Is natural gas the next coal? A framework for utilities and governments to think about the place of natural gas in the energy mix

Philippe Tordoir with Yara Beaini and Michael McGetrick

June 2021

M-RCBG Associate Working Paper Series | No. 174

The views expressed in the M-RCBG Associate Working Paper Series are those of the author(s) and do not necessarily reflect those of the Mossavar-Rahmani Center for Business & Government or of Harvard University. The papers in this series have not undergone formal review and approval; they are presented to elicit feedback and to encourage debate on important public policy challenges. Copyright belongs to the author(s). Papers may be downloaded for personal use only.

Mossavar-Rahmani Center for Business & Government Weil Hall | Harvard Kennedy School | www.hks.harvard.edu/mrcbg

Is natural gas the next coal? A framework for utilities and governments to think about the place of natural gas in the energy mix. Abstract

This study analyzes the role of natural gas in electricity generation and the implications for public policies by comparing the evolution of two hypothetical global electricity systems: one designed to reach zero-CO2 emissions by 2035 and another one with the objective to maximize global social welfare by instituting an international CO2 tax reflecting the estimated social cost of carbon. We show that the implicit CO2 cost of imposing a zero-CO2 emission target is an order of magnitude higher than the typically accepted social cost of carbon as a consequence of the non-dispatchable nature of renewable technologies. A tax approach would be much more cost-efficient and leads to a different mix in which natural gas still plays an important role as firm capacity provider. Public policies should better take into account the additional costs of integrating renewable technologies above 70% penetration level and dedicate more R&D efforts for alternative storage solutions, which may include batteries and green gases such as H2, before committing to replace natural gas. This paper provides a simple framework to structure a reflection around those issues.

1. Introduction

According to the International Energy Agency (IEA), the burning of coal accounts for 45% of the world’s total carbon emissions from fuel combustion, ahead of both oil (35%) and natural gas (20%). Coal is also the primary fuel used for electricity generation accounting for 38% of the global energy produced, followed at 23% by natural gas. There can be little substantial progress toward global emission reduction targets unless the world can begin to move away from its reliance on coal.

Pressure to reduce coal use has been increasing worldwide. A growing number of countries, including France and Germany, have decided to phase-out coal. Even China, where coal accounts for 65% of electricity generation, is investing heavily in renewable sources to reach carbon neutrality by 2060.

Natural gas emits about 50% less CO2 than coal for the same electricity output. The global reduction of power sector CO2 emissions in 20191 was partly driven by a demand shift away from coal to natural gas. This trend is especially pronounced in the US as a result of natural gas prices significant decline.2 However, while natural gas is cleaner than coal, it is still a fossil fuel. Therefore, it will eventually be phased out as the world strives to minimize or eliminate its carbon emissions.

The objectives of this study are to: (i) assess the future role of natural gas in global electricity generation and (ii) derive a framework that can be used by both the public and private sectors to analyze policy and investment options pertaining to their energy mix.

This study focused specifically on electricity generation and its carbon emissions. Section (2) below provides a global overview based on the latest IEA data. This data is then used throughout the paper as a basis for further analysis. The methodology and output of this study is intentionally simplistic. Its purpose is to distill complex academic data into easily understandable findings to aid in future decision- making by a broad audience.

1 reported by the IEA 2 “In mid-2016, the spot price for natural gas sold in the United States was just one-seventh of what natural gas had cost at its peak price a decade earlier”, Windfall, Meghan L. O’Sullivan ([1]) 1 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School This paper utilized two approaches to analyze the data. The first approach was hypothetical, independent of a specific policy mechanism, and consisted of modeling a global electricity generation trajectory that reaches a goal of zero CO2 emission by 2035. In this scenario, all fossil fuels were eliminated by 2035, leaving only renewables (on-shore and off-shore wind, solar photovoltaic ‘PV’), hydro, and nuclear. The second approach analyzed the effect of an international CO2 tax utilizing the estimated social cost of carbon. The resulting electricity generation profile was then driven by cost optimization rather than a CO2 emission target.

Despite the model’s simplicity, it provides valuable conclusions about the potential of natural gas in the future and the challenges associated with the integration of renewable technologies at higher penetration levels. While this paper references global-level data, country or regional specific conclusions can be made by altering numbers to match the geographic constraints of choice.

2. Electricity and heat generation: a global picture

According to 2018 International Energy Agency (IEA) data1, burning coal generates 38% of the world’s electricity. About half of that coal is burned in China.

Because coal generates twice as much CO2 as natural gas for the same energy output, shifting from coal to natural gas reduces emissions. However, burning natural gas still releases carbon dioxide and leakage during natural gas extraction, transportation, and storage is one of the major contributor to atmospheric methane.

China’s status as the world’s largest consumer of coal means that it must play a key role in the shift to cleaner energy sources.

1 www.iea.org 2 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School

Despite natural gas having a competitive levelized cost of energy (LCOE) globally, China has not begun to shift from coal to natural gas likely for political and economic reasons. China has very few domestic natural gas reserves and would need to import it.

Electricity generation per inhabitant is very uneven around the word but is critical to understanding likely future trends in global energy demand. This will be discussed later in the paper, but the below graph provides useful context for understanding current demand.

3 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School

3. Framework

This paper’s analytical framework uses two approaches. The first approach is hypothetical and consists of modeling a global electricity generation trajectory that reaches zero CO2 emissions by 2035. In this scenario, all fossil fuels are eliminated by 2035, leaving only renewables (on-shore and off-shore wind, solar photovoltaic ‘PV’), hydro, and nuclear. The second approach seeks to maximize global social welfare by instituting an international CO2 tax reflecting the estimated social cost of carbon. The resulting electricity generation profile is then driven by cost optimization rather than a CO2 emission target.

In general terms, the modeling developed for this paper builds upon a global electricity energy demand trajectory. It then meets that demand using an electricity generation mix determined primarily by policy goals (zero emission by a certain date – Approach 1) or market forces based on the relative Levelized Cost of Electricity (LCOE) for various sources, including a CO2 tax (Approach 2).

This basic modeling approach employed in this study does not provide enough granularity to directly analyze the effects of hourly/daily/seasonal fluctuations. However, these fluctuations are important to incorporate into any analysis of renewable sources, especially at higher penetration levels. The challenge of providing reliable energy at a reasonable cost increases significantly as renewable penetration level increases, especially above 70%.1

To simulate the impacts of intermittent renewable energy generation, we used the work of Ueckerdt et al. [2] and the resulting marginal system LCOE of variable renewable energy, presented in Reichenberg et al. [3], based on the European electricity system. The marginal system LCOE of variable renewable energy is the cost of substituting one Megawatt-Hour (MWh) of thermal electricity with one MWh of variable renewable energy at a given penetration level. The resulting “integration costs” includes both direct costs, like additional transmission and storage construction and indirect costs such as the decreasing returns. As the penetration level of renewables increases, the amount of energy

1 In this paper, the penetration level of renewables is defined as the percentage of solar and wind technologies out of the total generation. 4 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School provided to the grid for a given addition unit of generation capacity, referred to as the capacity factor, decreases. This is because of the intermittency issue associated with renewables and where new capacity is built. If the most appealing sites are already in use, additional capacity will inherently need to be built in less desirable locations. Capacity factor can also be explained as the proportion of time a given technology, such as a wind turbine, is producing energy. If the turbine is down for maintenance or it is not windy, the turbine is not producing energy which results in a lower capacity factor. Curtailment results when the energy generated excess demand and cannot be otherwise stored.

We have also compared the results from our approach to the outcomes of other studies in terms of capacity factor / curtailment of renewables.

The model and outcomes of the two approaches are described in the following pages.

4. First approach: targeting a zero-CO2 emission electricity generation target by 2035 Description of the model

The objective eliminate CO2 emissions from global electricity generation by 2035. In this scenario, all generation in 2035 comes from renewables (on-shore and off-shore wind, solar photovoltaic ‘PV’), hydro, and nuclear. The following sections show a generation profile that meets the 2035 objective, an analysis of CO2 generation from 2020-2035, and the trend of the LCOE (Levelized Cost of Electricity) during that period.

Electricity generation requirements

A projection for electricity generation was built using the following elements:

- Demographic evolution based on a 2020 publication in the Lancet [4] - An assumption of the evolution of electricity generation per inhabitant based on historical data taken from the International Energy Agency (IEA) for main consumption regions, i.e. the US, Europe, China, India, Russia, and the Rest of the World. o Yearly generation in the US is assumed to remain constant at 13.5 MWh/inhabitant; o Yearly generation in Europe is assumed to remain constant at 7.6 MWh/inhabitant; o Yearly generation in China is assumed to continue growing, reaching the level of Europe (7.6MWh/inhabitant) in 2025 and then remains constant; o Yearly generation in India is assumed to continue growing at its historical rate of 4.7% a year, reaching 2.5MWh/inhabitant in 2035; o Yearly generation in Russia is assumed to remain constant at 7.6 MWh/inhabitant; o Yearly generation for the rest of the world is assumed to continue growing at its historical rate of 0.8% a year, reaching 2.6MWh/inhabitant in 2035. - Additional electricity demand for electric vehicles (EV) and heat generation. o EV projections are based on the IEA 2020 Energy Outlook “Stated policies scenario”. o Per IEA data, heat generation has been nearly constant and is assumed to remain so at 4 million GWh annually.

The resulting electricity generation requirements evolve as per the chart below. Global electricity demand increases by 51% from 2019 to 2035. This equates to an increase of 2.7% per year which is approximately the same growth rate as the previous 13 years. A portion of that demand growth is due

5 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School to an increase in the used of electricity for heat generation and electric vehicle (EV) components. Without those effect, the projected annual growth would be 1.7% (30% overall over the considered period).

Electricity generation by sources

- Hydro generation is assumed to remain constant; - Nuclear generation is growing at 0.23% per year based on International Atomic Energy Agency outlook 19 (low scenario); - “Other sources,” mainly fuel oil and biofuels, are assumed to be slowly replaced by renewables; - Renewables (on-shore and off-shore wind, solar PV) are expected to grow at a rate that allows for all electricity not generated by hydro or nuclear to be generated by renewables by 2035. An “S” curve is proposed for that growth to reflect the fact that, as 2035 is approaching, the annual growth of renewables should converge with the global annual growth of electricity generation (plus the replacement rate of aging installations, not accounted for in this study). - On-shore wind, off-shore wind, and solar PV are treated separately in order to more accurately calculate the evolution of renewable LCOEs. Solar PV is expected to grow at the same rate as renewables globally, on-shore wind by half that rate, the rest being covered by off-shore wind. - Coal generation is projected using historical regional data from the IEA as detailed below:

▪ US: -4% per year ▪ Europe: -3% per year ▪ China: +1% per year ▪ RoW: +1% per year

6 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School - Natural gas generation then covers the remaining demand. It is assumed to be equally split between the US, Europe, and the Rest of the World for simplicity, each zone with a specific price level as described in appendix.

Levelized Cost of Electricity (LCOE – in 2019 USD/MWh)

LCOE is a key metric used to determine the relative cost competitiveness of each energy source. This section presents how the values were calculated and what the main underlying assumptions used in those calculations were.

LCOE measures the total expense of producing one MWh of electricity from a new project. It accounts for development, construction and equipment, financing, feedstock, operation, and maintenance costs. System LCOE

The following variables were used to calculate complete electrical system LCOE:

푇푒푐ℎ 퐿퐶푂퐸𝑖 : the “standalone” LCOE of a given technology (Tech) in year i. Technology can be coal, natural gas CCGT, on-shore wind, off- shore wind, solar PV, nuclear, or hydro.

푝 푚푎푟푔퐿퐶푂퐸푉푅퐸: the marginal system LCOE of variable renewable energy (on- shore and off-shore wind, solar PV). This represents the cost of integrating additional renewable energy, for a given penetration level p, as described in the following chapter.

푇푒푐ℎ 푉𝑖 : Energy (MWh) generated by a given technology (Tech) in year i

푇푒푐ℎ 푇푒푐ℎ 푇푒푐ℎ ∆푉𝑖 = 푉𝑖 − 푉𝑖−1 : Additional energy (MWh) generated by a given technology in year i compared to year i-1 (if positive)

In any given year i, the system LCOE is calculated by averaging the LCOEs for each technology weighted by their respective energy increase from the previous year (as the LCOE should reflect the impact of additional energy sources) as well as the marginal system LCOE:푆푦푠푡푒푚 퐿퐶푂퐸𝑖 = 푇푒푐ℎ 푇푒푐ℎ 푝 푉푅퐸 ∑푇푒푐ℎ 퐿퐶푂퐸𝑖 ×∆푉𝑖 +푚푎푟푔퐿퐶푂퐸푉푅퐸×∆푉𝑖 푇푒푐ℎ ∑푇푒푐ℎ ∆푉𝑖

푉푅퐸 푂푛−푠ℎ표푟푒 푤𝑖푛푑 푂푓푓−푠ℎ표푟푒 푤𝑖푛푑 푆표푙푎푟 푃푉 Where ∆푉𝑖 = ∆푉𝑖 + ∆푉𝑖 + ∆푉𝑖

푂푛−푠ℎ표푟푒 푤𝑖푛푑 푂푓푓−푠ℎ표푟푒 푤𝑖푛푑 푆표푙푎푟 푃푉 푉𝑖 +푉𝑖 +푉𝑖 And 푝 = 푇푒푐ℎ ∑푇푒푐ℎ 푉𝑖 Additional renewable capacity can be used either to satisfy the demand growth or to replace the decreasing share of fossil-fuel technologies.

7 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School 푇푒푐ℎ Technology LCOE (noted 퐿퐶푂퐸𝑖 above) The technology LCOE is a summation of 5 components:

- Capital costs - Variable Operations and Management (O&M) cost (excluding fuel) - Fuel costs - Fixed O&M cost - Transmission cost

This study utilizes the same assumptions as the 2020 Annual Energy Outlook of the US Energy Information Administration (US EIA). The following table summarizes those assumptions. All values are 2025 projections using 2019 US dollars:

Coal Natural Gas On-shore Off-shore Solar PV wind wind Ultra- Multi-shafts Technology supercritical CCGT 200MW 400MW 150MW 650MW 1083MW Capital costs 3661 954 1319 5446 1331 (USD/kW) Levelized O&M excl 4.48 1.86 0 0 0 fuel (USD/MWh) Fuel costs 17.79 25.02 0 0 0 (USD/MWh) Fixed O&M 40.41 12.15 26.34 109.54 15.19 (USD/kW-year) Transmission costs 1.17 1.2 2.8 2.65 3.59 (USD/MWh) Capacity Factor 85% 87% 40% 44% 29% Discount rate (real 4.3% 4.3% 4.3% 4.3% 4.3% after tax)

To calculate the resulting LCOE, capital costs and fixed O&M are converted to USD/MWh utilizing the capacity factor indicated in the table, the discount rate, and the assumed investment horizon.

The US EIA also published data for 2040, applying a learning rate between 2025 and 2040. We used both the 2025 and 2040 values projected by the EIA and assumed a linear trend between the two. We kept the LCOE constant between 2020 and 2025. A regional differentiation has been introduced through the cost of fuel for coal and natural gas, all other parameters are assumed equal for all regions.

The following charts represent the LCOE of the various technologies in 2025 and 2040 broken out by LCOE component. The investment horizon of new coal and natural gas capacity has been reduced to 10 years given the expectation that they would be out of service by 2035. The presented data is related to the US (other regions would differ through the cost of fuel).

8 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School

The increase of natural gas LCOE from 2025 to 2040 is driven by a change in fuel costs. The US EIA projects natural gas fuel costs will increase from 3.93 USD/mmbtu in 2025 to 4.81 USD/mmbtu in 2040 (in 2019 USD). This increase exceeds the expected decrease in natural gas capital costs.

Further details on LCOE assumptions are provided in the appendix.

푝 Integration costs (denoted as 푚푎푟푔퐿퐶푂퐸푉푅퐸 above) Renewable energy sources such as solar and wind are considered variable because the sun is not always shining, and the wind is not always blowing. As renewables assume a larger percentage of the total energy mix, also called penetration level, intermittent electricity generation increasingly becomes an issue. As a result, overcapacity is needed to ensure a reliable base load of energy can be supplied to consumers at all times. Because of this need for overcapacity, the marginal LCOE of the electrical 9 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School system increases as renewable penetration level increases. Ueckerdt et al. introduces the concept of “System LCOE” for variable energy sources as the sum of the traditional generation LCOE and their integration costs into the system, without factoring in variability. Integrations costs include:

(i) Balancing costs because of the uncertainty of variable energy sources requiring very short-term operating reserves (ii) Grid costs as additional transmission investments or additional grid constraints (iii) Capacity costs especially for peak-load periods (iv) Profile costs, the temporal generation profile of renewable sources does not match the load-profile, resulting in inefficient redundancy in the system.

This latest category is not necessarily a direct cost but should clearly be taken into account in an economic valuation. Further building on [2], [3] proposes an evaluation of integration costs, which in this context is equal to the marginal System LCOE of renewable sources (the cost of substituting one MWh of thermal electricity with one MWh of renewable electricity for a given level of penetration of renewables), including for high level of penetration, from 0% to 100%. Their model is based on Europe and we have assumed the results to be usable globally. The definition of the penetration level used in Reichenberg et al. excludes hydro power from renewables and from the total mix. We have adjusted the data to the penetration level definition we have used, i.e. 푝 = 푂푛−푠ℎ표푟푒 푤𝑖푛푑 푂푓푓−푠ℎ표푟푒 푤𝑖푛푑 푆표푙푎푟 푃푉 푉𝑖 +푉𝑖 +푉𝑖 푇푒푐ℎ ∑푇푒푐ℎ 푉𝑖 Using the study’s data, we have assumed a linear increase of integration costs from $30/MWh at 5% penetration to $70/MWh at 70% ($5/MWh increase per 10% additional penetration) and then doubling to $186/MWh at 95% penetration level. The costs tend to be infinite at 100%, reflecting the need for some baseload beyond variable sources such as nuclear or hydro.

10 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School Main outcomes Resulting generation profiles

11 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School In order to reach zero emissions by 2035, a high rate of renewable energy capacity integration onto the electrical grid is required. CO2 emissions

The following chart shows the increase in cumulative CO2 emissions from 2020 to 2035. The chart also includes a trajectory assuming the generation mix remains constant and a “no coal” trajectory that completely replaces all coal with natural gas immediately.

To put numbers in perspective, the above chart shows the remaining global CO2 budget for a two thirds chance to limit temperature increase to 1.5°C according to the Intergovernmental Panel on Climate Change (IPCC)1. The portion of that budget corresponding to the electricity and heat generation sectors is also depicted by the dotted line reflecting the 42% share of total emissions in 2018 as reported by the IEA. Levelized Cost of Electricity (LCOE) evolution

As indicated by the chart below, the global marginal LCOE increases by a factor 2.5 from 2021 to 2035 (in constant 2019 USD), climbing from $80/MWh to $200/MWh in this scenario. The LCOE increase is not linear and is primarily driven by the assumed increase in integration costs of renewable technologies at higher penetration levels.

1 The budget of 420Gt of CO2 published by IPCC is from the start of 2018. Assuming an annual CO2 generation of 33Gt (IEA), the remaining budget is 354Gt. 12 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School

Understanding LCOE and the increasing cost of renewables at higher penetration levels

The following chart represents the relative shares of the main LCOE components. The LCOE contribution of renewables increases in the first period because of the assumed growing share of higher cost offshore wind. After approximately 2027, the cost reduction resulting from learning overcomes the increased cost due to the addition of relatively expensive offshore wind capacity. This causes renewable LCOE to slowly decline.

By 2035, 65% of the LCOE is due to the integration costs, i.e. the non-dispatchable nature of renewable technologies.

13 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School These increased integration costs result from the decreasing returns of new units of renewables as their penetration level increases. The variable nature of renewables results in decreasing capacity factors and significant curtailment as their penetration level increases. To this point, capacity factor has been held constant. By factoring in the variation of integration cost, we are accounting for a varying capacity factor. This approach yields a decreasing capacity factor as penetration level increases. The chart below shows this relationship for on-shore wind. The trend is the same for off-shore wind and solar PV.

Utilizing this variable capacity factor, the following chart represents the equivalent LCOE components for natural gas and on-shore wind at different renewable penetration levels. The capacity factor of natural gas is assumed to remain constant to facilitate the comparison.

The system LCOE is then derived by calculating the weighted average of the above components, based on the annual electricity generation increase. The graph below continues the simplified analysis of on- shore wind and natural gas:

14 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School

This results in the following system LCOEs at different renewable penetration levels:

Total capital costs at 70% are twice those at 10%, and more than 4 times at 85%.

When renewables account for 85% of the total electricity generation above a hydro and nuclear baseline, , the resulting system is considerably oversized in terms of installed capacity. In the electricity demand trajectory that is proposed, the energy demand increases by 30% between 2020 and 20351.

Energy Demand Increase Required Additional Capacity Required Additional Capacity with Constant Capacity Factor with Reduced Capacity Factor 30% 110% 624%

The resulting installed power would be 7 times the current system capacity, which would imply either significant curtailment or improved storage capacities.

1 The same assumptions lead to a 55% energy increase by 2050. [5] compiled the results of 40 studies of electricity systems over different geographic scope and found out that the resulting demand increase is in the range 20%-120% by 2050 with a median at 52%. 15 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School Comparison with other studies

Integration costs increase significantly at high renewable penetration levels. This results in decreasing renewable capacity factors and leads to higher associated LCOEs. The following is a comparison between our conclusions and those of studies [5] and [6]

[5] is a compilation of 40 studies of electricity systems with varying geographical constraints, energy generation mixtures, and CO2 emission reductions targets between 80% and 100%. The main conclusions are that relying on a very high penetration level of non-dispatchable renewables:

1. Involves overbuilding total installed capacity by a factor up to 8 times peak demand to produce enough energy when solar and wind output are well below average 2. Results in a significant level of curtailment, reaching at least 40% of generated energy at 100% renewable penetration level 3. Results in an annual system cost at 100% renewables that is 3-5 times higher than the system cost at the current level of penetration 4. The cost and curtailment increases are non-linear with a significantly accelerated increase above 80% renewables

[6] examines what is required to maintain the New England electricity system reliability while reducing economy-wide emissions to net zero by 2050.

This study confirms that during weeks with prolonged low solar and wind generation, it becomes necessary to dispatch firm resources. The additional solar, wind (and storage) that would need to be built to serve load in such instances would be significant and would also lead to significant renewable oversupply during average generation weeks (curtailment up to 40%-50%). In a high-electrification scenario with zero-emission by 2050 (i.e. without thermal firm capacity), the required installed capacity is about 7 times higher than that in 2019. Half of that increase would need to be battery storage. Keeping a natural gas base capacity to optimize system costs

Given the significant increase in system cost for renewable penetration levels above 70% of the total energy mix, we analyzed the effects on system cost of maintaining 20% generation from natural gas. In this scenario, the 20% natural gas and 17% base-load capacity from hydro and nuclear account for 37% of the total energy mix. Renewables then cover the remaining 63%.

The LCOE increases to 137 USD/MWh in 2030 before going down to 122 USD/MWh in 2035. Compared with the zero-emissions system cost of 200 USD/MWh, maintaining 20% natural gas in the system leads to a 32% to 40% cost reduction. While CO2 emissions would not be reduced to zero in this scenario, they would significantly decrease from 15 billion to 3.5 billion tons per year.

The difference of LCOE between zero-CO2 emissions and 3.5 billion tons per year, applied to the energy generation projected in 2035, correlates to a price of 930 USD/ton of CO2 to remove the final 35 billion tons of CO2 emissions per year from the energy mix. This value is significantly higher than the typical range of proposed values for the social cost of carbon [50-100]USD/ton (see following pages for references). This shows that carbon emission abatement above 70% renewable penetration rapidly loses its cost effectiveness. If countries look to reduce their overall carbon emissions, complete decarbonization of the electricity sector is not currently be the most cost-effective way to do so.

16 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School One point of discussion should be the capacity factor assumption for natural gas plants: we have indeed taken the US EIA assumption of 87%, which is a very high number, especially in a scenario in which natural gas only plays a role of back-up firm capacity. The table below summarizes the outcome when using a capacity factor of 10% instead of 87%. The LCOE of natural gas is very much impacted (more than double) but the effect on the system LCOE and the conclusion on the implicit CO2 cost remain however limited.

System LCOE Implicit CO2 cost of a LCOE of natural gas (in (USD/MWh) zero-CO2 system 2040, USD/MWh) (USD/t) No natural gas 200 20% natural gas, 122 931 62.28 capacity factor 87% 20% natural gas, 133 800 128.22 capacity factor 10%

Potential impacts of batteries or firm H2 capacity

The generation model utilized thus far has not accounted for batteries or firm H2 capacity.

Batteries would reduce the need for constructing excess renewable capacity by storing energy that would have otherwise been curtailed. However, batteries are currently only capable of acting as a short-term storage solution. No large-scale, long-term (weeks or months) solutions are available. Current LCOEs in the range 100-150 USD/MWh remain comparable with the integration costs of this study. [6] concludes that unless enough batteries were added to double system capacity, batteries would not fundamentally change the above conclusion about the cost ineffectiveness of complete decarbonization of the electricity sector. The cost of batteries must be cut in half in order for the resulting system LCOE at higher renewable penetration levels to be comparable with a 20% natural gas scenario. Research and development efforts must yield a more cost-effective way to mitigate the issue of renewable intermittency before removing all fossil fuels from the energy mix.

Another possible alternative is to add a certain level of firm capacity based on H2 generated from renewable sources. The two main drawbacks of this approach are the low capacity factor H2 production plants, such as electrolyzers, and the high cost of H2. As a sensitivity, we ran a scenario with H2 accounting for 10% of the energy mix. We assumed the H2 was priced at 30 USD/mmbtu and was burned in CCGTs with a capacity factor of 10% (from [5]). The resulting LCOE in 2035 was 157 USD/MWh. This is equivalent to an implicit CO2 cost (vs. 20% natural gas) of 346 USD/t.

H2 prices are expected to decrease significantly in the future, potentially reaching 12 USD/mmbtu in 2050 ([13]). The implicit CO2 cost would decrease to 250 USD/t, still above the social cost of carbon typical range but in a much more reasonable measure.

17 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School 5. Second approach: system cost optimization with the inclusion of a CO2 tax monetizing the social cost of carbon.

In this approach, we consider that the emission of CO2 is a global welfare problem with a quantifiable cost to society. By imposing a CO2 tax in line with that social cost of carbon, the system should reach a global welfare optimum.

Research into the Social Cost of Carbon has yielded a wide range, mostly falling between 50-100 USD/ton of CO2. Below is a summary of some of those studies1, all of which 2019 dollar values for their projected values:

• US Interagency Working Group on the Social Cost of Carbon of Greenhouse Gases: from 53.3 USD/t of CO2 in 2020 to 87.6 USD/t in 2050 [7] • 101.24 USD/t (Pindyck, [8]) • Between 80 and 100 USD/t for experts with a high level of confidence (200 USD/t or higher with a reduced level of confidence) – Pindyck [9] • Between 44.4 USD/t for 2020 to 122 USD/t for 2050 (William Norhaus [10]).

In this approach, the objective is to choose the most cost-efficient mix of natural gas and renewables for new energy capacity and allow market forces to phase out fossil fuels. The system costs for a given year are calculated by applying the system LCOE, as previously defined, to the total energy generated that year (although the LCOE is calculated on the basis of new energy, it is assumed to impact the whole system). The LCOE of fossil-fuel technology includes an additional component: the CO2 tax. In the case of natural gas, the assumption is an emission of 0.419 tons of CO2 per MWh (US EIA), leading to a CO2 LCOE component of 20.95 USD/MWh for a CO2 tax of 50 USD/MWh.

It is further assumed that:

- There is no off-shore wind in the renewable mix: Its LCOE is high relative to other renewables and natural gas. The LCOE for renewables is calculated here as the average of on-shore wind and solar PV. - To simplify the model, coal is completely replaced by natural gas in year 1. While this scenario is unlikely for non-economic reasons, the tax makes the natural gas LCOE advantage over coal even more pronounced and new coal investment would not be viable. - The maximum amount of new renewable energy is assumed to be the S curve proposed in the first approach.

The optimal energy blend was determined using the method described in appendix.

The resulting renewable penetration level is represented below.

1 2019 USD, based on CPI Inflation calculator of the US Bureau of Labor Statistics 18 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School

When no CO2-tax is applied, natural gas displaces coal but renewables remain below 10%.

At 50 USD/t, renewables increase to around 45% penetration by 2046 and then relative growth stabilizes. At 100 USD/t, renewables quickly reach 60% by 2030 before relative growth stabilizes around 75% by 2046. A much higher tax level would be required to reach a zero CO2 emission system (above 300USD/t). The increase in penetration level between 2030 and 2045 is in fact driven by the natural gas price assumption: as this price is assumed to increase, the competitiveness of natural gas decreases overtime, giving more space for renewables despite their increasing integrations costs.

If a globally implemented carbon tax is between 50-100 USD/t, our analysis shows that that the renewable penetration level should reach 45% to maximum 75%, keeping ~5-35% natural gas to cover energy demand. The tax that each country would require to reach complete decarbonization would vary largely based on the LCOE of renewables relative to the LCOE of fossil fuels. However, the tax would need to be in the multiple hundreds of dollars per ton at a minimum.

19 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School 6. Comparison of both approaches

The relationship between a target renewable penetration level and associated CO2 costs found in our and other studies is shown on the chart below.

Typical Social Cost of Carbon range (50-100 USD/t)

Estimate by economists (Pindyck survey [10])

Estimate by climate scientists (Pindyck survey [10])

The implicit CO2 cost of a zero-emission system by 2035 without firm/dispatchable capacity (930USD/t) is completely out of the range of values typically considered as the social cost of carbon. This result is in line with the outcome of the New England study [6]. Due to their variability, renewable energy sources cannot reliably supply energy to the grid at all times. As a result, significant over capacity must be constructed to compensate which exponentially increases the cost of renewables above 70% penetration. Without any significant advances in battery storage, the addition of batteries would not make building a 100% renewable energy sector economical.

Keeping a minimum level of H2 firm capacity, produced from renewable sources, brings down the implicit CO2 cost down to 350 USD/t. While this figure is significantly less than the estimated 930USD/t with no firm capacity, it is still more than three times more expensive than the typically proposed social cost of carbon range proposed.

20 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School 7. Conclusions The role natural gas

This high-level study shows that keeping a certain level of natural gas as additional firm capacity above hydro and nuclear allows for significant CO2 emission reduction while maintaining a more reasonable system price. For example, the system price reached by keeping in the energy mix 20% natural gas would be 30% to 40% less expensive than a completely carbon free system. The cost increase is basically driven by the non-dispatchable nature of renewables, imposing to build overcapacity to compensate the lack of firm capacity. Battery-storage could provide an alternative to fossil fuel-based firm capacity but at current prices and technologies, they would not be cost effective and would only be short-term storage solutions (a few hours). Research and development that leads to a reduction in battery cost and that yields a cost-competitive technology that can store energy for longer than a few hours would be able to more substantially reduce the cost of complete decarbonization of the electricity sector. A more promising approach could be to switch firm capacity provided by natural gas to “green gases” such as H2 produces from renewables. Costs are still prohibitive today but the resulting implicit CO2 costs is more reasonable and H2 would provide long term storage capability similar to natural gas. R&D is also paramount to make such an alternative cost efficient.

A major assumption in the study is natural gas price increase over time from 3.9 USD/mmbtu to 4.8 USD/mmbtu between 2025 and 2040 (in 2019 USD). One could argue that further R&D in the sector and development of shale gas beyond the US could lead to more stable or even decreasing prices. With a 3.0 USD/mmbtu price in 2040, a 50USD/t CO2 tax would not lead to more than 15% renewables, and 60% with a 100USD/t tax.

Natural gas is essentially methane, a fossil fuel. While it emits about half of the CO2 as coal per MWh, it still releases CO2. Additionally, leakage between the extraction of natural gas until it is burned causes methane to be released to the atmosphere. Methane does not remain in the earth’s atmosphere for as long as CO2, but it causes an 80 times greater warming effect when it is present, typically around 20 years. 16.5% of the methane in the atmosphere comes from the oil and gas industry and the IEA evaluates that 75% of the methane released could be avoided with technologies available today [11]. While natural gas is needed to keep system LCOE costs to a reasonable level during the transitioned to green energy, addressing methane leakage is necessary to slow global warming. On public policies

The comparison of the two proposed approaches in terms of resulting CO2 cost has important implications for public policies. Setting a zero-emission target by a certain date with no additional firm capacity beyond the current nuclear and hydro basis would require a social cost of carbon much higher than usually accepted, and consequently is unlikely to be successful. A CO2 tax that reflects the social cost of carbon would use market forces to phase out of fossil fuels by making them less cost- competitive but would be ineffective unless set high enough.

However, many countries have elected to set emissions reductions targets. Governments should take into account the increasing costs of renewable energy at higher penetration levels to achieve the same grid reliability. Setting a zero-emission goal in the near to medium-term would require large-scale economic inefficiencies at high cost to government. As shown in this study, the CO2 tax rate necessary to completely decarbonize the electricity sector is indeed far above the value proposed by economists or climate scientists and keeping a small amount of natural gas as firm capacity greatly reduces the overall system cost because it mitigates the issue of renewables intermittency. This paper also

21 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School advocates the need for increased R&D efforts to find appropriate alternatives to fully displace natural gas in the longer term.

This study takes a global approach without taking into account local/regional specificities or geopolitical implications. A global CO2 tax that is universally accepted is highly unlikely (for an interesting discussion on how the necessity for a global approach and how to reach it, see [12]). The paper provides however a framework for analysis of decarbonization strategies to better understand their cost implications and evaluate their feasibility, which can be adjusted to local realities.

References

[1] Windfall: How the new energy abundance upends global politics and strengthens America’s power, Meghan L. O’Sullivan, Simon & Schuster, 2017

[2] System LCOE: What are the costs of variable renewables?, Falko Ueckerdt, Lion Hirtha, Gunnar Luderera , Ottmar Edenhofera, Energy 63, 2013

[3] The marginal system LCOE of variable renewables – Evaluating high penetration levels of wind and solar in Europe, Lina Reichenberg, Fredrik Hedenus, Mikael Odenberger, Filip Johnsson, Energy 152, 2018

[4] The Lancet, Fertility, mortality, migration, and population scenarios for 195 countries and territories from 2017 to 2100: a forecasting analysis for the Global Burden of Disease Study, published on-line 14 July 2020

[5] Getting to Zero Carbon Emissions in the Electric Power Sector, Jesse D. Jenkins, Max Luke, Samuel Thernstrom, Joule 2, December 19, 2018, Elsevier Inc.

[6] Net-Zero New England: Ensuring Electric Reliability in a Low-Carbon Future, jointly prepared by Energy+Environmental Economics and Energy Future Initiative, November 2020

[7] Valuing Climate Damages: Updating Estimation of the Social Cost of Carbon Dioxide, National Academies of Sciences, Engineering, and Medicine 2017, Washington, DC: The National Academies Press. https://doi.org/10.17226/24651

[8] Coase Lecture—Taxes, Targets and the Social Cost of Carbon, Robert S. Pindyck, Economica, London School of Economics and Political Science, 2017

[9] The social cost of carbon revisited, Robert S. Pindyck, Journal of Environmental Economics and Management, Elsevier, 2019

[10] Revisiting the social cost of carbon, William D. Nordhaus, Proceedings of the National Academy of Sciences, 2016

[11] The other greenhouse gas, The Economist, April 3rd 2021.

[12] The Climate Club, William Nordhaus, Foreign Affairs, May/June 2020

[13] The Future of Hydrogen, EIA, 2019

22 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School APPENDIX 1: LCOE additional details

In connection to its Annual Energy Outlook 2020, the US Energy Information Administration (US EIA) publishes significant details about its LCOE calculation, which have been used to reconstruct data that could be adjusted to specific investment horizons and fuel costs.

The following data has been used from the US EIA publications, for 2025 and 2040:

- Capital costs, fixed O&M costs, variable O&M (with and without fuel), transmission costs - Capacity factors - 30-year investment period - 4.3% real after-tax WACC

The resulting LCOE, as per the formula below, US based and in 2019 USD, are as follows

2019 USD/MWh LCOE 2025 LCOE 2040 Ultra-supercritical coal 58.35 56.09 Combined cycle 37.18 42.10 Nuclear 73.56 67.28 On-shore wind 32.89 28.20 Off-shore wind 116.01 81.84 Solar PV 40.98 33.68 Hydro 47.43 47.70

퐶푎푝푖푡푎푙 퐶표푠푡 + 푓푢푒푙 푐표푠푡푠 + 푣푎푟푖푎푏푙푒 푂&푀 + 푓푖푥푒푑 푂&푀 + 푡푟푎푛푠푚. 푐표푠푡푠 ∑30 푡 푡 푡 푡 푡 푡=1 (1 + 푟)푡 퐿퐶푂퐸 = 퐺푒푛푒푟푎푡푖표푛 ∑30 푡 푡=1 (1 + 푟)푡 푈푆퐷 퐶푎푝푖푡푎푙 퐶표푠푡푠 = 퐼푛푠푡푎푙푙푒푑 푝표푤푒푟 (푀푊)푥 푐푎푝푖푡푎푙 푐표푠푡푠 ( ) 푡 푀푊 푓푢푒푙 푐표푠푡푠푡 = fuel costs in year t

푣푎푟푖푎푏푙푒 푂&푀푡= variable O&M costs in year t

푓푖푥푒푑 푂&푀푡 = fixed O&M costs in year t

푡푟푎푛푠푚. 푐표푠푡푠푡 = levelized transmission costs in year t r = discount rate

퐺푒푛푒푟푎푡푖표푛푡 = generation (MWh) in year t = Power (MW) x 365 days x 24 hours x Capacity Factor

Although the investment period is assumed the same for all technologies (30 years), nuclear and hydro plant could be assumed to have longer life. As per the US Office of Nuclear Energy, the average age of nuclear plants in the US is already 40 years old. The US EIA mentions 50 to 100 years for hydro plants.

From the original US EIA methodology, the learning impact has been kept but additional 3% WACC on coal investment and specific tax subsidies for renewable technologies have not been included.

In our model, it has been further assumed that:

23 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School - Additional electricity generated from coal and natural gas would come from new investments with an investment horizon of 10 years instead of 30, impacting the 2025 LCOE as follows:

LCOE 2025 (2019 USD/MWh) 30-year 10-year Ultra-supercritical coal 58.35 90.39 Combined cycle 37.18 45.34

- Main regions are differentiated through the cost of fuel. o For coal, as per 2019, coal indices as follows (source BP Statistical Review 2020, https://www.bp.com/content/dam/bp/business- sites/en/global/corporate/pdfs/energy-economics/statistical-review/bp-stats- review-2020-coal.pdf ▪ US: 57.16 USD/t (US Central Appalachian spot price index) ▪ Europe: 60.86 USD/t (Northwest Europe marker price) ▪ China: 85.89 USD/t (China Qinhuangdao spot price) ▪ RoW: 80.81 USD/t (Asia marker price) o For natural gas (source BP Statistical Review 2020 https://www.bp.com/content/dam/bp/business- sites/en/global/corporate/pdfs/energy-economics/statistical-review/bp-stats- review-2020-natural-gas.pdf) : ▪ US Henri Hub: 2.53 USD/mmbtu ▪ Europe TTF: 4.45 USD/mmbtu ▪ Japan Korea Marker: 5.49 USD/mmbtu

The corresponding ratio vs. US will be applied to the LCOE, the US based fuel assumptions being as follows:

2019 USD/MWh LCOE 2025 LCOE 2040 US coal price 17.79 17.69 US natural gas price 25.02 30.66 (3.93USD/mmbtu) (4.81USD/mmbtu)

The resulting regional LCOEs are:

Ultra-supercritical coal LCOE 2025 LCOE 2040 (2019 (2019 USD/MWh) USD/MWh) US 58.35 56.09 Europe 59.50 57.24 China 67.29 64.99 RoW 65.71 63.41

24 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School Combined cycle LCOE 2025 LCOE 2040 (2019 (2019 USD/MWh) USD/MWh) US 37.18 42.10 Europe 56.17 65.36 RoW (based on Asia) 66.45 77.91

- The LCOE pre-2025 is assumed to be equal to the 2025 LCOE, which then evolves linearly until the 2040 value.

25 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School APPENDIX 2: system optimization with a C02 tax

Equation variables are define as follows:

푆푦푠푡푒푚 푉𝑖 : Total energy generated in year i 푆푦푠푡푒푚 ∆푉𝑖 : Required additional energy of the system for year i (i.e. system growth)

푅퐸푆 ∆푉𝑖,푚푎푥: Maximum available additional energy from renewables for year i

푅퐸푆 ∆푉𝑖 : Additional energy generated by renewable energy in year i

푁퐺 ∆푉𝑖 : Additional energy generated by natural gas in year i

푅퐸푆 퐿퐶푂퐸푝,𝑖 : LCOE of renewable energy, including integration costs for penetration level p, in year i

푁퐺 퐿퐶푂퐸𝑖 : LCOE of natural gas, including the CO2 tax in year i

푅퐸푆 푅퐸푆 푁퐺 푁퐺 푆푦푠푡푒푚 퐿퐶푂퐸푝,𝑖 ×∆푉𝑖 +퐿퐶푂퐸𝑖 ×∆푉𝑖 퐿퐶푂퐸𝑖 = 푅퐸푆 푁퐺 : System LCOE in year i ∆푉𝑖 +∆푉𝑖

푆푦푠푡푒푚 푆푦푠푡푒푚 푆푦푠푡푒푚 푐표푠푡푠 = 퐿퐶푂퐸𝑖 × 푉𝑖 : Total system costs

As an example, A CO2 tax of 50 USD/t would increase the LCOE of natural gas generation by 20.95 USD/MWh, from 53.3 USD/MWh (average between the three considered regions, 30-year horizon) to 74.2 USD/MWh. As long as the LCOE of renewables (on-shore wind and solar PV in this case), including the integration costs for the resulting penetration level, is below the LCOE of natural gas, the full amount of available new renewable energy is used. However, this is only the case up to a penetration level of 15%. Above that level, the LCOE-minimizing energy mix includes natural gas. The LCOE of natural gas is assumed to grow over time driven by an increase in fuel cost. This is what drives the continued, slow increase of the relative share of renewable. If natural gas fuel costs actually decrease, the renewable penetration level would remain fixed at or below that level.

26 Philippe Tordoir Mossavar-Rahmani Center for Business and Government Senior Fellow Harvard Kennedy School