Assessing Alternative Renewable Energy Policies in Korea’s Electricity Market^†

A. Nonrenewable Generation

The generation of nonrenewable energy source i is denoted by x_it, and its aggregate capacity is denoted by K_it. Let ζ_i denote the annualized construction cost for adding one unit (MW) of capacity for source i, taking its average lifespan into account. For each source at stage t, the wholesale settlement price according to the energy source is p_it, which is a function of the SMP, as explained below. In addition, τ_i represents various policy charges levied on the amount of power generated by each nonrenewable energy source.

Under the assumed cost structure, for each energy source, the sum of individual firms’ cost functions is that of the representative firm, as is the profit. The cost function of the representative nonrenewable-based firm for source i at stage t is C_i(x_it, K_it), which corresponds to the annual power generation costs (excluding construction costs) for generating x_it.3 Suppressing subscripts for simplicity, it is assumed that cost functions satisfy C_x ≡ ∂C / ∂x > 0 and C_xx ≡ ∂²C / ∂x² > 0 . That is, total costs and marginal costs increase in the generation amount. Further we assume C_K < 0 and C_KK > 0 . Therefore, total costs decrease in the capacity, while the marginal reduction of generation costs decrease in the capacity. By assuming C_xK < 0 , we deal with situations where an additional capacity increase leads to a reduction in the marginal cost.

A representative nonrenewable energy firm has the follows profits:

Equation (1) reflects the additional costs α_tB_t incurred to comply with the RPS obligation for one unit of nonrenewable generation. If a firm is simultaneously generating using nonrenewables and renewables, due to the separable cost structure, it can be understood that REC purchases and REC sales are occurring separately, where equation (1) only shows purchases.

In a competitive market, individual firms determine the amount of generation where the marginal cost curve (i.e., supply curve) matches the given price (e.g., Stoft, 2002; Borenstein and Holland, 2005; Biggar and Hesamzadeh, 2014). Given a linearly increasing individual supply curve, the slope becomes infinite (i.e., vertical marginal cost curve) after a firm’s generation reaches its maximum capacity. Thus, there are two situations: i) if the market price belongs to the vertical part of the supply curve, a firm generates to the maximum level of capacity, and ii) if the market price falls within the incremental part, it generates at the point where the price meets its supply curve. Note that the market price is determined at the point where the market-level supply curve and demand curve meet, where the annual market supply curve can be derived by horizontally summing individual marginal cost curves with respect to firms and hours.

The SMP in the wholesale electricity market is determined by the variable costs of a marginal generator. From an annual perspective, we consider that a representative gas-fired firm’s generation is determined at the point at which the aggregate marginal costs equal the wholesale settlement price under the given capacity.4 Assuming an interior solution, we exclude situations where the annual gas-fired generation becomes zero or reaches the maximum aggregate capacity. Then, for t = 1,2 and i = l equation (2) is established as a short-term equilibrium condition that determines the annual gas-fired generation.

On the other hand, we consider corner solution situations for nuclear and coal. The annual generation can be determined by multiplying the given capacity by the annual average capacity factor (β_it) and by 8,760 hours (= 24 hours × 365 days). Therefore, for t = 1,2 and i = n, c, equation (3) holds as a short-term equilibrium condition.

In the long run, the generation capacity can increase or decrease freely. In the end, the long-run equilibrium condition is that the utility’s profit becomes zero (Borenstein and Holland, 2005). Therefore, over the long term, for t = 1,2 and i = n, c, l, equation (4) holds together with the short-term equilibrium conditions.

In the pricing structure of the domestic electricity market in South Korea, the settlement price for nonrenewable energy, p_it, is determined by an adjustment process based on the SMP (to maintain the financial balance between KEPCO and its generation subsidiaries). Specifically, a typical settlement price is calculated by where is the SMP, C_x,i (x_it, K_it) denotes the variable costs, and λ_i represents the adjustment coefficients. Furthermore, we consider that the per-MWh compliance costs and various policy charges are also offset for producers, reflecting the actual settlement process. Taking this feature into consideration, we have equation (5) for t = 1,2 and i = n, c, l.

B. Renewable Generation

Following Fischer and Newell (2008) and Fischer, Preonas, and Newell (2017), we assume that the accumulative knowledge is S_jt = S_j(H_jt, L_jt). Note that S_jt is a function of knowledge from R&D, H_jt, and that from learning-by-doing, L_jt. We assume that S_H > 0 and S_L > 0 , as well as S_HL = S_LH. The annual new R&D knowledge generated in each stage, h_jt, increases the cumulative R&D knowledge; i.e., H_j2 = H_j1 + n₁h_j1, and the annual production in each stage increases the accumulative experience knowledge; i.e., L_j2 = L_j1 + n₁y_j1. The R&D expenditure, R_j(h_j1), is increasing and convex, i.e., R_h > 0 and R_hh > 0 , and R(0) = 0. The government supports proportion σ_j of renewable energy R&D expenditures.

When the above is reflected, profits for the representative renewable-based firm can be expressed as equation (6).

As mentioned above, the generation amount (MWh) is directly determined according to the capacity factor, and renewable generation satisfies equation (7).

With regard to knowledge accumulation, the degree of appropriability of newly acquired knowledge, ρ, can be considered, where we assume that solar power and wind power have the same value.5 Below, the incomplete transfer rate is considered (0<ρ<1), and for knowledge that a specific firm appropriates, other firms can use it by paying a license royalty. In particular, it is assumed that newly acquired knowledge is ultimately utilized through the use of either imitation or permission.6 As a result, the license revenue does not appear in equation (6), as this represents a simple transfer between firms.

The model takes into account second-stage cost savings which arise due to appropriable parts of new R&D-based knowledge in the first stage. Therefore, ρ appears in the representative firm’s first-order conditions, which are derived from those of the individual firms (see Appendix A. for the derivation). Finally, for j = s, w, the first-order condition associated with setting the value of the annual R&D knowledge, h_jt, can be expressed as equation (8).

The first term in equation (8) is the amount of R&D net investment that must be paid in the first stage to acquire additional units of knowledge, and the negative of the second term represents discounted gains in the second stage from appropriating the new knowledge. Finally, the amount of new R&D knowledge is determined at the point where the cost and benefit are equal.

The long-term equilibrium condition of the representative renewable-based firm is to obtain a zero profit. This means that the excess profits from R&D investment are transferred to the consumers in the long run. For j = s, w, the long-term equilibrium condition is as follows:

In the case of renewable-energy sources, the SMP itself is the settlement price; thereby, .

C. Consumers

The electricity demand function is derived from the quasi-linear utility function of the representative consumer such that the consumer welfare can be consistently evaluated for each scenario in the simulation later. Specifically, the utility function can be expressed as U = I +Θ(P)−M. Here, I is the monetary income expressed by reference goods, and the price of the reference goods is normalized by 1. The demand function then becomes D(P) = −∂Θ / ∂P. For the sake of simplicity, peak demand during the year and resulting fluctuations in demand are not taken into consideration. In addition, it is assumed that the consumer considers environmental costs, M. Specifically, such costs apply to only coal and gas, which heavily emit GHG, and to the three major air pollutants NO_X, SO_X and PM_2.5. Furthermore, it is assumed that consumers consider the possible nuclear accident costs (NAC), which are proportional to the capacity of nuclear plants. The five external costs above are expressed as ε_i,m, where m = GHG, NO_X, SO_X, PM_2.5 for i = c, l and m = NAC for i = n. Finally, the total external cost can be expressed as M_t = Σ_i∈{c,l}Σ_{m∈{GHG, NOX, SOX, PM2.5}}ε_i,mx_it + ε_n,ACK_nt for t = 1,2 . Except for external costs, the consumer welfare for the two stages is expressed by equation (11).

The retail electricity price, P_t, is calculated by summing the aggregate settlement price, , and the transmission and distribution costs per unit, Δ, and then adding the ad valorem sales tax ψ. As a result, for t = 1,2 equation (12) holds.

To be concrete, the aggregate settlement price, , is equal to the ‘total expenditure paid’ divided by the ‘total amount of electricity purchased’ by the distributer in the wholesale market:

where φ_XK_Xt implies capacity charges for other nonrenewable energy sources.

In South Korea, although the discussion on price rationalization continues, the government regulates the retail electricity price, ensuring that it is not directly connected to the wholesale price. Under this circumstance, changes in the RPS obligation compliance costs cannot easily be fully reflected in the retail price. To examine the mid- and long-term market effects, however, in this paper we assume that the wholesale price is connected to the retail price in the future as well as in earlier periods.

D. Equilibrium

First, for t = 1,2 equilibrium in the power market must satisfy the equation

where ℓ is the loss ratio due to transmission and distribution.

Assuming full compliance under the RPS, the total amount of RECs, ∑_jχ_jy_jt + Y_t, equals the obligation amount, α_t×(∑_ix_it + X_t). As a result, the supply-obligation rate, α_t, can be expressed as

Note that the numerator in equation (15) is based on effective units of renewable generation considering REC weights; accordingly, the physical amount of renewable generation decreases as the weights exceed 1, and vice versa. Later, we will look at the renewable share based on physical units, which is essentially related to achieving the policy goal.

Equilibrium in the integrated REC market requires that the demand and supply for RECs meet across renewable energy sources. In order to grasp the amount that a (marginal) buyer is willing to pay, equation (2) can be rewritten as

From the short-term standpoint of gas-fired firms, the left-hand side of equation (16) corresponds to the per-MWh gain that a gas-fired firm can earn in the wholesale market, and the right-hand side corresponds to per-MWh compliance costs. Regarding the long-term behavior, rearranging equation (4) yields the following expression for t = 1,2 and i = n, c, l,

where the left-hand side of equation (17) corresponds to the per-MWh gain in the long run. Note that for a gas-fired firm, it holds that C_x,i(x_it, K_it) = [C_i(x_it,K_it) + (ζ_i - φ_i)K_it] / x_it from equations (16) and (17), indicating that for the marginal energy source, marginal costs equal average costs in the long run. Let the left-hand side of equation (17), identical for all i, be Φ_t. In a competitive integrated REC market, arbitrage cannot occur across energy sources. Thus, all nonrenewable-based firms pay Φ_t to comply with their per-MWh obligations.

For renewable sources, equations (9) and (10) can be rearranged as follows:

The left-hand sides of equations (18) and (19) correspond to the minimum acceptable amount for a REC seller, while the right-hand sides correspond to the REC sales revenue per 1MWh of renewable generation. In the competitive REC market, each left-hand side has a unique value regardless of the source, denoted by Ψ_t.

Using equations (17)-(19), for t = 1,2 REC market equilibrium can be defined as

Finally, equation (20) implies that the total compliance costs for nonrenewable generation and the total REC revenue for renewable generation are balanced. As long as the equilibrium REC price is greater than zero, the RPS imposes a burden on nonrenewable generation through the market mechanism and assists in the development of renewables. However, the burden of nonrenewables will eventually be passed on to consumers due to the compliance cost return mechanism, as explained in the context related to equation (13).

Under the RPS, equilibrium can be characterized by 34 equations from equations (2)-(5), (7)-(10), (12), (14), and (20), which can be solved for 34 endogenous variables (x_it, K_it, y_jt, Z_jt, h_t, p_it, , P_t, B_t for i = n, c, l, j = s, w, and t = 1,2 ). In order to analyze a situation without the RPS, equation (20) should be dropped. Furthermore, in order to consider renewable energy subsidies, b_jt, in the absence of the RPS, one can replace B_t with b_jt in equations (9) and (10).

After finding the equilibrium values, the social welfare value can then be calculated. The consumer surplus, except for the externality costs, is equal to equation (11), and the external costs are equal to EX = n₁M₁ + δn₂M₂. The government net revenue is given by equation (21),

where Q_t ≡ ∑_ix_it + ∑_jy_jt + X_t + Y_t and is based on equation (13). Social welfare can be defined as W = PS + CS + RV + EX, where PS = 0 in the long run. This welfare measure has its limitations such that it ignores benefits from enhancing national energy self-sufficiency or potential grid costs due to the unstable characteristics of renewable energy sources. In the context of a lack of related research in Korea, however, it still provides a useful metric of resulting welfare effects caused by alternative renewable energy policies.

A. Primary Data

Information about capacity, generation, and price levels in the first stage are based on 2016 data from KEPCO (2017). For the second stage, we use projected values for the year 2030 provided by MOTIE (2017), reflecting RE3020. Other nonrenewable energy generation, X_t, is assumed to be fixed over the first and second stages; these values are calculated using the 2016 raw data of KEPCO (2017). For other renewable energy generation, Y₁ is based on KEPCO’s (2017) raw data, while Y₂ is constructed using projected values from MOTIE (2017), reflecting an increasing trend of the other renewable category. For the amount of electricity that is actually supplied in the wholesale market, based on KPX (2018), intra-plant load factors, ω_i, are considered for nonrenewable sources though not for renewable sources.

Figures regarding R&D spending on core renewable technologies is collected from internal data of Korea Energy Technology Evaluation and Planning (KETEP). Cumulative R&D knowledge of solar power in the first stage is normalized as H_s1 =1, and cumulative knowledge of wind power is calculated as H_w1 = 0.69 by considering the proportion of relative R&D expenditures during 2007-2016. Cumulative experience knowledge, L_j1, is estimated using cumulative generation volumes during 2005-2016. The amount of new R&D knowledge obtained through the first stage, h_j1, can be calculated using equation (9), which requires information about second-stage generation. Therefore, we start with the projected 2030 capacity values to calculate and then input in equation (9) to obtain the estimated capacity values. Iterating this procedure until precisely replicating the first stage value can give us the reference value of h_j1.

Hourly CP rates for nonrenewable sources are derived using internal data of the KPX, and corresponding annual rates are calculated by considering plant utilization factors (KPX, 2014). Annualized fixed costs for nonrenewable and renewable energy sources come from KPX (2018)’s data, which are similar to ‘overnight costs divided by the plant lifetime’ based on figures from the International Energy Agency (IEA) and Nuclear Energy Agency (NEA) (2015). Retail costs are calculated using KEPCO’s Electricity Cost Information for 2013-2017. Specifically, the total sales figure of the retail market minus the cost of purchasing electricity is regarded as the residual retail cost, which is ₩19,355/MWh. The degree of appropriability is set to 0.5 for both solar power and wind power in keeping with Fischer, Preonas, and Newell (2017), meaning that the social return of R&D is double the private return.

B. Functional Assumptions

For nonrenewable sources i = n, c, l, we employ quadratic cost functions that satisfy foregoing assumptions, , where the superscript BL indicates baseline values. Because , we can calibrate c_0,i, which corresponds to all fixed costs except for the annualized construction costs, based on equation (4). Given the above cost functions, marginal cost functions are , and at the baseline . First, one can obtain c_{1, l}, which corresponds to gas power marginal costs at the baseline, using equation (2). Based on the internal data of KPX, we horizontally sum individual marginal cost functions to obtain aggregate marginal cost functions according to the energy type and derive c_1,i for i = n, c and c_2,i for i = n, c, l.

Cost functions for renewable sources j=s,w are specified as , where g_0,j captures fixed costs apart from annualized construction costs. One can calibrate g_0,j using equation (9). For j=s,w, knowledge functions are S_j(H_jt, L_jt) = (H_jt / H_j1)^s_1,j(L_jt / L_j1)^s_2,j, where the cumulative knowledge stock has constant elasticity with respect to R&D knowledge and experience knowledge, and the first-stage knowledge stock is normalized to 1, i.e., S_j(H_j1, L_j1) = 1 . R&D knowledge parameters are set to s_1,s = s_1,w = 0.3 and learning knowledge parameters are set to s_2,s = s_2,w = 0.3; thus, projected outcomes under RE3020 are roughly implemented in the baseline. R&D expenditure functions are for j = s, w, which yield constant elasticities, γ_2,j. Following Fischer and Newell (2008) and Fischer, Preonas, and Newell (2017), we set γ_2,j = 1.2 for j=s,w,9 with γ_1,j calculated using equation (8).

A constant elasticity aggregate demand function is utilized, in this case D(P_t, N_t) = N_t × (P_t)^η, where the price elasticity of electricity demand, η, is assumed to be -0.3 from a mid- to long-term perspective. Several recent domestic studies estimate the elasticities of electricity demand as follows. Kim and Park (2013), using monthly consumption data for 1981-2011, suggest that the price elasticity rates of industrial (residential) consumption are -0.127 (0.143) before 1997 and -0.088 (0.123) after 1997. Lim, Lim, and Yoo (2013), based on survey data of 521 households in 2012, estimate the price elasticity of residential consumption as - 0.68. As examples of elasticity application for a model simulation, Borenstein and Holland (2005) assume that the demand elasticity is -0.1 in the short run and -0.3 or -0.5 in the long run. In Fischer and Newell (2008), it is assumed to be -0.2 from a long-term perspective, while Fischer, Preonas, and Newell (2017) apply -0.1 from a short-term perspective. Note that N_t represents exogenous changes in the power demand level. We apply N₁ = 1.64 × 10¹⁰ considering the baseline price and quantity and N₂ = 1.85×10¹⁰ (= N₁ × 1.125) based on MOTIE’s (2017) projected increase in demand.

The solar and wind capacity factors in the first stage are calculated as β_s1 = 0.138 and β_w1 = 0.182, respectively, using aggregate generation and capacity data from KEPCO (2017). For the second stage, β_s2 = 0.14 which is the average of the predicted capacity factor values for 2017-2031. We also set β_w2 = 0.21 by assuming that the wind capacity factor increases at a rate identical to that of the solar power.

C. Externality Costs

GHG social costs in the first stage are assumed using Yi’s (2018) results, which are the basis of the recently published Financial Reform Special Committee (2018): ₩35,680/MWh for coal and ₩15,720/MWh for gas. When referring to figures in old publications—emission factors in MOTIE (2014) and marginal damages in MOTIE (2015)—we have externality costs of ₩20,575/MWh for coal and ₩9,063/MWh for gas. Accordingly, it appears that external costs are experiencing upward adjustments to reflect increasing environmental damages.

With respect to GHG externality costs in the second stage, an additional discussion regarding the trends of the estimates in the literature is necessary. The US Government (2016) specifies that under three discount rate assumptions (5%, 3%, and 2.5%), the US CO₂ social costs in 2030 compared to 2015 are greatly increased (46%, 39%, and 30%). Nordhaus (2017), relying on the updated Dynamic Integrated Model of Climate and the Economy (DICE), presents estimates of the global CO₂ social cost of $31.2/tCO₂ in 2015 and $51.6/tCO₂ in 2030, indicating an increase of 65%. In addition, applying the discount rate of Stern (2007), which is approximately 1.4%, the corresponding values are estimated to be $197.4/tCO₂ in 2015 and $376.2/tCO₂ in 2030, resulting in an increase of 91%. Along with the trends in these estimates, we assume that the discounted GHG social costs in the second stage are 34% higher than those in the first stage, applying a discount rate of 7%; these outcomes are ₩47,759/MWh for coal and ₩21,042/MWh for gas. Thereby, we specify costs as ε_it,GHG for i = n, c and t = 1, 2.

Air pollutant social costs follow the values given by Yi (2018): NO_X, SO_X and PM_2.5, respectively, cause social costs of ₩16,590/MWh, ₩15,740/MWh, and ₩800/MWh for coal, and ₩4,630/MWh, ₩310/MWh, and ₩320/MWh for gas. These figures are applied in both stages without trends because air pollutants cause damages locally and do not stay in the atmosphere for a long time. In addition, the nuclear accident risk cost is based on the value provided by KPX (2018), where several alternatives are specified depending on how the probability of an accident is handled. We employ the estimate of ₩67,644,000/MW, as finally determined in the analysis by KPX (2018).

A. Baseline RPS and Repeal of the RPS

In the baseline scenario, the first-stage market results replicate the raw data, while the second-stage results are of interest because there are no values to be replicated other than the renewable generation capacity. It was found that if an RPS supply-obligation rate of 28% is applied in the second stage, the renewable share increases to 19.6%, where the annual solar and wind power generation levels are similar to each other. Specifically, the annual generation levels are 32,544 MWh for solar and 46,627 MWh for wind. This is quite similar to the plans for these two power sources (33,530MWh for solar and 42,566MWh for wind) by MOTIE (2017).

In the baseline, the annual R&D expenditure in the first stage is approximately ₩87.1 billion/year for solar (₩435.5 billion in total during the first stage) and about ₩70.1 billion/year for wind (₩350.5 billion in total). Compared to the first stage, in the second stage the generation costs are reduced by 24.7% for solar and 17.3% for wind, owing to technology development and learning effects. Under the cost reduction, the REC price drops to ₩5,781/REC but remains positive, meaning that the RPS is binding even in the second stage. In other words, generation costs for renewable energy are still higher than the costs for fossil fuels.

The second-stage retail price fell by 4.1% from that in the first stage due mainly to a decline in the generation costs for renewable energy. Owing to the reduced REC price, the annual RPS compliance cost (= REC price×weighted renewable energy generation) dropped greatly from ₩2.7 trillion/year in the first stage to ₩76 billion/year in the second stage. In addition, due to the expansion of renewable energy, the generation volume of nonrenewable energy in the second stage is reduced compared to that in the first stage. The generation mix consists of 30% nuclear, 39% coal, 23% gas, 3% other nonrenewables, and 5% renewable in the first stage, and with corresponding rates of 26%, 35%, 17%, 3% and 20% in the second stage. The resulting capacity factors for nuclear, coal and gas are 76%, 73%, and 42% in the first stage and 76%, 73%, and 35% in the second stage, respectively, where there is a decrease only in gas-fired generation. The reduced generation volume contributes to lower nonrenewable-energy marginal costs, resulting in a slight decline of 0.03% in the wholesale electricity price. As a result of this price effect, annual electricity consumption in the second stage is increased by 13.9% from the first stage, which exceeds the exogenous demand increase of 12.5%.

When the RPS is abolished, there is no renewable generation in either the first or second stages, and no R&D investment in renewable energy occurs. The wholesale settlement price rises slightly by 0.11% from the first stage to the second because marginal costs increase as nonrenewable generation increases due to the increase in exogenous demand. The retail price declines by 0.5% in the second stage compared to the first stage because there is a decrease in capacity payments in the second stage, which overwhelms the increase in the wholesale settlement price. Compared to the repeal of RPS scenario, in the baseline scenario retail prices are increased by 5.0% in the first stage and 1.1% in the second stage. In other words, under the RPS, it can be understood that the payment for the RPS compliance cost is placed on top of the wholesale adjusted price, causing an increase in retail prices. As a result, total consumption in the baseline scenario is decreased by 1.4% in the first stage and 0.3% in the second stage, relative to the no RPS scenario.

With regard to welfare effects, the baseline scenario, compared to the no RPS scenario, shows a welfare increase of ₩44.7 trillion, which means a relative gap in the discounted 25-year welfare sum. Looking at distributional effects, government revenue in the baseline is reduced compared to the no RPS scenario mainly due to i) an additional R&D expenditure and ii) lower wholesale tax revenue. Compared to the no RPS scenario, in the baseline increased retail prices lower consumer surplus by ₩18.5 trillion, while the reduced amount of nonrenewable generation results in lower externality costs of ₩63.6 trillion.

B. Past RPS and Optimal RPS

In terms of the second-stage supply-obligation rate, which is 28% in the baseline, the past RPS and optimal RPS scenarios respectively apply a lower rate of 17% and a higher rate of 31%. Note that as the second-stage rate increases, i.e., when we go through ‘no RPS’ → ‘past RPS’ → ‘baseline RPS’ → ‘optimal RPS’, the second-stage wholesale price decreases. This occurs due to the merit-order effect, by which a decrease in nonrenewable energy generation due to the strengthening of the RPS results in lower marginal costs at the margin. On the other hand, first-stage retail prices increase as the RPS is strengthened because a stronger second-stage supply-obligation rate necessitates a larger expansion of wind power, which is more expensive than solar power, thereby leading to rises in the REC price and retail price in the first stage.

It is important to examine changes in R&D investments. First, as the second-stage supply-obligation rate increases, the annual R&D investment increases significantly for both solar and wind power. This arises because if second-stage renewable generation is forcibly increased, the R&D incentive becomes stronger because the marginal effects of the second-stage cost reduction from the first stage of R&D increases. It appears that as the second-stage supply-obligation rate increases, the greater cost-reduction effects realizes. In addition, strengthening the RPS results in a larger role of wind power; thereby, wind power experiences a greater cost reduction from the baseline than solar power in the optimal RPS scenario. This arises because the first-stage accumulative knowledge stock for wind is lower than that of solar, meaning that the marginal cost-reduction effects from additional knowledge is higher for wind than for solar.

In term of welfare, when normalizing the welfare increase from the no RPS scenario to the past RPS scenario as 1, the baseline undergoes an increase in welfare by 1.7 times and the optimal RPS scenario experiences an increase by 1.9 times. In detail, the stronger the RPS supply-obligation rate is in the second stage, the higher the incentive for R&D investment becomes, resulting in higher government expenditures. A stronger second-stage supply-obligation rate also lowers consumer surplus with an increase in retail prices. Despite these negative effects, overall welfare increases in the second-stage RPS target rate owing to the considerable decrease in externality costs.

C. Renewable Generation Subsidy and Nuclear Power Reduction

In the subsidy scenario, it turns out that when subsidy levels are precisely equal to REC prices, the resulting renewable energy generation and cost saving effects are identical to those in the baseline scenario. As the RPS obligation is not incurred, however, retail electricity price falls by 4.8% in the first stage and by 1.2% in the second stage compared to the baseline scenario. Wholesale prices are higher in the subsidy scenario because nonrenewable energy generation is greater, yielding higher marginal costs than in the baseline.

As a result, replacing the RPS with a renewable generation subsidy yields a gain of ₩19.1 trillion in consumer surplus compared to the baseline, as electricity is supplied at lower retail prices. On the other hand, compared to the baseline, there is an increase in the government expenditure of ₩19.3 trillion, and the amount of nonrenewable energy generation increases, resulting in reduced external costs of ₩1.8 trillion. In summary, the effect of raising overall welfare from the no RPS scenario is ₩2.6 trillion lower than that in the baseline scenario.

As shown in Table 2, an introduction of the RPS from its absence mainly reduces gas-fired generation, which is marginal in terms of generation, and has minor effects on the remaining nonrenewable energy sources. In the baseline scenario, the second-stage nuclear power capacity is determined to be approximately 23,124MW, and nuclear generation accounts for 26.0% of the total in the second stage. The nuclear reduction scenario is to, from the baseline, limit aggregate nuclear capacity to 21,968 MW, which is a 5% reduction, thereby reducing nuclear generation from 153 TWh to 146 TWh under the baseline nuclear capacity factor, accounting for 24.6% of the total generation.

The forced reduction in the second-stage aggregate nuclear capacity results in an increase in the retail price owing to the increases in other expensive forms of nonrenewable generation. As a result, compared to the baseline, consumer surplus is lowered by about ₩300 billion, and producer surplus in nuclear power experiences a loss of ₩3.6 trillion as the long-term equilibrium condition is not met with increased marginal costs. The costs associated with the nuclear accident risk are reduced by about ₩640 billion compared to the baseline scenario, but the environment external costs are more than ₩3.2 trillion due mainly to the increased amount of gas-fired generation. As a result, overall welfare is lower by ₩7.9 trillion than in the baseline.

The nuclear power reduction scenario can be thought of as a further reflection of MOTIE (2017) when considering the schedule of nuclear power capacity. One can also interpret the result of this scenario in the following way. When setting the nuclear power reduction scenario as the baseline, the original baseline scenario can be treated as a scenario in which market principles are applied to the nuclear power sector ceteris paribus. The implication is then as follows: if letting the market mechanism fully govern the nuclear power sector, there would be welfare gains partially from an increase in nuclear power generation and a decrease in gas-fired generation. This outcome mainly arises because nuclear power generation is less expensive than generation by other energy sources, resulting in overall welfare gains even if considering the negative impacts from nuclear accident risks.

D. Sensitivity Analysis

A sensitivity analysis is performed while changing major parameter values. Table 4 summarizes the variations of the parameters and the resulting welfare effects. To be specific, the analysis deals with changes in the discount rate, the length of the second stage, the level of demand elasticity, the R&D effect, the learning effect, and the appropriability rate, while also removing exogenous demand growth, second-stage efficiency improvements, and any increase in the second-stage GHG social costs, with 15 cases analyzed in total. It should be noted that the relative welfare rank among the six scenarios remains the same throughout the analysis. Specifically, from the highest welfare scenario, the order is as follows: optimal RPS, baseline RPS, renewable generation subsidy, nuclear reduction, past RPS, and repeal of RPS, where the magnitude of the welfare effect continues to vary.

TABLE 4

SENSITIVITY ANALYSIS

Note: 1) Welfare effects are changes relative to the no RPS scenario. 2) In the case of ‘low GHG social costs’, it is assumed that the second-stage discounted GHG costs of coal and gas are identical to those in the first stage.

There are several notable features. If we increase the discount rate from 7% to 9%, the present value of future benefits and costs become smaller; thus, the overall welfare effect gap between the scenarios is reduced compared to that in the baseline. On the other hand, if the discount rate is reduced to 5%, the relative gap becomes larger than in the baseline results. Note that reducing or increasing the number of years in the second stage has the same effect as raising or lowering the discount rate applied to the second stage, thereby yielding similar results compared to an adjustment of the discount rate.

Lowering the demand elasticity reduces the magnitude of the overall welfare effects for all scenarios, while increasing it results in larger welfare changes, where for the latter demand responses are more significant. The changes in welfare by varying the R&D effect, learning effect, and appropriability are lower compared to the former variations. The elimination of the natural demand growth in the second stage compared to the first lowers welfare for all scenarios, and removing the improvement in the renewable capacity factor also results in lower welfare. When assuming that the discounted second-stage GHG social costs are maintained at the level of the first stage, the magnitude of the welfare effects is lower than in the baseline for all scenarios.