Applied Energy 142 (2015) 33–43

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Optimal subsidy estimation method using system dynamics

and the real option model: Photovoltaic technology case
Chanwoong Jeon a, Jeongjin Lee b, Juneseuk Shin c,⇑
a
Graduate School of Management of Technology, Sungkyunkwan University, 2066, Seobu-Ro, Jangan-gu, Suwon, Gyeonggi-do 440-746, Republic of Korea
b
School of Computer Science & Engineering, Soongsil University, 369 Sangdo-Ro, Dongjak-Gu, Seoul 156-743, Republic of Korea
c
Department of Systems Management Engineering, Sungkyunkwan University, 2066, Seobu-Ro, Jangan-gu, Suwon, Gyeonggi-do 440-746, Republic of Korea

h i g h l i g h t s

 We propose a method of estimating optimal financial subsidy and public R&D investment.
 System dynamics is used to analyze a complex system of energy subsidies.
 Appropriate real option valuation models are matched and used to estimate optimal subsidies.
 Our approach can reduce the total Korean photovoltaic subsidy by $US 359.5 million.

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, we propose a method of optimizing financial subsidies and public research and develop-
Received 21 May 2014 ment investments for renewable energy technologies, rather than optimizing financial subsidy alone.
Received in revised form 20 November 2014 By combining system dynamics with real option models, we capture dynamic complex interactions
Accepted 27 December 2014
among investors, consumers, and policymakers, as well as future uncertainties of key energy, economic,
and environmental factors. Our method thereby makes subsidy optimization more accurate and flexible.
To evaluate our model, we apply it to the Korean photovoltaic subsidy and determine that the govern-
Keywords:
ment can achieve the photovoltaic diffusion target, even if it reduces the total subsidy by $US 359.5
Optimization
Subsidy
million. The optimal approach is to increase research and development funding by $US 310.4 million
System dynamics while reducing the financial subsidy by $US 669.9 million. Our method helps policymakers optimize their
Real option subsidy allocation and therefore reduces subsidy inefficiencies.
Photovoltaic Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction replaced [5]. However, renewable energy technologies require

more than a decade to attain grid parity [6]. Moreover, an individ-
Public subsidies have been primary policy instruments for ual firm does not expect positive externalities from a learning-
spurring adoption and progress of renewable energy technologies. by-doing approach, which therefore constrains investments in
Global renewable energy subsidies have been steadily increasing; renewable technologies [5]. Thus, a market failure will occur with
in 2011, they reached US $88 billion [1]. Since 2000, renewable investments below the socially optimal level [7]. Finally, lock-in
energy technologies have received higher subsidies for electricity effects and incentives to continue exploiting the existing energy
generated than fossil fuels or nuclear energy [2]. Through these infrastructure hinder efforts to replace legacy energy technologies
generous subsidies, several countries, including Germany, the [8]. Without some policy rents, such as subsidies, renewable
United States, China, and Japan, have increased electricity energy technologies must be developed and adopted far slower
production from renewable technologies [3,4]. than is socially needed.
There are several reasons why renewable energy technologies For renewable energy technologies, a large proportion of subsi-
have been subsidy-driven. Given the risk of environmental degra- dies has been provided through a feed-in tariff (FIT), production tax
dation and energy insecurity, unsustainable technologies must be credit (PTC), and research and development (R&D) funding [2].
Several studies have examined the effects of these subsidies on
⇑ Corresponding author. Tel.: +82 31 290 7607; fax: +82 31 290 7610. the diffusion of renewable energy technologies and determined
E-mail addresses: [email protected] (C. Jeon), [email protected] (J. Lee),
that FIT is more efficient and effective than alternative policies
[email protected] (J. Shin). [9–12]. However, the adoption of renewable energy technologies

http://dx.doi.org/10.1016/j.apenergy.2014.12.067
0306-2619/Ó 2015 Elsevier Ltd. All rights reserved.
34 C. Jeon et al. / Applied Energy 142 (2015) 33–43

could not reach the given target when the financial subsidy alone [27], and R&D investments [28,29]. Further, recent studies have
was implemented. This implies that financial subsidies should be evaluated the overall technology investment, including subsidy
used together with R&D funding [13]. At a minimum, subsidy opti- and infrastructure investments [30,31]. Considering various mar-
mization should consider both the financial subsidy and R&D ket and policy uncertainties, these studies avoid the over- and
funding. underestimation of subsidies by DCF and therefore improve the
In response to tight national budgets and broad impacts of glo- accuracy of estimation.
bal low economic growth, many countries, including Germany and Recently, under conditions of global low economic growth and
Italy, have significantly reduced FIT rates for renewable energy fiscal burdens, the necessity of subsidy optimization has been
technologies [3]. In addition, public R&D investments in solar increasing for many governments, which warrants more advanced
power, wind, biomass, and biofuels have decreased in most devel- approaches [14–16]. Some researchers have used a consumer
oped countries, including the United States, Germany, and Japan choice model to forecast future PV demands and estimate the min-
[3]. Under severe budget constraints, many governments are seek- imal subsidy for achieving the policy adoption of targets [7]. This
ing new ways to meet increasing policy targets of development optimization approach has the advantage of considering uncertain-
and deployment of renewable energy technologies and to maxi- ties in consumer demand and PV technology development, as well
mize the benefits of these technologies, including reduction of as of maximizing the overall social value conferred, including con-
greenhouse gas emissions, enhanced energy security, and sumer, investor, and policymaker benefits and costs.
economic growth [14–16]. Nevertheless, governments intend to Other researchers have suggested optimization models to
optimize energy subsidies at any given moment. include important but hard-to-measure factors, including different
Among subsidy estimation methods, the discounted cash flow investor attitudes toward market risk [32,33], different subsidy
(DCF) method is widely used. It is appropriate for short-term esti- designs [34], and network effects [18]. Using option valuation
mation with certainty; however, it is not appropriate for long-term models, these studies account for different exposures and attitudes
estimation with severe uncertainties [17]. Instead, researchers sug- of investors, consumers, and policymakers to market, technology,
gest optimization approaches, including the real option model [17], and policy risks. Accordingly, they make optimization more
consumer choice model with the learning-by-doing effect [7], and realizable.
the consumer choice model with the option valuation model [18]. Despite their contributions, previous studies do not consider
These approaches have the advantage of reflecting various factors interactions among multiple stakeholders, which influence factors
and their uncertainties; nevertheless, these models cannot con- and uncertainties. For instance, PV module manufacturers decide
sider their interactions. Additionally, the appropriate option model investment levels for plants, which affect PV pricing and consumer
varies with the type of investment, which implies that the most adoption. Our SD model provides a means of analyzing their inter-
appropriate option model exists for each subsidy type [19,20]. actions as well as circular causalities; it therefore traces dynamic
To address these issues, we propose a new method to determine changes of key variables and their effects on subsidies. Our model
the optimal amount of financial subsidy and R&D funding for can make dynamic subsidy optimization more reliable and identify
renewable energy technologies. System dynamics (SD) enables this time-varying effects of key variables on subsidies as well as other
method to reflect various factors as well as their interactions. Fur- variables. Moreover, subsidy planning can become more accurate
ther, we match the most appropriate real option model to each and flexible.
type of subsidy and thereby increase the optimization accuracy. In addition, there has been no effort toward multi-subsidy opti-
By leveraging the advantages of both methods, our approach can mization. For example, many previous studies have improved ways
improve the optimal subsidy estimation for renewable energy of optimizing FIT; however, they do not try to optimize more than
technologies. As an example, we estimate the optimal financial two types of subsidies together. Financial subsidies and R&D fund-
subsidy and R&D funding for Korean photovoltaic (PV) technology. ing can create a synergy for facilitating consumer adoption. Thus,
the optimal subsidy for either one is not optimal but sub-optimal.
Our model provides a way of optimizing two types of subsidies
2. Literature review: renewable energy subsidy optimization together; it therefore can be a first step toward multi-subsidy opti-
mization. Finally, there has been little effort at identifying the most
Given the policy target, the appropriate level of subsidy is deter- appropriate subsidy optimization model for a specific subsidy pol-
mined in practice by using DCF. DCF estimates and sums all icy with different objectives, benefits, and uncertainties. Our model
expected future cash flows to set the amount of subsidy regardless strives to match an appropriate model to a specific subsidy.
of different subsidy design and implementation options [21,22].
DCF is suitable for short-term subsidy estimation with little uncer-
tainty; however, it is not appropriate for strategic investment, such 3. Methodology
as renewable energy subsidies affected by severe uncertainties
[23]. A wide array of uncertainties exist, including advancing com- 3.1. Research framework
petitive energy technologies, constricting budget reductions, stron-
ger environmental regulations, infrastructure accidents, and As noted above, renewable energy technology subsidy optimi-
others. Moreover, various other factors amplify uncertainties and zation must consider: (1) complex interactions among various fac-
make DCF inappropriate; these include subsidies, generation costs, tors, (2) dynamic uncertainties and their influences, and (3)
and environmental regulations, which interact in complicated appropriateness of an optimization model for a specific subsidy.
ways. Researchers conclude that other methods are better than Most deterministic valuation and optimization methods, including
DCF in terms of robustness and accuracy [24,25]. DCF and the consumer choice model, can partially meet the third
The real option approach is a recommended alternative. It can requirement but cannot meet the others. The real option approach
reflect various factors and their uncertainties in estimation; there- was proven to be appropriate for the second requirement; it can
fore, it is more effective than DCF [26]. Davis and Owens [17] esti- also provide various subsidies with well-matched optimization
mate the renewable electric R&D expenditure from both the real models. SD was initially developed to address the dynamics of
option and DCF perspective; they demonstrate that the real option complex systems characterized by uncertainties and sophisticated
is more effective than DCF. Using real option models, researchers interactions; it therefore satisfies the first and second require-
have attempted to estimate financial subsidies, including FIT ments. By combining SD with appropriate real option models, we
 C. Jeon et al. / Applied Energy 142 (2015) 33–43 35

can develop an estimation method of optimal renewable energy complex systems in which many variables show dynamic non-lin-
subsidies that satisfy the three key requirements. earity through complex interactions.
SD can be either deterministic or probabilistic. Dhawan [35] Renewable energy technologies are affected by various interact-
conducted performance tests of deterministic SD, SD with sensitiv- ing factors, including government policies, supply–demand condi-
ity analysis, and probabilistic SD. Probabilistic SD can represent the tions, the cost per unit of electricity, and subsidies. Using SD, we
real behavior of complex systems more effectively than the others. can therefore represent complex dynamic behavior of a renewable
Among probabilistic SD models, previous studies have shown that energy technology system, estimate its volatility, and consequently
SD with Monte Carlo simulation is useful, particularly in investi- analyze the way it affects subsidies.
gating uncertainties [36,37]. Considering our methodological
requirements as noted above, we adopt SD with Monte Carlo 3.3. Real option models
simulation.
Our method consists of three modules: (1) a renewable energy The real option approach employs the concept of financial
technology system model, (2) a volatility estimation model of over- options and values opportunities over uncertainties [43]. DCF
all system and key variables, and (3) an optimal subsidy estimation assumes that a firm will embark on an inflexible path forward
model. In the system, internal and external variables influence and ignore any uncertainties and options embedded in the oppor-
future returns and risks of a renewable energy technology, thereby tunity. For instance, the net present value (NPV) of a project cannot
creating interactions in complicated ways. We review several consider the values of options to delay, expand, or abandon it. In
energy frameworks and models and select key variables that influ- contrast, the real option approach assumes that firms continuously
ence financial subsidies and R&D funding. Using SD, we then repre- respond and adjust to changes and uncertainties that may occur in
sent the dynamic causal relationships among the variables. A the future. It therefore captures the values of options and thereby
renewable energy technology system model is constructed, which confers strategic flexibility.
reflects complex interactions among the key variables. Monte Carlo Amram and Kulatilaka [44] organized various real option
simulation makes the SD model probabilistic and thus enables esti- models in three categories comprising: (1) a partial differential
mation of volatility of both an individual variable and the overall equation, (2) dynamic programming, and (3) simulation. The first
system. With volatility estimates, an appropriate real option model approach expresses the value of an option and its dynamics by a
is matched to a specific subsidy. The optimal subsidy is then esti- partial differential equation; it then develops analytical solutions.
mated to achieve the policy target while reducing the fiscal burden The most basic approach, the Black–Scholes model [45] produces
as much as possible. The modules, processes, and methods are a closed-form solution for the value of European options on a
shown in Table 1. non-dividend-paying opportunity. Several researchers have
improved this approach; their models include Roll–Geske–Whaley,
Barone-Adesi–Whaley, and others [46,47].
3.2. System dynamics The above models are useful for simple-structured opportuni-
ties; however, they are inadequate in addressing opportunities that
In the 1950s, Forrester developed SD to analyze the core issues have multiple flexible options and decisions. The dynamic pro-
that determine the success or failure of corporations. It was ini- gramming approach makes intermediate decisions, values, and
tially applied exclusively to managerial problems; however, its options visible; it therefore clarifies complex decision structures.
applications have broadened to include urban dynamics, global In the renowned binominal tree pricing model [48], the binominal
socio-economic systems, and others. In particular, it has become tree represents different possible paths of an option’s probabilistic
a modeling tool of energy systems characterized by uncertainties, value for discrete time intervals over its lifetime. It thereby esti-
circular causalities, and complex interactions [38,39]. A number mates the value of an option at any given time interval. Several
of SD energy policy models have been developed, including Ford variations also exist, including trinomial trees, the accelerated
[40] and Naill [41]. However, there has been minimal focus on binomial option pricing model, and others [49,50].
investigating subsidies using SD models. However, these models are inappropriate for real-life opportu-
To date, what makes SD unique is its stock-flows-feedback nities with many flexible options. A simulation approach generates
structure. It enables researchers to describe circular causalities thousands of random value paths of an opportunity from the pres-
and mutual interactions among various factors in a systematic ent to the final decision date in an option, calculates the payoff at
way; accordingly, it explains how systems display non-linearity the end of each path, and thus identifies the optimal investment
[42]. Further, SD can address interlocking and time-delayed rela- strategy. It can handle opportunities characterized by complicated
tionships among factors, thereby overcoming the weakness of lin- decisions and complex relationships between the option value and
ear causal models, such as regression. It is one of the most opportunity. Monte Carlo simulation is one of the frequently used
appropriate methods for analyzing the dynamic behavior of methods.

Table 1
Research framework.

Module Process Method

Renewable energy technology system (1) Conceptual framework review Literature review
(2) Key variable identification System dynamics
(3) Causal loop diagram analysis
(4) Stock-flow diagram analysis
(5) System dynamics model validation
Volatility estimation (1) Key uncertainty identification Monte Carlo simulation
(2) Monte Carlo simulation
(3) Volatility estimation of key individual uncertainties
(4) System volatility estimation
Optimal subsidy estimation (1) Appropriate option model identification for each subsidy Real option models (Black Scholes & Geske compound option)
(2) Optimal subsidy estimation
36 C. Jeon et al. / Applied Energy 142 (2015) 33–43

In reality, many opportunities can be regarded as a sequence of exacerbated the situation. To alleviate the fiscal burden, some gov-
real options; that is, as compound options. Geske [51] presented a ernments have significantly cut financial subsidies and R&D fund-
method of pricing an option on an option that is a twofold com- ing for renewable energy technologies. Further, they have
pound option. This method is useful; however, it has difficulty val- transitioned from FIT to a renewable portfolio standard, which
uing opportunities of more than two phases. To resolve this issue, forces suppliers to deliver a fixed proportion of electricity from
researchers have suggested new models and solutions, including renewable energy sources [60].
the generalized n-fold compound option model [19,52], binominal These changes can lead to a vicious cycle, as shown in Fig. 1,
lattice [53], Monte Carlo simulation [54], and others. which begins from the reduced policy incentives for renewable
In the energy sector, several previous studies have used the technologies. Some governments focus on renewable energy tech-
Black–Scholes and relevant partial differential equation models nologies of lower costs and less pollution and reduce budgets for
to evaluate relatively inflexible projects characterized by a limited less competitive renewable energies. Rationalization is a key driver
number of options, including offshore oil investments [55], power of reduced incentives and is boosted by low economic growth.
generation investments [56], tariff policy evaluations [57], and oth- With less expected revenue, companies will reduce investments
ers. However, with increasing competitiveness and the complex for PV power system installation, thereby slowing PV market
processes of regulation, deregulation, and technological innova- growth. Fewer PV system installations will result in less R&D and
tion, the uncertainties in the energy sector have proliferated to capital investments by suppliers, including module producers
an unprecedented level. More sophisticated real option models and system integrators, which will impede PV technology develop-
[58] have therefore been needed. In particular, renewable energy ment and cost reduction. PV technology requires a longer time to
technology projects and policies affected by these features have reach grid parity than expected; it thus cannot meet the policy tar-
been evaluated by dynamic programming [59] and dynamic pro- get to reduce the percentage of carbon dioxide emissions. Without
gramming with partial differential equations [29]. additional policy incentives, governments cannot use other renew-
able energy technologies to compensate for the deficiency because
4. Empirical analysis and results PV policy incentives have already been spent. Moreover, to com-
pensate for the immediate deficiency of energy production, gov-
4.1. PV industry background: vicious vs. virtuous cycles ernments depend on fossil fuels rather than renewable energies
because of the low cost advantage. Therefore, energy dependency
Accelerating PV market growth has been primarily due to gov- on fossil fuels will not be significantly reduced, thereby making
ernment policies, including FIT and subsidies. Driven by long-term countries vulnerable to energy insecurity, such as oil shocks. Seek-
cost-based compensation policies, private companies have decided ing solutions, some governments are likely to increase public sup-
to enter the PV market and have continuously expanded PV gener- port for low-carbon and low-cost alternative energy, such as
ation capacity. As long as these policies are in place, they can pre- natural gas. Energy security and environmental externalities affect
vent deficits while reducing the PV cost per unit of electricity by investment decisions in alternative energies [61,62]. In cases in
R&D and economies of scale. Then, by attaining grid parity, they which alternative energies are specified, energy-specific externali-
will produce large profits. However, PV costs have declined more ties must be added in the cycle. Policy incentives for PV will be fur-
slowly than expected, which has caused anxiety for companies. ther reduced, and the vicious cycle will be further established.
Some companies have withheld additional PV investments or have The primary cause of this cycle is the PV cost per unit of elec-
even withdrawn from the market. Global low economic growth has tricity, which is higher than that of other fossil fuel technologies.

Fig. 1. Vicious cycle of PV technology system.

 C. Jeon et al. / Applied Energy 142 (2015) 33–43 37

Fig. 2. Virtuous cycle of PV technology system.

However, if PV becomes competitive with oil, coal, and natural gas public energy policies and strategic decisions of private energy
in terms of cost, its adoption will increase, even with marginal pol- companies. Causalities are not linear but circular.
icy support. A higher level of adoption will lead to increasing R&D Focusing on PV technology, we elaborate on such common cau-
investments and more learning effects, reduce PV costs below sal relationships as a causal loop, which is illustrated in Fig. 3. Con-
other energies, and thus accelerate its diffusion. It has been shown sidering that policy targets of PV electricity production are affected
that PV costs can be steadily reduced on account of R&D invest- by GDP and current levels of carbon dioxide emissions, govern-
ments, process innovations, and capacity expansion [63]; a virtu- ments determine policy incentives. R&D investments and cumula-
ous cycle, as shown in Fig. 2, could be in place. Obviously, R&D tive PV system installations drive down module costs as well as
investments can trigger this cycle. other generation costs, thereby reducing the PV unit cost per elec-
Governments are seeking ways to create this virtuous cycle; tricity. On account of these factors, power generation companies
however, they cannot significantly increase public policy incen- decide their PV investments, which are affected by macro-eco-
tives. Above all, they must optimize public R&D funding to reduce nomic variables. Depreciation and indirect costs then occur. PV
PV costs before the vicious cycle begins. However, with insufficient sales depend mainly on these costs, the unit cost per electricity,
financial subsidies, many private power generation companies generation capacity, and module efficiency. The overall profit and
might withdraw from the PV market because of expected loss. value of PV technology is determined by sales, economic variables,
Governments must provide companies with the optimal amount and public incentives, which lead to changes in policies and eco-
of financial subsidies to help them remain in the PV market. Thus, nomic variables.
financial subsidies must be optimized to maximize the learning To build a probabilistic SD model, we define key variables with
effects and therefore accelerate PV cost reduction. appropriate measures. Key variables in the energy category include
PV module cost, efficiency, and capacity. These are crucial to PV
4.2. PV technology system dynamics model technology development as well as its economic attractiveness.
The Korean government has set a policy target of reducing the
As explained above, financial subsidies and R&D funding are PV module cost to 0.5$/Wp by 2015, and then to 0.3$/Wp by
important factors in the PV technology system and are directly 2030 [68]. In this regard, we assume that the PV module cost is
and indirectly affected by various variables. Some are directly con- 0.5$/Wp. Based on Korean FIT rules, PV module efficiency is
nected to subsidies; others are indirectly connected to them. The expected to decrease from 80% to 2% each year [69]. Given the pol-
optimal level of subsidy depends not only on these factors, but icy target of the annual cumulative number of new PV power sys-
on their complex interactions. Accordingly, when focusing on PV tems and total electricity generation by 2030, we set the average
subsidies, a model must be constructed to identify the key vari- PV plant size as 414,346 kW [70]. Power generation companies
ables, form circular causal relationships among them, and thus rep- have purchased lands in several regions for PV power system
resent their current and future behaviors. installations. We therefore calculate the average daily solar radia-
Reviewing several macro-level energy models, including tion of these regions and use it in our SD [71].
ENTICE [64], IMACLIM-R [65], ReMIND-R [66], and WITCH [67], Investments, costs, sales, and profits are key variables in the
the key factors can be classified into three categories: energy, econ- economy category. Investments in PV power system installations
omy, and environment. Environmental factors, such as policies and are specified in Korean FIT rules. The operation of the PV power
regulations, promote green energy technology adoption and rele- system lasts an average of 20 years; it should be more than
vant R&D investments; consequently, they affect the national 15 years, as enforced by Korean FIT rules. Module cost, efficiency,
energy mix. Economic variables, such as the gross domestic prod- and capacity are as previously assumed. Other important expenses
uct (GDP), are affected by these industrial changes and influence include construction, land purchase, and on-grid system costs [69].
38 C. Jeon et al. / Applied Energy 142 (2015) 33–43

Fig. 3. Causal loop of PV technology system.

Costs are also affected by the investment structure. A review of estimate the average rate of the cost of equity as 16.12%. Addition-
Korean PV companies reveals that their most frequently used ally, we estimate an average market risk premium of 11% and a
investment structure is 22% equity-financed and 78% debt- risk-free interest rate of 2.8%. Using these figures, we can calculate
financed. the weighted average cost of capital and use it as the discount rate.
There are no direct costs in the PV power system. Electricity is In the environment category, the Korean government strives for
produced by using only solar radiation and is immediately sold a PV electricity generation capacity of 462,698 MW h by 2015, and
to the grid. All PV plants are assumed to be directly connected to 547,042 by 2020 [68]. The R&D target is the reduction of PV module
the grid. Having no inventory costs, PV electricity cannot be stored. costs, as previously noted. Over the last several years, the corporate
Only indirect costs, including those for maintenance, occur. In tax rate has been 27.5% with minimal change. No subsidy for the
Korea, however, inflation-adjusted PV indirect costs have been construction of PV power systems is provided in Korea. However,
almost constant over the last decade and are thus assumed to once electricity is generated, governments provide generators with
remain unchanged [72]. financial subsidies determined by environmental policies, rules,
Sales are basically determined by the unit price of electricity. In and costs. In Table 2, all variables are classified and briefly
the Korean electricity market, the Korea Power Exchange deter- described with relevant references. An appropriate time lag is con-
mines the unit price of electricity as the highest cost of power from sidered, determined, and used in every causal relationship.
available producers every hour. This is called the system marginal Many tests exist to validate SD models. However, Barlas [73]
price (SMP). Coal and nuclear power are excluded because they are suggested that the behavioral validity test is sufficient only for
cheaper than the others. Over the last decade, SMP has continu- SD-based valuation models. Thus, we select four key variables
ously increased in a typical pattern of exponential curves. Thus, comprising generation capacities, investments, sales, and discount
eliminating some outliers of the 2008 economic crisis, we fit an rates and compare the model-simulated behavior with the
exponential curve to SMP data during the 2001 to 2013 period, observed behavior of the real system between 1984 and 2013.
estimate parameters, project the future SMP, and use it as the Referring to Barlas [73], we choose the root mean square percent-
future unit price of electricity. The exponential curve accounts age error (RMSPE) as a statistical validity indicator because it has
for 85.4% of variances. greater advantages in reliability, sensitivity, and outlier protection
Fluctuating macroeconomic variables can be assumed to be than the others. The values of RMSPE are 3.9% for generation capac-
constant over the long term because the average operating period ities, 4.8% for investments, 4.3% for sales, and 6.3% for discount
is 20 years. Using the regression method with the interest rate data rates. Overall, our SD model well matches with past data.
of AA corporate bonds between 2004 and 2013, we forecast the
future interest rates over the next 20 years and calculate the aver- 4.3. Volatility estimation
age as 2.93%. The inflation and exchange rates are similarly deter-
mined. However, the discount rate is set to be a variable to reflect For volatility estimation of the overall PV technology system,
corporate financial risks. Using the capital asset pricing model, we we review previous studies and identify key uncertainties
 C. Jeon et al. / Applied Energy 142 (2015) 33–43 39

Table 2
Key variables of the PV technology system model.

Category Variable Unit Description Reference

Energy Module cost $/W 0.5 KETEP, 2011
PV plant size kW 414,346 MKE, 2008
Module efficiency % Initially 80% with annual 2% reduction KERI, 2010
Daily solar radiation kW h/m2/day 3411 KIER, 2013
Generation capacity kW h/year Daily solar radiation  module capacity  module efficiency
Operation period Year 20 KERI, 2010
Economy Unit price of electricity KRW/kW h SMP (projected value by curve fitting)
Sales KRW Unit price of electricity  generation capacity
Indirect cost KRW/year 1% of total investment KERI, 2010
Investment structure % 22% equity and 78% debt
Inverter price KRW/kW 0.4 million (US$ 363.6) KERI, 2010.
Shaft price KRW/kW 0.3 million (US$ 272.7) KERI, 2010
Construction cost KRW/kW 100 million (US$ 908.9) KERI, 2010
On-grid system cost KRW/kW 0.3 million (US$ 272.7) KERI, 2010
Land purchase cost KRW/kW 0.36 million (US$ 327.2) KERI, 2010
Investment KRW Module capacity  (module cost  exchange rate + inverter KERI, 2010
price + shaft price + construction cost + land purchase cost + on-grid system cost)
Scrap value KRW 10%  investment KERI, 2010
Inflation rate % 3.12 (2004–2013 average) KOSIS, 2013
Interest rate % 2.93 (AA corporate bond, 2004–2013 average) KOSIS, 2013
Foreign exchange rate KRW/$ 1100.21 (2003–2012 average) KOSIS, 2013
Risk-free interest rate % 2.80
Cost of equity % 16.12
Discount rate % (Debit  interest rate + equity  cost of equity)/(debit + equity)
Environment Generation target MW h 462,898 (year 2015), 547,042 (year 2020) MKE, 2008
R&D target $/W 0.5 (year 2015), 0.3 (year 2030) KETEP, 2011
Tax incentive % 27.5 KERI, 2010
Subsidy KRW 0 (Initial value)

Note: KRW (Korean won), $US (US dollar), kW (kilowatt), and kW h (kilowatt hour).

Table 3
Parameters for Monte Carlo simulation.

Uncertainty Unit Distribution Mean Std. dev. Min. Max. Reference

Solar radiation (Annual) kW h/m2/year Normal 1395.86 342.48 569.40 2230.2 KIER, 2013
Interest rate % Uniform 2.93 2.25 6.35 KOSIS, 2013
Exchange rate Korean Won/$ Uniform 1100.21 771.04 1398.88 KOSIS, 2013
Risk premium % Uniform 11.00 8.50 17.30
Risk free rate % Uniform 2.80 2.13 5.93
SMP % Normal 0 8.36 31.36 15.97

comprising (1) solar radiation, (2) SMP, (3) interest rates, (4) for- standard deviation of NPV is 43.49%. Additionally, the probability
eign exchange rates, (5) risk premiums, and (6) risk-free rates. of loss is slightly above 30% in 2035, and is more than 75% in
These variables are key sources of volatilities and therefore affect 2025. Obviously, the volatility of PV technology system is large,
the value of PV technology more than others. Solar radiation and implying high uncertainties, possibility of loss, and risk of
SMP volatility data follow a normal distribution for a year and incurring the vicious cycle.
are expected to move accordingly. SMP is measured by the propor-
tion of price variations from the annual average SMP. Other vari-
ables do not move in trends over the long term and can thus be 4.4. Optimal subsidy estimation
assumed to follow a uniform distribution between historical min-
imum and maximum values according to a random walk hypothe- 4.4.1. Optimal financial subsidy: Black–Scholes model
sis [74]. Detailed information of these six variables are presented in The optimal financial subsidy can be defined as the minimal
Table 3. subsidy for achieving the policy target of PV electricity generation.
Monte Carlo simulation is used to individually estimate the vol- If the expected loss is above a tolerable level, private generation
atility of the six key variables as well as the overall PV system. We companies are likely to withdraw from the PV market, or at least
randomly generate 1000 inputs of individual variables from their to significantly reduce electricity generation. To avoid this situa-
probability distributions and calculate their variances. Solar radia- tion, government should provide companies with insurance
tion explains 61% of the total variance of NPV, thereby showing the against losses from reduced returns due to volatility; in other
importance of the PV power system location. SMP follows with the words, a put option. Moreover, the PV generation business charac-
second highest uncertainty, which explains 23%. Of the macroeco- terized by a huge initial investment is set on a rigid path with little
nomic variables, the interest rate and foreign exchange slightly strategic flexibility. Further, the Korean PV subsidy scheme has
affect PV volatility. The risk premium and risk-free rate have little been stable with little volatility and is expected to remain that
influence. As shown in Fig. 4, the NPV of the overall PV technology way [68]. These are typical characteristics of the European option.
value fluctuates from an 847 billion KRW (US$ 769 million) loss to Neither margin trading nor dividends exist. Thus, the Black–Scho-
2062 billion KRW (US$ 1.87 billion) by 2035. The normalized les model is the most appropriate model for estimating the value of
40 C. Jeon et al. / Applied Energy 142 (2015) 33–43

Fig. 5. Sensitivity analysis results.

Fig. 4. Monte Carlo simulation of PV technology value.

4.4.2. Optimal R&D investment: Geske compound option model
The value of the R&D investment is determined not only by the
a European put option on a non-dividend-paying opportunity. The value of the initial R&D investment, but by the value of future
value of a put option (Eq. (1)) is given by [45]. investment opportunities resulting from the initial investment
[20]. In the PV business, there are three sequential investment
P ¼ KerT Nðd2 Þ  SNðd1 Þ ð1Þ opportunities: initial R&D investment, follow-up R&D investment,
and PV system installation. At the time of the initial investment, a
      company has the first call option to invest in the follow-up R&D
1 S r2
d1 ¼ pffiffiffi ln þ rþ T ; d2 investment; it additionally has the second call option of the PV sys-
r T K 2
      tem installation. Once the PV power system is completed, the com-
1 S r2 pany must produce electricity for more than 15 years with no
¼ pffiffiffi ln þ r T ð2Þ
r T K 2 options. Furthermore, the initial R&D investment is so small as to
be ignorable. This is a typical twofold compound option, which is
Five factors affect the value of this put option: (1) the value of
a call on a call option. The value of this compound option (C,
the underlying asset (S), (2) the strike value (K), (3) the time to
Eq. (3)) is given as [51]:
expiration (T), (4) the volatility of the asset (r), and (5) the risk-free
sffiffiffiffiffi! sffiffiffiffiffi!
interest rate (r). Using our SD model, we can obtain the total PV
T1 rT 2 T1
investment (K) to achieve the policy target over the next 20 years C ¼ S2 M a1 ; b1 ;  I2 e M a2 ; b2 ;
T2 T2
(T). Given overall PV value volatility (r) and the risk-free interest
rate (r), we can calculate the value of a put option that makes  I1 erT 1 Nða2 Þ ð3Þ
the overall income (S) equal to the total investment (K). Detailed
information of input parameters is provided in Table 4.
     
1 S2 r2
The value of the put option is 205.90 billion KRW ($US 187 mil- a1 ¼ pffiffiffiffiffi ln  þ r þ v T 1 ; a2
rv T 1 S 2
lion). Dividing this figure by the PV electricity generation target of      
1 S2 r2
462,698 MW h by 2015, we can obtain the optimal unit subsidy per ¼ pffiffiffiffiffi ln þ r  v T1 ð4Þ
electricity of 445 KRW/KWh ($US 0.404). The value of a put option rv T 1 S2 2
should vary with the risk-free interest rate, volatility, and invest-      
ment. Government has difficulty controlling the first two factors; 1 S2 r2
b1 ¼ pffiffiffiffiffi ln þ r þ v T2 ; b2
however, it can reduce the required total investment by increasing rv T 2 I2 2
     
R&D funding. For instance, the module cost is assumed to be 1 S2 r2v
0.5$/W in the above model; nevertheless, it can be further reduced. ¼ pffiffiffiffiffi ln þ r T2 ð5Þ
rv T 2 I2 2
Thus, we trace how the total subsidy changes in response to the
varying volatility and investment. As shown in Fig. 5, as the PV sys- where M is the cumulative bivariate normal distribution function
tem becomes more volatile, the R&D investment can further drive with a and b as the respective p upper
ffiffiffiffiffiffiffiffiffiffiffiffiffi and lower integral limits,
down the total subsidy by reducing the total investment, which and the correlation coefficient is T 1 =T 2 . The variable S⁄ is the crit-
implies the importance of the R&D investment under severe ical value of the PV R&D above which the follow-up R&D (the first
uncertainties. call option) will be conducted. Given the policy targets comprising
the module cost of 0.5$/W by 2014, and the PV electricity genera-
tion of 462,698 MW h by 2015, we calculate the expected income
Table 4 (S2) and installation investment (I2) using our SD model. With sim-
Input parameters of the put option in the Black–Scholes model.
ilar notations as used above, detailed information of input parame-
Contents Unit Value ters is presented in Table 5.
S (income) 100 million KRW 11,482 ($US 1096.0 million) The value of the PV R&D investment is now captured by the
K (total investment) 100 million KRW 11,482 ($US 1096.0 million) expanded NPV, which is the sum of the static NPV and Geske call
T (operation period) Year 20 option. As previously shown, increasing the public R&D investment
r (risk-free interest rate) % 2.80
can reduce the financial subsidy, but lower the expanded NPV. The
r (volatility) % 43.49
negative expanded NPV can lead to reduced investments of private
 C. Jeon et al. / Applied Energy 142 (2015) 33–43 41

Table 5
Input parameters for the Geske compound option.

Contents Unit Value

T1 (follow-up R&D investment time) Year 1
T2 (installation investment time) Year 3
r (risk-free interest rate) % 2.80
rv (volatility) % 43.49
S2 (income) 100 million KRW 19,680 (US$ 1788.7 million)
I2 (installation investment) 100 million KRW 12,058 (US$ 1096.0 million)
I1 (follow-up R&D investment) 100 million KRW Variable

companies and should thus be avoided. Therefore, the optimal sophisticated ways, create uncertainties, and continuously change
public R&D investment is defined as the investment of maximizing options. Real option valuation has difficulty handling the opportu-
the overall reduction of R&D investment and financial subsidy nity, in which sources of change are not clear. SD with Monte Carlo
while keeping the strategic NPV positive. simulation enables us to analyze the dynamics of a complex sys-
The Korean government sets a plan of providing companies tem, quantify its variables and interactions, estimate its volatility,
with the financial subsidy of 942.9 billion KRW ($US 857.0 mil- and thus enable real option valuation to overcome one of its
lion), and of investing 1158.5 billion KRW ($US 1.053 billion) in limitations. Moreover, our approach is useful for identifying key
PV R&D [68,70]. As shown in Fig. 6, the government can increase uncertainties, which therefore makes it easy to reflect those
the R&D investment up to 1500 billion KRW ($US 1.363 billion) uncertainties in valuation and optimization.
while keeping the expanded NPV positive. Our real option model From a practical standpoint, using our method, policymakers
estimates the optimal R&D investment of 1500 billion KRW, which can optimize financial subsidies and public R&D investment
can reduce the financial subsidy to 205.9 billion KRW ($US together. In other words, they can achieve the fundamental policy
187.1 million). Compared to the figures in the Korean government targets of renewable energy technology development and deploy-
plan, the R&D investment increases by 341.5 billion KRW ($US ment with less money and therefore efficiently enhance energy
310.4 million); however, the financial subsidy is reduced by security and environmental sustainability. As previously shown,
737.0 billion KRW ($US 669.9 million). It leads to the overall bud- using our method, the Korean government can cut the overall PV
get reduction of 395.5 billion KRW ($US 359.5 million), which subsidy by 18.8% while achieving the policy target. Rebalancing
demonstrates that our approach can make energy subsides more of the subsidy portfolio makes it possible with more PV R&D
efficient. The policy target of PV electricity production can be investment and less financial subsidy. Applied to other renewable
achieved in both subsidy plans, which are analyzed in our SD energy technologies, our approach may alleviate the fiscal burden
model. In other words, the solution of achieving governmental pol- to some extent. Taking a step further, it can help policymakers turn
icy targets as well as managerial objectives of private companies is a vicious cycle of any renewable energy technology into a virtuous
neither more R&D investments nor more financial subsidies. The cycle. They can understand what the trigger of the shift is, and fur-
real solution is an optimal subsidy estimation of a single policy; ther estimate the optimal subsidy they need.
it should be an optimal subsidy portfolio from a broad perspective. In terms of policy design and implementation, we suggest a
dynamic subsidy budget planning and rebalancing process based
on our multi-subsidy optimization method. An integrated budget
5. Conclusion and policy implications proposal should be made for financial subsidies together with
R&D funding, as well as for several types of subsidies. Then, from
Our primary academic contribution is to reflect dynamic uncer- a multi-subsidy optimization perspective, it should be examined
tainties resulting from complex interactions among various factors and fixed. Through this process, the overall budget is likely to be
in real option valuation. Real option valuation is useful for valuing less than the sum of separately decided subsidies. In addition, by
opportunities under uncertainties. However, in a complex system, observing changing interactions and uncertainties among key
internal and external factors underlying an opportunity interact in stakeholders and factors, policymakers should continuously

Fig. 6. Relationship between expanded NPV and R&D investment.

42 C. Jeon et al. / Applied Energy 142 (2015) 33–43

fine-tune the subsidy allocation and can thereby reduce overall [12] Midttun A, Gautesen K. Feed in or certificates, competition or
complementarity? Combining a static efficiency and a dynamic innovation
costs while meeting policy targets. For the Korean PV policy, we
perspective on the greening of the energy industry. Energy Policy
find that it is better for the Korean government to more aggres- 2007;35:1419–22.
sively invest in PV R&D subsidies. Under increasing uncertainties [13] Nemet GF, Baker E. Demand subsidies versus R&D: comparing the uncertain
of PV R&D, the government is going to reduce R&D funding more impacts of policy on a pre-commercial low-carbon energy technology. Energy J
2009;30.
than financial subsidies. However, it is not an efficient policy; it [14] Altenburg T, Engelmeier T. Boosting solar investment with limited subsidies:
squanders approximately $US 359.5 million. The most efficient rent management and policy learning in India. Energy Policy 2013;59:866–74.
way is to increase R&D funding by $US 310.4 million. This increase [15] Dartanto T. Reducing fuel subsidies and the implication on fiscal balance and
poverty in Indonesia: a simulation analysis. Energy Policy 2013;58:117–34.
will lead to the reduction of financial subsidies by $US [16] Hong L, Liang D, Di W. Economic and environmental gains of China’s fossil
669.9 million. energy subsidies reform: a rebound effect case study with EIMO model. Energy
If various renewable energy technology subsidies are deter- Policy 2013;54:335–42.
[17] Davis GA, Owens B. Optimizing the level of renewable electric R&D
mined in this way, corporate managers can rationally expect the expenditures using real options analysis. Energy Policy 2003;31:1589–608.
amount and timing of public subsidies over the long term. Further- [18] Kim K, Lee C. Evaluation and optimization of feed-in tariffs. Energy Policy
more, they can identify various future uncertainties, understand 2012;49:192–203.
[19] Cassimon D, Engelen P, Yordanov V. Compound real option valuation with
their effects on investment opportunities, and thus make more phase-specific volatility: a multi-phase mobile payments case study.
effective private R&D and power system investment decisions. Technovation 2011;31:240–55.
However, if subsidies are not efficiently designed, corporate man- [20] Perlitz M, Peske T, Schrank R. Real options valuation: the new frontier in R&D
project evaluation? R&D Manage 1999;29:255–70.
agers will recognize that the probability of expected loss will
[21] Cory K, Couture T, Kreycik C, Williams E. A policymaker’s guide to feed-in tariff
increase. They will hold investments, or withdraw from the policy design. Colorado (USA): National Renewable Energy Laboratory; 2010.
market. [22] Tamás MM, Bade Shrestha S, Zhou H. Feed-in tariff and tradable green
Our method is valuable but has several limitations. Above all, certificate in oligopoly. Energy Policy 2010;38:4040–7.
[23] Amram M, Kulatilaka N. Strategy and shareholder value creation: the real
subsidy policies could differ in design and implementation options options frontier. J Appl Corp Finance 2000;13:15–28.
over energy sources, regions, and countries. Subsequently, appro- [24] Menegaki A. Valuation for renewable energy: a comparative review. Renew
priate optimization models might vary. These are not sufficiently Sustain Energy Rev 2008;12:2422–37.
[25] Lee S. Using real option analysis for highly uncertain technology investments:
considered in our study. The energy environment will become the case of wind energy technology. Renew Sustain Energy Rev
increasingly more volatile, thereby making traditional real option 2011;15:4443–50.
models inappropriate. To address this issue, we require more [26] Venetsanos K, Angelopoulou P, Tsoutsos T. Renewable energy sources project
appraisal under uncertainty: the case of wind energy exploitation within a
advanced real option models, including the jump model, regime- changing energy market environment. Energy Policy 2002;30:293–307.
switching model, and others. A sum of optimal subsidies might [27] Lee S, Shih L. Renewable energy policy evaluation using real option model—the
not lead to overall optimal portfolio in many cases, which suggests case of Taiwan. Energy Econ 2010;32:S67–78.
[28] Bednyagin D, Gnansounou E. Estimating spillover benefits of large R&D
that further studies are needed on overall subsidy portfolio optimi- projects: application of real options modelling approach to the case of
zation. Furthermore, a causal link between PV R&D investments thermonuclear fusion R&D programme. Energy Policy 2012;41:269–79.
and cost reductions might not be assumed in another energy tech- [29] Siddiqui AS, Marnay C, Wiser RH. Real options valuation of US federal
renewable energy research, development, demonstration, and deployment.
nology. Moreover, evidence exists that such causality in PV tech-
Energy Policy 2007;35:265–79.
nology is weakening in some countries because the marginal cost [30] Zhou W, Zhu B, Fuss S, Szolgayová J, Obersteiner M, Fei W. Uncertainty
of PV R&D as well as investment risk becomes ever higher [63]. modeling of CCS investment strategy in China’s power sector. Appl Energy
In these cases, causal loops should be carefully modified. Finally, 2010;87:2392–400.
[31] Zhu L, Fan Y. A real options-based CCS investment evaluation model: case
some disruptive technological innovations and extreme events study of China’s power generation sector. Appl Energy 2011;88:4320–33.
could render useless the subsidy policies of an energy technology; [32] Doherty R, O’Malley M. The efficiency of Ireland’s Renewable Energy Feed-In
however, these are not considered in our SD model. Tariff (REFIT) for wind generation. Energy Policy 2011;39:4911–9.
[33] Devine M, Farrell N, Lee W. Managing investor and consumer exposure to
electricity market price risks through Feed-in Tariff design; 2014.
Acknowledgements [34] Farrell N, Devine M, Lee W, Gleeson J, Lyons S. Specifying an efficient
renewable energy Feed-in Tariff; 2013.
[35] Dhawan R. System dynamics and its impact on managerial decision making.
This research was supported by the Basic Science Research Pro- Special Integration Group on Systems Applications in Business and Industry;
gram through the National Research Foundation of Korea (NRF) Saturday 2005.
funded by the Ministry of Education (NRF-2013R1A1A2009406). [36] He Y, Wang B, Wang J, Xiong W, Xia T. Residential demand response behavior
analysis based on Monte Carlo simulation: the case of Yinchuan in China.
Energy 2012;47:230–6.
References [37] Kräupl S, Wieckert C. Economic evaluation of the solar carbothermic reduction
of ZnO by using a single sensitivity analysis and a Monte-Carlo risk analysis.
Energy 2007;32:1134–47.
[1] IEA. World energy outlook 2012. Paris: IEA; 2012.
[38] Ansari N, Seifi A. A system dynamics analysis of energy consumption and
[2] Badcock J, Lenzen M. Subsidies for electricity-generating technologies: a
corrective policies in Iranian iron and steel industry. Energy 2012;43:334–43.
review. Energy Policy 2010;38:5038–47.
[39] Marquez AC, Blanchar C. A decision support system for evaluating operations
[3] REN 21 (Renewable Energy Policy Network for the 21st Century). Renewables
investments in high-technology business. Decis Support Syst 2006;41:472–87.
2013 Global Status Report. Paris: REN21 Secretariat; 2013.
[40] Ford A. Using simulation for policy evaluation in the electric utility industry.
[4] Zhang Y, Song J, Hamori S. Impact of subsidy policies on diffusion of
Simulation 1983;40:85–92.
photovoltaic power generation. Energy Policy 2011;39:1958–64.
[41] Naill RF. A system dynamics model for national energy policy planning. Syst
[5] Stern NH, Britain G, Treasury H. Stern Review: the economics of climate
Dyn Rev 1992;8:1–19.
change. London: HM Treasury; 2006.
[42] Steman JD. Business dynamics; system thinking and modeling for complex
[6] Kramer GJ, Haigh M. No quick switch to low-carbon energy. Nature
world. Irwin: McGraw Hill; 2000.
2009;462:568–9.
[43] Schulmerich M. Real options valuation. Springer; 2005.
[7] Van Benthem A, Gillingham K, Sweeney J. Learning-by-doing and the optimal
[44] Amram M, Kulatilaka N. Real options: managing strategic investment in an
solar policy in California. Energy J 2008:131–51.
uncertain world. Boston: Wiley Online Library; 1999.
[8] Unruh GC. Understanding carbon lock-in. Energy Policy 2000;28:817–30.
[45] Black F, Scholes M. The pricing of options and corporate liabilities. J Polit Econ
[9] Butler L, Neuhoff K. Comparison of feed-in tariff, quota and auction
1973:637–54.
mechanisms to support wind power development. Renew Energy
[46] Barone-Adesi G, Whaley RE. Efficient analytic approximation of American
2008;33:1854–67.
option values. J Finance 1987;42:301–20.
[10] Couture T, Gagnon Y. An analysis of feed-in tariff remuneration models:
[47] Geske R, Roll R. On valuing American call options with the Black–Scholes
implications for renewable energy investment. Energy Policy 2010;38:955–65.
European formula. J Finance 1984;39:443–55.
[11] Dusonchet L, Telaretti E. Economic analysis of different supporting policies for
[48] Cox JC, Ross SA, Rubinstein M. Option pricing: a simplified approach. J Finance
the production of electrical energy by solar photovoltaics in western European
Econ 1979;7:229–63.
Union countries. Energy Policy 2010;38:3297–308.
 C. Jeon et al. / Applied Energy 142 (2015) 33–43 43

[49] Breen R. The accelerated binomial option pricing model. J Financial Quant Anal [63] Candelise C, Winskel M, Gross RJ. The dynamics of solar PV costs and prices as
1991;26:153–64. a challenge for technology forecasting. Renew Sustain Energy Rev
[50] Derman E, Kani I, Chriss N. Implied trinomial tress of the volatility smile. J 2013;26:96–107.
Derivatives 1996;3:7–22. [64] Popp D. ENTICE: endogenous technological change in the DICE model of global
[51] Geske R. The valuation of compound options. J Finance Econ 1979;7:63–81. warming. J Environ Econ Manage 2004;48:742–68.
[52] Cassimon D, Engelen P, Thomassen L, Van Wouwe M. The valuation of a NDA [65] Waisman H, Guivarch C, Grazi F, Hourcade JC. The IMACLIM-R model:
using a 6-fold compound option. Res Policy 2004;33:41–51. infrastructures, technical inertia and the costs of low carbon futures under
[53] Shockley RL, Curtis S, Jafari J, Tibbs K. The option value of an early-stage imperfect foresight. Clim Change 2012;114:101–20.
biotechnology investment. J Appl Corp Finance 2002;15:44–55. [66] Leimbach M, Bauer N, Baumstark L, Edenhofer O. Mitigation costs in a
[54] Cortelezzi F, Villani G. Valuation of R&D sequential exchange options using globalized world: climate policy analysis with REMIND-R. Environ Model
Monte Carlo approach. Comput Econ 2009;33:209–36. Assess 2010;15:155–73.
[55] Siegel DR, Smith JL, Paddock JL. Valuing offshore oil properties with option [67] Bosetti V, Carraro C, Galeotti M, Massetti E, Tavoni M. A world induced
pricing models. Midland Corp Finance J 1987;5:22–30. technical change hybrid model. Energy J 2006;27.
[56] Frayer J, Uludere NZ. What is it worth? Application of real options theory to [68] KETEP (Korea Institute of Energy Technology Evaluation and Planning). Green
the valuation of generation assets. Electr J 2001;14:40–51. energy strategic roadmap 2011. Seoul: KETEP; 2011.
[57] Yu W, Sheblé GB, Lopes JAP, Matos MA. Valuation of switchable tariff for wind [69] KERI (Korea Electro-technology Research Institute). Amendments of the
energy. Electr Power Syst Res 2006;76:382–8. Korean Feed-in-Tariffs in 2011. Changwon: KERI; 2010.
[58] Fernandes B, Cunha J, Ferreira P. The use of real options approach in energy [70] MKE (Ministry of Knowledge Economy). The third basic plan for new and
sector investments. Renew Sustain Energy Rev 2011;15:4491–7. renewable energy R&D and development. Seoul: MKE; 2008.
[59] Kumbaroğlu G, Madlener R, Demirel M. A real options evaluation model for the [71] KIER (Korea Institute of Energy Research). Renewable energy data center.
diffusion prospects of new renewable power generation technologies. Energy http://kredc.kier.re.kr/kier/default.asp.
Econ 2008;30:1882–908. [72] KOSIS (Korean Statistic Information Service). Database. http://www.kosis.kr.
[60] IRENA (International Renewable Energy Agency). Renewable power generation [73] Barlas Y. Formal aspects of model validity and validation in system dynamics.
costs in 2012: an overview. Abu Dhabi: IRENA; 2013. Syst Dyn Rev 1996;12:183–210.
[61] Owen AD. Renewable energy: externality costs as market barriers. Energy [74] Perron P. Trends and random walks in macroeconomic time series: further
Policy 2006;34:632–42. evidence from a new approach. J Econ Dyn Control 1988;12:297–332.
[62] Longo A, Markandya A, Petrucci M. The internalization of externalities in the
production of electricity: willingness to pay for the attributes of a policy for
renewable energy. Ecol Econ 2008;67:140–52.