preferences in purchasing HEVs. The questionnaire consists of two sections. The ﬁrst, asked questions related to responders’ socioeconomic status such as age, gender, income, residence, vehicle possession and environment awareness. A quasi-stated preference experiment is also contained included asking respondents to choose from hypothetical choice sets. The choices are described by bundles of attributes values including gasoline price, acceleration

per-formance,1_{vehicle prices and WOM inﬂuences. The second collected data for calibrating the parameters in the small-world}

net-work and the WOM function. The respondents are asked how many people, as well as the social relationships they discussed their choices before purchasing a vehicle. The respondents were also asked for what they thought was the level of influence exerted by various social relationships such as strong or weak ties and social media. The level of influence was measured on a five-point scale, ranging from extremely important, very important to unimportant with scores of 100, 75, 50, 25 and 0.

Max-imum likelihood estimation is used to estimate ^

a

; ^b; ^h; ^

c

and ^din Eq. (1) and ^

a

0_{; ^}_b0_{; ^}_h0_{; ^}

_c

0_{and ^}_d0_{in Eq. (2).}

1

(3)

An Internet survey is used to collect the data, with e-mails for pre-notiﬁcation to improve the response rate. The main disadvantage of an Internet survey is the possibility of getting a biased sample due to a decreased access rate to the Internet

(Iragüen and Ortúzar, 2004), but in Taiwan it is over 70% of the population have Internet access (TWNIC, 2009). The survey

was administrated for four weeks in February 2011.

The majority of the inexperienced respondents were under 30 years of age (65.3%),2_{with ﬁrst-time vehicle buyers mainly}

from the younger consumer group. Most of the experienced respondents were young, highly educated consumers with a high income, and the major respondents with these characteristics are more likely to be attracted by low-pollution vehicles in line

with the ﬁndings ofEwing and Sarigollu (1998).

Six and seven scenarios were presented giving 645 observations for inexperienced and 917 for experienced consumers. A binary logit model was calibrated for estimating the parameters in = equation 2, while a multi logit model was used for Eq.

(1). The results are shown inTable 1.3_{As seen in the upper part relating to inexperienced customers, vehicle price and}

perfor-mance, consumer’s monthly income, residence and environmental awareness, and gasoline price and the WOM effect are all

signiﬁcant, with the parameter signs consistent with other studies includingPotoglou and Kanaroglou (2007), Erdem et al.

(2010), andAxsen et al. (2009). Fuel efﬁciency and pollution emission are not listed in the upper part ofTable 1because they are not signiﬁcant.

Turning to experienced consumers, the lower part ofTable 1, vehicle price, monthly income, accumulated mileage and

age of the present vehicle are signiﬁcant in affecting the intention of trading in an existing vehicle for a new one; results

in line withBrownstone et al. (2000)andMarell et al. (2004). Because gasoline is required for both new and used vehicles,

its price is excluded from the model alternative as trading in an existing for a new gasoline vehicle. The impact of WOM on the purchase of HEVs is significant and positive as are government subsidies. Compared with several prior studies, variables such as fuel efficiency, performance, emissions and environmental awareness are not included in our model because of significance.

Table 1

The parameters in the utility functions.

Variable type Explanatory variable Coefﬁcient

Inexperienced consumers

Alternative speciﬁc constant Purchase gasoline vehicle 3.8451***

Generic variable Vehicle Price 5.63E06***

Acceleration performance 1.45E01***

Alternative speciﬁc variable (HEV) Gasoline price 1.13E01***

WOM effect 1.30E03***

Monthly income 1.23E06***

Environment awareness 4.39E01***

Residence 9.16E01***

Goodness of ﬁt statistics: LL(0) = 401.3, LL(c) = 383.5, LL(^b) = 331.4216,q2

= 0.1742 Experienced consumers

Alternative speciﬁc constant Stay in status quo 3.0678***

Sell the used vehicle with a replacement of HEV 3.2548***

Sell the used vehicle with a replacement of gasoline vehicle 1.2632

Additional purchase of HEV 1.00E01

Additional purchase of gasoline vehicle 8.41E01*

Scrap the used vehicle with a replacement of HEV 1.6729

Scrap the used vehicle with a replacement of gasoline vehicle 9.71E02

Alternative speciﬁc variable Vehicle price 2.40E06***

Gasoline price 7.32E02***

WOM effect 1.12E03***

Accumulated mileage 1.06E05***

Monthly income 5.86E07**

Alternative speciﬁc variable Vehicle price 3.52E06***

Accumulated mileage 8.0E06***

Age of the original vehicle 6.2E02***

Monthly income 9.82E07***

Alternative speciﬁc variable WOM effect 1.03E03*

Alternative speciﬁc variable Age of the original vehicle 9.96E02*

Government subsidy 1.06332*

Alternative speciﬁc variable Age of the original vehicle 1.78E01**

Goodness of ﬁt statistics: LL(0) = 1167.2800, LL(c) = 861.3024, LL(^b) = 791.0511,q2

= 0.32231

*_{Signiﬁcant at the 10% level.} **

Signiﬁcant at the 5% level.

*** _{Signiﬁcant at the 1% level.}

2

Details can be found athttp://www.motc.gov.tw/ch/.

3

The t-statistics have been reduced ðpffiffiffi6Þ1_{and ð}pffiffiffi₇_Þ1_{for the inexperienced and experienced consumers compared to the original values due to the}

(4)

Table 2shows the values of the level of influence of social relationships. The average and median of influential level of different social relationship are significantly different from each other by the assessment of t-test. Negative information is seen as more influential than positive information in all kinds of social relationships. Our survey also shows the average number of people involved in the discussions on vehicle purchase and usage experience is 3.4, while those occasionally ran into and chatted number 1.2.

4. Case study

A case study is used to demonstrate the application of the model. It is assume that 80% of consumers in the market are experienced in according with the survey. The simulation has several stages

Step 0: WOM’s influence is initialized as zero. A database is created of consumer profiles including their average monthly income, vehicle possession, age of vehicle, accumulated mileage, WOM influence, environmental awareness and location of residence. Accumulated mileage, average monthly income and age of vehicle are randomly assigned to the node

fol-lowing the real data distribution inMinistry of Transport and Communications (2008). The assignment of environmental

awareness and location of residence to a node is based on our survey.

Step 1: At each time step t, consumer utility is estimated with respective to alternatives in which the utility gained by consumers can be calculated from equations 1 and 2. In addition to the consumer proﬁle, the input data includes vehicle and gasoline prices. We assume that consumers are maximizing utility in their decision-making. If the optimal alternative is to maintain the status quo or scrap the present vehicle without replacement, the experienced consumer becomes a non-adopter. He or she is labeled a HEV/gasoline adopter if selling or scrapping the existing car and replaced it by a HEV or gasoline vehicle maximizes utility. Inexperienced consumers are labeled as HEVs/gasoline adopters if acquisition of a HEVs or gasoline leads to maximum utility.

Step 2: Generate a social network based on the small world theory. Each node is connected to the nearest nodes with strong ties. Based on the survey, the number of weak ties is one with a randomly chosen probability between 0.01 and

0.1.4 _{For a node with weak ties to others, we randomly generate a number between one and 20 as the length of the}

discussion.

Step 3: Estimate of WOM inﬂuences are based on Eq. (3).

Step 4. Update the database of the consumer proﬁle, including a vehicle’s accumulated mileage and age.

Step 5: Proceed to the next time step t = t + 1. Inexperienced consumers who make their purchase decisions at time step t are randomly chosen with a probability of 0.1. Go to Step 2.

The model is veriﬁed by comparing the simulation results with data of the market share of the HEVs in the US. Taking Toyota as an example, the prices of gasoline vehicle and HEVs with engine over 1799 cc are $16,360 and $23,810; while those

with engines over 3456 cc are $30,955 and $38,300 (Toyota, 2011). In the simulation, the price of the HEV is set to be 1.45

times that of a gasoline vehicle. Gasoline price data are taken from the US Energy Information Administration (EIA). In

addi-tion, a US government purchase subsidy ranging from $500 to $3000 per HEV is considered (Gallagher and Muehlegger,

2011). For simpliﬁcation, we assume there are one weak and four strong ties, and the probability representing the

random-ness of the network is 0.05.

Fig. 1gives the simulation results set against the historical data of the market share of HEVs; differences in the estimated and the historical data are between 0.05% and 1.7%. The estimated market share of HEVs varies with the gasoline price as also

shown inFig. 1. There was an increase in the market share of HEVs in 2009, at a time when consumers were being

encour-aged to scrap their existing vehicles if their consumption was 18 mpg or less as part of the US ‘‘Cash for Clunkers Program’’. In looking at the impacts of government policies on the market for HEVs, we assume the socioeconomic characteristics of

the consumers and their vehicles are distributed following theMinistry of Transport and Communications (2008)survey.

Table 2

The inﬂuence of social relationships.

Social relationship Information category Inﬂuential level

Average Median

Strong tie Positive 58.63 50

Negative 62.72 75

Weak tie Positive 35.82 25

Negative 40.50 50

Social media Positive 53.66 50

Negative 58.71 50

4

The link-rewired probability p (0 < p < 1) represents the randomness of the resulting small-world network and the dynamic properties of network structure (Watts and Strogatz, 1998). The network is most ﬁtted to the characteristics of small-world network as (0.01 < p < 0.1).

(5)

The gasoline price when HEVs enter the market is assumed to be NT$30 per liter5_{in the ﬁrst year, with an annual growth rate}

of 1.8% (Energy Information Administration, 2011). The price of a gasoline vehicle is taken as NT$600,000, while that of a HEV is

NT$1.5 million with an annual decreasing rate of 9% (Axsen et al., 2009). There are assumed to be 6.65 million potential vehicle

consumers in Taiwan.6

Fig. 2shows the simulation results. The lower vehicle and higher gasoline prices result in a 95% of HEVs market share in the 10th year after HEVs are introduction. To examine the sensitivity of the results to trends in fuel and vehicle prices, two further scenarios applied. In the ﬁrst it is assumes that the price of HEVs remains 1.56 times that of similar gasoline vehicle after the 5th year, and in the second, that in addition the gasoline price remains at NT$32.8/liter after the 5th year without any further increases.

As shown inFig. 3, market HEV share initially decreases to 30% and 9% in the ﬁrst and second scenarios, suggesting that

without a considerable rise in gasoline prices, the HEV market will shrink because their high capital cost. After maintaining the same level during the 5th and 6th years, the market share of HEVs increases even though there is no downward trending

in their price, and this may be the result of WOM effects as well as the increase in gasoline prices.Fig. 4examines the HEVs

market share under various growth rate combinations for gasoline price and HEV price.

If HEVs are priced at over NT$1.25 million (Fig. 4), the impact of increasing gasoline prices on the HEV market share is

negligible. There are some effects at lower prices; e.g. if an HEV costs NT$1.02 million, the HEV market shares are 3.4% and 8.0% when price of gasoline rises 1.8% and 3.6%, and 41.9% and 78.0% when the increases are 7.2% and 10%. These results

are in line withLave and MacLean (2002)ﬁnding that HEVs will have signiﬁcant sales as fuel prices rise several-fold.

Turning to relationships between the HEV market share and the average accumulated mileage of a consumers’ existing

vehicles,Fig. 5shows a quadratic speciﬁcations7when gasoline is priced at NT$32.91/liter. The data refers to HEVs market

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 2003 2004 2005 2006 2007 2008 2009 Year market share (%) 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5

gasoline price (US$/gallon)

Actual Estimated gasoline price (US$/gallon)

Fig. 1. Simulation results and historical data of the market share of HEVs in the US.

5_{NT$100 = $3.44 (January, 2013)} 6

To minimize the computing load, the study reduces the sample size at a rate of 10% at each trial and run simulations with different random seeds. After several trials, we found that the differences in HEVs market shares are small and the variations converged to an average value.

0 10 20 30 40 50 60 70 80 90 100 1 2 3 4 5 6 7 8 9 10 11

Year since HEV introduced

market share (%) 0 5 10 15 20 25 30 35 Market penetration (%)

Market Share (%) Market penetration (%)

Fig. 2. Simulation results.

7

(6)

share after they have been available for 10 years, with x-axis referring to average accumulated kilometers of used vehicles in the market in the 10th year. The HEVs market share rises as the accumulated mileage traveled of the used vehicle increases. This relationship, however, becomes convex (right side of ﬁgure) if gasoline prices become NT$36.91 per liter; high gasoline prices

0 5 10 15 20 25 30 35 0 1 2 3 4 5 6 7 8 9 10

market share (%)

1st scenario 2nd scenario

Fig. 3. Market share of HEVs under various scenarios.

0 10 20 30 40 50 60 70 80 90 100 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5

HEV price (unit: NT$ 1000000)

m ar k et s h ar e ( % )

growth rate=1.8% growth rate=3.6% growth rate=7.2% growth rate=10%

(1.02, 78%)

(1.02, 42%)

(1.02, 8%)

(1.02, 3%)

Fig. 4. HEVs market share under combinations of growth rates of gasoline and HEV prices.

(7)

lead to greater intentions regarding the purchasing of HEVs for both inexperienced and experienced consumers as shown in

Table 2.

To assess the impact of government subsidies on promoting HEVs, we take a benchmark scenario that assumes an initial gasoline price of NT$32.91/liter with an annual growth rate of 1.8%, with HEVs and gasoline vehicle priced at NT$1 million and NT$600,000. We consider four subsidy policies; Scenarios A and B assume a subsidy of NT$100,000 for disposing of a vehicle nine years or older or with fuel consumption of 12.5 liter per 100 km or over, and replacing it with a HEV during the ﬁrst four years after the latter has entered into the market. Scenarios C and D assume a fuel taxes of 3% and 5% dependent on vehicle usage.

As seen inFig. 6, government subsidies in scenarios A and B promote the initial adoption of HEVs but their market share

falls immediately to a normal level when the subsidies end. Although the policies under scenarios C and D result in a smaller market share during the ﬁrst four years compared to scenarios A and B, the share continues to grow, with D shows emerging

as the most effective promoter of HEV penetration. This followsMorrow et al. (2010)who concluded that fuel taxes generate

the greatest reductions in CO2emission by increasing the cost of driving.

Considering the impact the Internet-WOM on HEV, if the market lacks these, the level of inﬂuence from strong and weak ties are 58.6 and 35.8. About 71% of new and used vehicle shoppers use the Internet to shop for vehicles and 33% of shoppers

review the evaluations of speciﬁc types of vehicle from other shoppers (PolkView, 2011). Thus, the probability of receiving

WOM through the Internet can be estimated as 0.234; i.e. for markets with Internet-WOM, the probability of receiving WOM through the Internet is 0.234 with an inﬂuence level of 53.66.

As depicted inFig. 7, the HEV market share can be increased by 5–6% with Internet-WOM compared to the market share

without Internet-WOM. The difference in market share is estimated to be 85 to 10% with further increases in gasoline prices. So far we have dealt with the positive WOM influence on the market share of HEVs, but there can also be negative implica-tions. Three sceneries are investigated. First we assume a social network without negative WOM, and then two scenarios with negative WOM spreading via strong and weak ties and social media. Given the survey showed negative information is 1.34 times more influential than of positive information, we assume that 5% of adopters are dissatisfied with HEVs and

spread negative information, with the remainder providing positive information.8_{In addition we assume that the}

probabil-ities for the adopters spreading positive and negative information are one and 0.12.

Fig. 8shows that the market with negative WOM via social network is the one least inﬂuenced by WOM, suggesting that

when there is negative WOM, social media has more inﬂuence in spreading negative information than social network. The diffusion of negative WOM leads a smaller market share. Since negative WOM via social networks is less inﬂuential, the dif-ference in market shares between markets with and without negative WOM range between 1% and 6%, although HEVs share is decreased by about 25% due to the negative WOM via social media.

5. Conclusions

We formulated a dynamic marketing model and explored the market share and penetration of HEVs in Taiwan. When the price of gasoline rises, consumers are more concerned with the fuel cost savings associated with HEVs and less with the in-creased price of the HEV, resulting in an inin-creased HEV market share. However, the impact of gasoline price on the HEV mar-ket becomes negligible if the HEV is priced beyond an acceptable level. The evidence indicates that Internet-WOM increases the market share of HEVs, but when there is negative WOM, the Internet is more inﬂuential media than social networking in spreading negative information. The implementation of additional fuel taxes is more effective for promoting HEVs than pro-viding a subsidy for disposing of old vehicles when purchasing a HEV. However, the government in Taiwan should pay atten-tion to any public backlash when implementing a fuel tax policy.

0 10 20 30 40 50 60 0 1 2 3 4 5 6 7 8 9 10

market share (%)

Benchmark Scenario A Scenario B Scenario C Scenario D

Fig. 6. Simulation results under different government subsidy policies.

8

(8)

References

Axsen, J., Mountain, D.C., Jaccard, M., 2009. Combining stated and revealed choice research to simulate the neighbor effect: the case of hybrid-electric vehicles. Resource and Energy Economics 31, 221–238.

Bettman, J.R., Park, C.W., 1980. Effects of prior knowledge and experience and phase of the choice process on consumer decision processes: a protocol analysis. Journal of Consumer Research 7, 234–278.

Brownstone, D., Bunch, D.S., Train, K., 2000. Joint mixed logit models of stated and revealed preferences for alternative fuel vehicles. Transportation Research Part B 34, 315–338.

Dodds, P.S., Watts, D.J., 2005. A generalized model of social and biological contagion. Journal of Theoretical Biology 232, 587–604.

East, R., Hammond, K., Wright, M., 2007. The relative incidence of positive and negative word of mouth: a multi-category study. International Journal of Research in Marketing 24, 175–184.

Energy Information Administration, 2011. Annual Energy Outlook. <http://www.eia.gov/forecasts/aeo/index.cfm>.

Erdem, C., Sßentürk, _I., Sßimsßek, T., 2010. Identifying the factors affecting the willingness to pay for fuel-efﬁcient vehicles in Turkey: a case of hybrid. Energy Policy 38, 3008–3043.

Ewing, G., Sarigollu, E., 1998. Assessing consumer preferences for clean-fuel vehicles: a discrete choice experiment. Journal of Public Policy and Marketing 19, 106–118.

Gallagher, K.S., Muehlegger, E., 2011. Giving green to get green? Incentives and consumer adoption of hybrid vehicle technology. Journal of Environmental Economics and Management 61, 1–15.

0 10 20 30 40 50 60 70 80 90 100 0 2 4 6 8 10 12 14 16 18 20

marekt share (%)

w/o internet wom w/ internet wom

Fig. 7. Simulation results with and without Internet-WOM.

-100 -50 0 50 100 150 200 250

WOM

w/o N-WOM w/N-WOM fm I w/N-WOM

Word of mouth

0 10 20 30 40 50 60 70 80 90 100 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20

market share (%)

w/o N-WOM w/N-WOM fm I w/N-WOM

Market share

(9)

Goldenberg, J., Libai, B., Moldovan, S., Muller, E., 2007. The NPV of bad news. International Journal of Research in Marketing 24, 186–200.

Iragüen, P., Ortúzar, J.d.D., 2004. Willingness-to-pay for reducing fatal accident risk in urban areas: an Internet-based Web page stated preference survey. Accident Analysis and Prevention 36, 513–524.

Lave, L.B., MacLean, H.L., 2002. An environmental-economic evaluation of hybrid electric vehicles: Toyota’s Prius vs. its conventional internal combustion engine Corolla. Transportation Research Part D 7, 155–162.

Marell, A., Davidsson, P., Gärling, T., Laitila, T., 2004. Direct and indirect effects on households’ intentions to replace the old car. Journal of Retailing and Consumer Services 11, 1–8.

Ministry of Transport and Communications, 2008. <http://www.motc.gov.tw/mocwebGIP/wSite/ct?xItem=4300&ctNode=553&mp=1>.

Morrow, W.R., Gallagher, K.S., Collantes, G., Lee, H., 2010. Analysis of policies to reduce oil consumption and greenhouse-gas emissions from the US transportation sector. Energy Policy 38, 1305–1320.

Potoglou, D., Kanaroglou, P.S., 2007. Household demand and willingness to pay for clean vehicles. Transportation Research Part D 12, 264–274. PolkViews, 2011. The Role of the Internet in the new and used vehicle purchase process.<https://www.polk.com/knowledge/polk_views/

role_of_internet_in_new_and_used_vehicle_purchase_process>.

Roorda, M.J., Carrasco, J.A., Miller, E.J., 2009. An integrated model of vehicle transactions, activity scheduling and mode choice. Transportation Research Part B 43, 217–229.

Toyota, 2011. <http://www.toyota.com>.

TWINC, 2009. <http://www.twnic.net.tw/index4.php>.

Wardman, M., 1998. A comparison of revealed preference and stated preference models of travel behaviour. Journal of Transport Economics and Policy 22, 71–91.