Consumer cost effectiveness of CO 2 mitigation policies in restructured electricity markets

We examine the cost of carbon dioxide mitigation to consumers in restructured USA markets under two policy instruments, a carbon price and a renewable portfolio standard (RPS). To estimate the effect of policies on market clearing prices, we constructed hourly economic dispatch models of the generators in PJM and in ERCOT. We ﬁ nd that the cost effectiveness of policies for consumers is strongly dependent on the price of natural gas and on the characteristics of the generators in the dispatch stack. If gas prices are low ( ∼ $4/MMBTU), a technology-agnostic, rational consumer seeking to minimize costs would prefer a carbon price over an RPS in both regions. Expensive gas ( ∼ $7/MMBTU) requires a high carbon price to induce fuel switching and this leads to wealth transfers from consumers to low carbon producers. The RPS may be more cost effective for consumers because the added energy supply lowers market clearing prices and reduces CO 2 emissions. We ﬁ nd that both policies have consequences in capacity markets and that the RPS can be more cost effective than a carbon price under certain circumstances: continued excess supply of capacity, retention of nuclear generators, and high natural gas prices. S Online


Introduction
The US Environmental Protection Agency (EPA) has begun a rulemaking process to regulate greenhouse gas emissions from existing power plants through section 111(d) of the Clean Air Act [1]. This section requires states to meet federal standards through EPA approved State Implementation Plans (SIPS). SIPS may include 'market-based instruments, performance standards, and other regulatory flexibilities' [1]. One of the significant state-to-state differences is the presence or absence of organized electric power markets. Here we examine the cost effectiveness of CO 2 mitigation policies in PJM and ERCOT.
There appears to be a consensus among economists that a price on carbon is the favored policy mechanism for its efficiency [2][3][4][5][6]. As an example, Metcalf writes, 'For economists, the obvious choice is to move toward market-based environmental mechanisms that put a price on greenhouse gas emissions' [6].
However, policy-makers have unique perspectives when quantifying costs. Policymakers do not generally follow the perspective of neoclassical welfare or public economics that wealth transfers away from consumers are welfare neutral. On the contrary, such transfer payments are important to elected Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. and appointed officials; they are sensitive to costs from the perspective of their constituents-consumers. Federal and state administrations have explicitly cited increased prices for consumers as undesirable, using language such as '…ensure that the standards are developed … with the continued provision of reliable and affordable electric power for consumers and businesses' [1]. Lisa Jackson, EPA Administrator from 2009-13, echoed this perspective by stating that a key principle of the regulations will be to '…implement the most costeffective measures that do not burden small businesses and nonprofit organizations' [7]. State policy-makers generally view costs from the consumer perspective. A meta-study by Lawrence Berkeley National Laboratory (LBNL) on state renewable portfolio standards (RPS) quantified carbon abatement costs using, '… consumer costs, which often include wealth transfers to generators and do not necessarily reflect the true social cost of each state RPS policy' [8].
The assumption of neutrality towards wealth transfer payments is particularly important in restructured markets. Electricity is a commodity in restructured markets, where consumer costs are driven by wholesale payments quantified by market clearing prices. Carbon policies may either raise or lower market clearing prices and the differences have large transfer payment implications. Existing low carbon generators could receive a windfall profit from a carbon price. Carbon intense generators would lose a portion of their producer surplus to tax revenue (that we assume here is used for reduction of consumer taxes).
In order to estimate consumer cost and tax revenue effects of carbon mitigation policies in restructured markets, we examine two policies to reduce CO 2 emissions: a carbon price and renewable energy standards. For each policy, we observe how sensitive cost effectiveness is to the price of natural gas by varying the price from $4 to $7/MMBTU. We compare the differences in cost effectiveness if transfer payments away from consumers are considered neutral or a cost. We also examine how the change in profits in energy markets affects PJM's capacity market. In addition to estimating the costs in PJM, we examine ERCOT to see if our results are sensitive to a different mix of generators.
We find that from the social perspective, where wealth transfers are neutral, a carbon price is indeed the most cost effective mechanism. For consumers, however, an RPS may be more cost effective than a carbon price when natural gas is expensive (price to electric power generators of $7/MMBTU or more). Expensive gas requires a high carbon price to induce fuel switching and this leads to wealth transfers from consumers to low carbon producers. The RPS may be more cost effective for consumers because the added energy supply lowers market clearing prices and reduces carbon emissions. We find that both policies have consequences in capacity markets because they affect the profits of fossil generators. Renewables supply energy but supply very little capacity [9], and the RPS is more cost effective than a price on carbon for consumers only if existing capacity supply remains adequate in addition to high gas prices.

Methods
We use as a metric for cost effectiveness the cost per unit of reduced greenhouse gas emissions, dollars per tonne of CO 2 ($/tCO 2 ) [10]. We estimate cost effectiveness from two perspectives: the 'social perspective' and the 'consumer perspective'. We define the first perspective as the 'social perspective' because we assume that wealth transfers are neutral. Our metric for estimating cost effectiveness from two perspectives is shown in equations (1) and (2) below. The social perspective includes social costs (capital, O&M, fuel costs) but exclude wealth transfer payments (profits or taxes).
Social costs CO Emissions .
(1) 2 We define the second perspective as the consumer's perspective. From the perspective of consumers in restructured markets, costs are quantified by differences in wholesale payments net of any related change in tax revenue. Tax revenue is increased by a carbon price and decreased by renewable energy subsidies. It has been shown that the changes in tax revenue would not equitably affect consumers even if the tax revenue was redistributed to households because lower income households spend a larger share of their income on energy [6]. Here, we do not consider the issue of equitable tax distribution and assume the changes in tax revenue affect consumers equally. In this approximation, cost effectiveness of carbon mitigation policies can be estimated for consumers as: Cost effectiveness (consumer) Wholsale payments Tax revenue CO Emissions .
( 2 ) 2 To examine the difference of wholesale payments in energy markets under a carbon price and under an RPS, we created an hourly economic dispatch model of the generators in PJM and ERCOT. The dispatch model calculates marginal costs for all generators then dispatches the least expensive generators necessary to meet load on an hourly basis. All variable costs, including a price on carbon, were assumed to be passed on as marginal costs in the bids of generators. The market clearing price is set by the marginal generator, and all generators receive the market clearing price for that hour. The model might be thought of as using a load duration curve approach, because it does not take into account unit commitment.
Hourly load and hourly wind generation for 2012 were obtained from ERCOT and PJM databases [11][12][13][14]. Power plant fuel costs, heat rates, variable O&M costs, and carbon intensities for each region were obtained from Ventyx Velocity Suite [15]. We assumed that fuel costs remained constant except for natural gas which we varied from $4 to $7/ MMBTU. Marginal costs for nuclear and coal plants are shown in figure 2 below.
The dispatch model neglects transmission, ramping, and security constraints in addition to forced outages. Under some circumstances, these simplifications could mask appreciable increases in locational marginal prices if the combined effect of their inclusion were to severely limit supply. Exploring the nuances of any singular or combined limitation on supply is beyond the scope of this analysis. Rather, our intention is to demonstrate how consumers could be affected under distinct mixes of generators-PJM (diverse mix of generators with significant nuclear) and ERCOT (gas-heavy with signifcant coal) 4 .
On the consumer side, we make the assumption that demand is inelastic to price changes because the value is relatively small and uncertain [16][17][18]. There is also concern over how much of the reduction would occur because of interstate leakage [19]. Fischer et al showed that supply elasticity values can influence the cost of policies to consumers [20]. However, Fischer et al as well as other research [8] note that the elasticity values are uncertain and that policies would affect natural gas prices on the order of a few cents per MMBTU. Therefore, we neglect the long term supply elasticity values of fossil fuels.
Because we neglect consumer price elasticities, we emphasize that this research should not be interpreted as a net social welfare analysis. Our objective is to inform policymakers how cost estimates differ when viewed from the perspective of consumers given a wide range of plausible costs for renewables and natural gas.

Time frame and power plant turnover assumptions
Policy cost estimates inherently require uncertain assumptions over some arbitrarily chosen time horizon. We make the following simplifying assumptions.
We limit the analysis to the short term by assuming demand and the mix of generators in each region stays the same as it was in 2012. We assume demand remains constant because it has been from 2005 through 2012 [21]. Both PJM and ERCOT project peak demand to increase by only 1% annually from 2014 to 2024 [22,23].
With excess capacity and relatively flat demand, construction of new capacity is expected to be low [24]. Inexpensive shale gas further disincentivizes new power plants because of lower market clearing prices in energy markets. New power plants are not profitable at low carbon prices according to our dispatch model. Therefore, we model a mix of generators identical to that in 2012.
It is likely that older, less efficient power plants may retire in both PJM and ERCOT. In ERCOT, 10 GW of natural gas steam generators have already retired since 2005 [25]. In PJM, as much as 20 GW of coal plants are thought to be at risk of retirement due to pending environmental regulations [26]. In the supporting data (SD), available at stacks.iop.org/ ERL/9/104019/mmedia, we examine an alternate scenario in which 18 GW [26] of small, old coal plants are retired in PJM. We find that the high heat rates of these generators preclude them from greatly affecting energy markets; thus their exclusion does not substantially alter our findings.
Though the retirement of older power plants may not affect our estimates in energy markets, retirements may affect the capacity market. When new capacity is needed (due to retirements or demand increases), the decision will be based on the conditions of energy markets and capacity markets along with environmentally related incentives. After presenting our results for policies in energy markets, we examine how carbon policies may affect capacity market bids if either existing coal or new NGCC power plants are on the margins.

Carbon dioxide mitigation quantity
A price on CO 2 disadvantages coal fueled generators and causes other generators to be dispatched first; this is termed fuel switching. Our model estimates for each RTO the amount of carbon reduced due to a given carbon price (figure 1) over a baseline set by the CO 2 emissions output of our dispatch model at a given natural gas price and 2012 modeling data. Figure 1 shows that the effectiveness of the carbon price is dependent on the price of natural gas and the amount of gas capacity in each region. For each region, the line stops when market conditions induce either new NGCC plants or new wind plants.
Of the three regions examined, MISO has the least fuel diversity and the dominance of coal means that market clearing prices rise quickly with carbon prices. We performed the analysis for MISO but found that this research was not as applicable to this region because new plants are induced at such a low carbon price and the lack of fuel diversity limits wealth transfers. Methods and results for MISO are given in in the supporting data.

Transfer payment implications of a carbon price
Should gas prices reach $7/MMBTU, a higher carbon price would be necessary to induce new power plants or fuel switching. These price changes lead to transfer payments (figure 2). In PJM, the carbon price raises the market clearing price and leads to increased profits for low carbon generators. In ERCOT, a carbon price may cause transfers from producer surplus to tax revenue ( figure 2(b)). ERCOT does not have as many existing low carbon generators to take advantage of the carbon price. If coal is dispatched despite the carbon price, its producer surplus is transferred to tax revenue. The price of electricity is (to first order) unaffected, since gas sets the market clearing price. The changes in profits affect capacity markets and may cause some generators to extend or cease operations; we discuss this further below.

Renewable portfolio costs and transfer payment implications
A price on carbon raises market clearing prices; renewables lower it. Renewable generators increase energy supply with very low short-run marginal cost, push the energy supply curve to the right, and lower market clearing prices (SD figure  S2) [8,28]. The lowered market clearing prices decrease the profits of fossil generators [8,28]. Here, we make the assumption that these savings are passed on to consumers in the form of lower wholesale power prices.
We assume that whatever technology is used to meet the RPS has the same hourly production pattern as wind energy did in 2012 5 . We assumed the cost of the renewable energy to consumers is equal to its levelized cost of electricity (LCOE). Renewable energy is induced through a combination of revenue from bilateral power purchase agreements, renewable energy credits, and other subsidies. The sum of the revenue received by renewable energy developers would be approximately equal to the LCOE.
The US DOE estimates that 500 GW of wind are available at ∼$85/MWh or less, not including integration costs or subsidies [29]. We find that the most recently-available assumptions available would also yield a levelized cost of about $85/MWh: a capital cost of $1940/kW [30], a fixed charge factor of 12%, a capacity factor of 35% [30], and O&M costs of $25/kW-year [30].
Wind also has costs due to variability and transmission not typically included in LCOE estimates [31]. Utilities would pass these costs onto consumers, so we include them here. The 2012 DOE Wind Technology Market Report estimates variability costs to be in the range $2.5-$10/MWh [30]. For transmission costs unique to wind, LBNL performed a metastudy that found the cost of transmission to be $10-$15/MWh [32]. We add these costs, which utilities would pass onto consumers, to estimate a total levelized cost of approximately ∼$100/MWh of wind.
We examine wider bounds than the costs described above because of the wide range of costs found in literature [31,33] and the unpredictability of technology and subsidies. The federal government (sometimes) provides a production tax credit of $23/MWh [34]. In our results, we examine cost effectiveness if the added energy cost $80, $100, or $120/ MWh from the perspective of consumers. In PJM, low-carbon generators benefit from a price on carbon but do not change their order in the dispatch stack. In ERCOT, carbon intense generators that remain in the dispatch stack lose profit to tax revenue. 5 In the supporting data, we show that our conclusion is unchanged if the RPS production pattern is base-load.

PJM Energy market
In figure 3, we show the marginal cost effectiveness of policies in PJM's energy market. From the social perspective, where wealth transfers are neutral, our results indicate that a carbon price is (as expected) the most cost-effective option. As theory would suggest, the marginal cost of abatement is equivalent to the carbon price. If wealth transfers are neutral, the RPS would cost approximately ∼$40-$80/t CO 2 more than a carbon price.
From the consumer perspective, the most cost effective policy is dependent on market conditions. A carbon price is the most cost effective option for consumers at low natural gas prices and low carbon prices (less than ∼$15/t CO 2 ). With high gas prices, consumers may pay less per tonne offset with an RPS. A carbon price is not cost effective with high gas prices because the high carbon price necessary (figure 1) leads to a wealth transfer at the expense of consumers. The cost effectiveness of the RPS is improved by higher natural gas prices because the supply of renewable energy does more to suppress market clearing prices. In the next section we consider capacity markets.

PJM capacity market implications
Power plant bids in capacity markets are driven by fixed costs (PJM refers to these as 'avoidable costs' [35]) less profits made in energy markets [36]. Carbon dioxide mitigation policies affect capacity markets because they affect the profits of generators in energy markets. If generators increase their bids in capacity markets as a result of carbon policies, the cost to consumers can be appreciable [8].
Capacity markets are volatile [36][37][38]. For the PJM Base Residual Auctions from 2007 to 2017, the RTO resource clearing price has varied between $16-$174/MW-day [39]. This is a consequence of uncertain demand [38] and the steep supply curve of capacity market bids in PJM; it increases by over $300/MW-day for the last 10 GW offered [39].
Given the volatility of capacity markets, models of the market do not exhibit a high degree of accuracy. However, it is feasible to examine how policies affect the bids of plants that may be on the margin in capacity markets now and in the future-existing coal generators or new NGCC power plants.
Coal plants appear to be on the margins in the PJM capacity market, as 10 GW of coal plants did not clear in the 2016/ 2017 capacity market auction [39].
In figure 4, we show how capacity market bids of coal and NGCC power plants change as a result of carbon policies. Without revenue from energy markets, the avoidable costs of an existing coal fired power plant and a new NGCC plant are  approximately $160/MW-day and $370/MW-day, respectively [27,35]. Lower bids are submitted to capacity markets as generators earn revenue in energy markets. We estimate that revenues from energy markets 6 would allow the power plants to make bids of $105/MW-day and $350/MW-day, respectively 7 . We use The Brattle Group's estimates for the cost of new entry (CONE) in PJM for NGCC plants [27]. Coal plant performance is based on data from the 2012 PJM State of the Market Report [35] and marginal costs of existing power plants in our dispatch model [15]. The bids of generators change as profits are affected by carbon policies ( figure 4).
A CO 2 price has a larger effect on capacity bids than does an RPS because gas and coal plants switch their positions in the dispatch stack. The RPS does not change the order of the dispatch stack; it simply displaces the marginal generator. For a 20% reduction in CO 2 , the RPS raises the bid of an existing coal power plant by $15/MW-day whereas a carbon price raises it by $40/MW-day.
As long as existing generators satisfy capacity supply, the capacity market reaction from policies appears moderate and unlikely to change the decision of a policy-maker. Over a year, a $15/MW-day or $40/MW-day increase in PJM capacity prices would result in additional costs to consumers of $0.9B to $2.4B, respectively (with 165 GW of capacity in PJM [39]). We summarize as follows for a 20% reduction in CO 2 emissions. For the carbon price, the cost of mitigation increases from ∼$65 to $95/tCO 2 . For an RPS, the mitigation cost increases from ∼$75 to $90/tCO 2 .
Existing generators are unlikely to satisfy capacity supply forever, and renewables may force generators into premature retirement by reducing their profits. Let's suppose that renewables shortened capacity supply and increased capacity prices by $100/MW-day. If we add this cost to the 20% reduction RPS case, the mitigation cost increases from ∼$75 to $150/tCO 2 .
Another possible outcome of an RPS is that capacity prices remain low but renewables cause nuclear retirements. Three of Exelon's nuclear power plants did not clear in the most recent PJM capacity auction and are no longer considered 'in the money' [40]. Suppose that renewables pushed these generators (4.8 GW) to retirement and emissions reduced only 12% instead of 20% because of the retirements. The cost of mitigation would then increase from ∼$75 to $140/tCO 2 .
Policy-makers may favor a carbon price simply to increase capacity supply and decrease dependence on volatile capacity markets. A carbon price makes new gas plants and existing nuclear plants more competitive by increasing profits in energy markets. An RPS would increase dependence on capacity markets by undercutting fossil profits in energy markets [8]. In order for the RPS to be cost effective for consumers, markets must retain nuclear generators, attract new capacity, and do so without causing drastic increases to capacity market prices.

ERCOT
As shown in figure 2 above, the wealth transfer implications of a carbon price in ERCOT are different from PJM. ERCOT has few nuclear generators, large gas capacity, inexpensive coal, and no capacity market. The dominant wealth transfer effect of a carbon price is coal generators losing profits to tax revenue. Figure 5 below shows the two perspectives of carbon policies in ERCOT. Figure 5 shows that a carbon price in ERCOT is more cost effective from the consumer point of view than an RPS under the assumption that tax revenues from the CO 2 price accrue to consumers. In the expensive gas case, the RPS can be cost effective for consumers because it lowers the market clearing price in gas-heavy ERCOT.
ERCOT does not have a capacity market, so we cannot model whether coal generators would cease operations as a result of lost profits. The RTO expects capacity to become extremely tight with low gas prices and no capacity market [25]. The short capacity situation in ERCOT is expected to lead to larger price spikes, which in theory effectively act as capacity payments, though market changes have been suggested [41].
We summarize our results for PJM and ERCOT for a 20% reduction in carbon dioxide emissions from the baseline year of 2012 in table 1.

Discussion
We examined the cost effectiveness of a carbon price and of an RPS in restructured markets. From the social perspective, where wealth transfers are neutral, we find that a carbon price is (as expected) the most cost effective mechanism. This research adds a perspective that is relevant to policy-makers: in the short term, how will these policies affect consumers?
We find that the cost effectiveness of policies for consumers is strongly dependent on the price of natural gas and the characteristics of the generators in the dispatch stack. If gas prices are low (∼$4/MMBTU), a technology-agnostic, rational consumer seeking to minimize costs would prefer a carbon price over an RPS in both PJM and ERCOT. A relatively low carbon price is required to induce fuel switching when gas is inexpensive. The low carbon price minimizes wealth transfers and the marginal cost of mitigation to consumers is ≲$50/t CO 2 .
If gas prices are high ($7/MMBTU), for a 20% reduction of CO 2 in PJM, a consumer would find that an RPS mitigates CO 2 for an average of $60/tCO 2 , much lower than the average 6 Our estimates explore energy market revenues and neglect ancillary service market revenues. Revenue from ancillary services amounted to $6/ MW-day in PJM [27]. Revenues in ancillary service markets may increase due to the variability created by wind (RPSs in the PJM states require 14% renewables by 2026). However, because the amount is less than 10% of the capacity clearing price, we did not include it in this analysis. 7 These estimates assume that a coal plant has a marginal cost of $25/MWh per data from Ventyx Velocity Suite [15]. We assume that the NGCC plant has a marginal cost of $29/MWh per NGCC performance data from Brattle's CONE analysis [27] and a gas price of $4/MMBTU. The coal plant earns higher profits in energy markets and can make lower capacity market bids as a result. carbon price mitigation cost of $190/tCO 2 (SD figure S5). However, in ERCOT, the consumer would find that a carbon price is considerably less expensive than an RPS as a mitigation strategy because a portion of coal generators' producer surplus is converted to tax revenue (SD figure S6).
As long as existing generators can supply adequate capacity, the effect of policies on capacity markets is limited and would not affect a policy maker's decision. However, if new capacity is needed, a carbon price substantially reduces the capacity market bids of new NGCC plants. Policy-makers concerned with the low capacity supplied by an RPS could include other low carbon technologies that may have a higher LCOE [33] but that supply more capacity, such as coal with carbon capture and sequestration or nuclear.  Figure 5(a) assumes transfer payments are welfare neutral. Figure 5(b) assumes the consumer perspective. From the consumer perspective, so much producer surplus is transferred to tax revenue by the carbon price that consumers surplus may increase. This causes the negative cost effectiveness estimates. Figures showing alternate costs of wind are in the supporting data. If wind costs are $80/MWh, costs are approximately $30/tCO 2 less expensive than in the figure. If wind costs are $120/MWh, costs are approximately $40/tCO 2 more expensive than in the figure above. In the supporting data, we show the average cost effectiveness if a 20% reduction is required (SD figure S6).