2012 International Conference on Traffic and Transportation Engineering (ICTTE 2012)
IPCSIT vol. 26 (2012) © (2012) IACSIT Press, Singapore
Forecasting Market Share of Innovative Technology by Integrating
Diffusion Model with Discrete Choice Model
Duk Hee Lee1, Jong Wook Kim2, Sang Yong Park2+∗
1
Department of Management Science, Korea Advanced Institute of Science and Technology, 373-1,
Guseong-dong, Yuseong-gu, Daejeon 305-701, Republic of Korea
2
Energy Policy Research Department, Korea Institute of Energy Research, 71-2, Jang-dong, Yuseong-gu,
Daejeon, 305-343, Republic of Korea
Abstract. The Korean government voluntarily announced on November 17, 2009 its target to reduce green
house gas emissions by 30% compared with BAU (Business As Usual) by 2020. This 30% reduction target
can be accomplished only when actively adopting innovative technologies which can reduce green house gas
emission dramatically. However the future market of innovative technology is unclear. Most previous studies
have been conducted by qualitative methodologies which rely on expert survey results and the reliability of
forecasting results depends on subjective views of respondents. Hence this study suggests new quantitative
forecasting methodology by integrating diffusion model with discrete choice model and apply developed
model to green car technologies. The empirical study can verify the validity of suggested model and will
offer meaningful insights in regard to the market penetration process of innovative technologies.
Keywords: market penetration forecasting, green car, diffusion model, discrete choice model
1. Introduction
Korea established green growth as its new national development paradigm to help solve the problems of
global warming, the energy crisis, etc. and to create a new growth engine and it has been investing in various
efforts to accomplish this new goal (Presidential Commission on Green Growth [1]). As a part of these
efforts, the government voluntarily announced on November 17, 2009 its target to reduce green house gas
emissions by 30% compared with BAU (Business As Usual) by 2020. The development and distribution of
green car technology that can drastically reduce green house emissions in the transport sector will be the key
to accomplishing the green house gas reduction target.
This study attempts to envisage the future of transportation by developing the model which can forecast
the market penetration process of alternative green car technologies according to a dynamic change of factors
such as fuel efficiency, feedstock cost, vehicle price and degree of infrastructure development that affects the
vehicle consumer utility. To consider how the dynamic change, mostly from green car technology, impacts
competitive relationships among the technologies, chapter 2 studies the two key methodologies for market
forecasting model development. Chapter 3 presents the developed market penetration forecasting model
developed using the methodologies presented above and chapter 4 deducts the empirical result of the market
penetration forecasting of green car technologies using develop model. Finally, chapter 5 summarizes the
significance of this study and future study directions.
2. Methodologies
+
Corresponding author. Tel.: +82-42-860-3037, fax: +82-42-860-3135
E-mail address: gspeed@kier.re.kr
1
2.1. Diffusion Model
This study forecasts the total green car market size using the Bass diffusion model. The Bass diffusion
model, proposed by Bass in 1969, is the most widely used diffusion model. It defines the probability of a
consumer adopting a new product at time T in terms of the innovator group’s adoption probability and
imitator group’s adoption probability, called the innovation factor p and imitation factor q, respectively (Bass
[2]). While the adoption probability of the innovator group remains the same as the innovation factor during
the prediction period, the adoption probability of the imitator group is determined as the product of the
imitation factor and cumulative adopter rate. Defining the innovation factor as p, imitation factor as q,
density function of product adoption at time T as f(T), cumulative density function of product adoption at
time T as F(T), the conditional probability of a consumer not adopting the product until time T to select the
product at time T can be presented in the following mathematical form:
(1)
2.2. Discrete Choice Model
To estimate the market share of each green car technology, this study considers the impact of factors like
fuel price, fuel efficiency, vehicle price and degree of infrastructure development on purchase probability of
the product. Since the consumers make the choice to maximize utilization after considering various product
attributes, the result of consumer choice is mainly the categorical variable with the discrete value. The
discrete choice model is a statistical analysis method used when the dependent variable is a qualitative
variable or categorical variable. In the discrete model, the multinomial logit model is the most widely used
because of ease of estimation. Like the discrete choice model, the multinomial logit model is based on the
random utility model consisting of the deterministic term, which is the observation of the consumer utility
and statistical stochastic term. The random utility model of the multinomial logit model can be expressed as
follows (Train [3]):
(2)
Here, n means the consumer, i means the alternative, x means the product attribute vector, and β means
estimation coefficient vector. In this case, the probability of a consumer n choosing an alternative j can be
expressed as follows:
,
,
(3)
Since the error term εni in the above equation is a probability variable, the model will vary according to
which probability distribution is assumed. As the multinomial logit model assumes the independent and
homoscedastic type I extreme value distribution density function, the choice probability equation after a
deduction like Train’s (2003) can lead to the following equation.
′
∑
∑
(4)
′
In this study, the estimation coefficient vector estimated with the ranking ordered logit model developed
by the Korea Energy Economics Institute in 2008 was applied to develop the green car market penetration
forecasting model.
2
3. Model
To propose an integrated model considering all models, including the Bass diffusion model, discrete
choice model, repurchase, and feedback, this study defined the state transition matrix as follows:
Pijt = Probability of a consumer owning i type of vehicle purchasing a j type of vehicle at time t
In this study, the integrated forecasting model was developed by defining each element of the above
state transition matrix using the diffusion model and discrete choice model with some assumptions.
Assumption 1. The probability of a consumer owning an internal combustion vehicle who is purchasing a
green car conforms to the Bass diffusion model.
1
(5)
p = Innovation factor
q = Imitation factor
Y(t) = Number of accumulated customers owning a green car at time t
m = Potential market size of passenger vehicles
Assumption 2. Selecting which type of green car will conform to the discrete choice model.
,
∑
j = 1, 2, 3
(6)
(7)
(8)
∑
Ui(t) = Utility of i type of green car at time t
Vi(t) = Quantitative part of utility of i type of green car at time t
εi(t) = Probabilistic part of utility of i type of green car at time t
αk = Weight factor of the attribute k of a green car
Xik(t) = Value of the attribute k of i type green car at time t
This study considered following three attributes for green car technologies.
k = 1; fuel efficiency
k = 2; vehicle price
k = 3; fueling/recharging infrastructure
4. Empirical Study
4.1. Data
For the diffusion model, the innovation coefficient, imitation coefficient, and data on the potential
market size are needed. In this study, the coefficient value, which is estimated in Park’s study was used in
order to forecast the overall market penetration curve of the green car (Park [4]). To apply the discrete choice
model to forecast market penetration of the green car technology, the estimated weight factor (αk) of the
impact of each attribute of fuel efficiency, fuel cost, vehicle cost and level of infrastructure on consumer
utility and predicted attribute value (Xik(t)) of each green car technology in each year are needed. First, the
3
study data by
b Korea Eneergy Econom
mic Institute (KEEI) weree used for thee weight facttor of each attribute.
a
Thee
KEEI studyy estimated the weight factor of atttributes like the fuel, vehicle type, fuel efficiency, fuelingg
accessibilityy, and purchhase price ussing the conjjoint analysiis. Next, the prediction oof the techniical attributee
value in eacch year appllied the studdy result from
m the Korea Institute off Energy Tecchnology Evaluation andd
Planning (K
KETEP) in 20011. KETEP
P performed a long-term prediction
p
thhrough the yeear 2050 of key
k technicall
attributes suuch as the uttilization ratee, consumed energy, effiiciency, inveestment cost and mainten
nance cost off
15 green ennergy technollogies such as
a solar cells,, wind energy, fuel cells and green caars based on the opinionss
of the mosst renowned experts in each area. However, since the preediction of fueling acceess for eachh
technology was not perrformed in thhe KETP stuudy, this stud
dy assumed that the rechharging infraastructure off
HEV was 100% by using
u
the exxisting rechaarging infrasstructure, whhile the scennario of thee rechargingg
infrastructure of EV andd HFCV lineearly increassing beginnin
ng in 2010 and reaching 100% in 2050 was usedd
as the baseline scenario for the empiirical study. Furthermoree, the price prediction
p
of each energy
y source usedd
as the basicc assumptionn of the long--term energyy demand forrecasting stuudy performeed by KEEI in 2010 wass
used for thee fuel price.
4.2. Resuults
The basseline scenarrio indicatess that only HEV
H
enters the market early
e
to havve the largest number off
distributed vehicles
v
by around
a
2035, while EV and
a HFCV will
w begin to appear in the market in around
a
20300
and will reeach the accuumulated nuumber of disstributed veh
hicles similaar to that off HEV by arround 2050..
However, experts
e
prediict that the consumer
c
uttility of EV will be highher than thatt of HFCV and
a thus thee
market sharre of EV willl be slightly higher
h
than that
t of HFCV
V.
Figurre 1. Accumullated Number of Green Carss Distributed – Baseline Scenario
However, the analyssis of the accumulated number
n
of veehicles distriibuted does not give a clear
c
idea off
which technnology is supperior per yeear. The analyysis of the diiffusion of each green caar technology
y in terms off
the year-by--year markett share indicates that the market sharre of HEV will
w graduallyy decrease frrom 100% inn
2010, and thhen the decreease will acccelerate from
m around 2030 when the distribution
d
oof EV and HFCV
H
will bee
more activee. EV and HF
FCV are expeected to enteer the markett around 2015 and 2023, respectively
y. Near 2037,,
market sharres of both EV and HFCV
V will surpasss that of HE
EV to becomee the leadingg green car teechnologies.
4
T
– Baseline Scennario
Figuure 2. Market Share of Eachh Green Car Technology
5. Concllusions
This studdy developedd market pennetration forecasting mod
del to analyzze the impacct of the dynaamic changee
of technoloogy attributess on the maarket share among
a
the competitive technologies
t
. The baseliine scenario,,
which appllied the techhnology devvelopment prediction
p
by
y the experrts, predictedd that whilee HEV willl
completely dominate the green car market
m
initiaally, three greeen car technnologies of H
HEV, EV and
d HFCV willl
o accumulatted sales vollume by arou
und 2050 as the EV andd HFCV tech
hnologies aree
have a simiilar number of
developed. This study developed
d
a new
n demandd forecasting
g methodologgy by effectivvely integratting the basss
m
and disccrete choice model.
diffusion model
In next study,
s
variouus empirical studies can be
b conducted
d by generatting scenarioos related witth long-term
m
forecast of technology
t
a
attributes
likee the fuel priice, fuel efficciency, and vehicle
v
price,, which affecct the markett
share. The application
a
o the developed model too the green car
of
c area can also
a lead to a significant result in thee
R&D and infrastructurre developm
ment viewpoiint that can be used ass a referencee for successsful markett
g
car techhnologies.
penetration strategy of green
6. Acknoowledgmeent
This stuudy was suppported by thee HERC(Hyydrogen Enerrgy R&D Ceenter). The auuthors thank
k anonymouss
reviewers foor their helpfful commentts and suggesstions.
7. Referrences
[1] Presidenntial Commission on Greenn Growth, Roaad to Our Futu
ure: Green Grrowth, Presideential Commisssion on
Green Growth,
G
2009.. (www.greenngrowth.go.kr))
[2] F.M. Baass, A new product growth for model connsumer durables, Managem
ment Science 115 (5) (1969) 215-227.
2
[3] K. Trainn, Discrete chhoice method with
w simulatioon, Cambridgee University Press,
P
Cambriddge, 2003.
[4] S.Y. Paark, J.W. Kim,, D.H. Lee, Deevelopment of a market pen
netration foreccasting modell for Hydrogen
n Fuel Cell
Vehiclees consideringg infrastructuree and cost redduction effectss. Energy Policcy 39 (2011) 33307-3315.
5