Assessment of Current Technologies for and Policies Supporting Increasingly Clean Electric Power Generation
The United States has made significant progress in reducing air pollution and its harmful effects since pollution control laws such as the Clean Air Act (originally passed in 1963, with major amendments in 1970, 1977, and 1990) were introduced. “Killer fog” in America is, at present, a thing of the past. Tragedies such as the Donora smog of 1948 and the “Great Smog” of 1952 that killed thousands of people are essentially unheard of in developed nations. Acid rain and even the once-famous smog in Los Angeles have significantly dissipated. Notwithstanding the measured decreases since the 1960s, however, pollution from the production of electric power continues to cause tangible harm, nor does the price of electricity currently include all of the societal costs of electricity generation.
A 2010 National Research Council study, for example, found that air pollution from coal-fired electric power plants in the aggregate still caused significant harms to human health, including, among others, asthma and premature deaths (NRC, 2010b). These harms arise from sulfur dioxide (SO2), oxides of nitrogen (NOx), particulate matter (PM2.5 and PM10), ammonia (NH3), and volatile organic compounds (VOCs), referred to collectively as criteria pollutants as they are regulated under the Clean Air Act. The 2010 National Research Council study estimates that in 2005, the emissions of criteria pollutants from coal-fired power plants caused damages costing, on average, $0.032/kilowatt hour (kWh) of electricity generated. The human health harms from all coal-generated electricity thus cost about 33 percent of the value of all electric power produced that year.1 The 2005 emissions from gas-fired plants
1The National Research Council (2010b) study reports damages for the year 2005 but in 2007 dollars. The average retail price of electricity in the United States in 2005 was $0.0814/kWh (EIA, 2007). The Bureau of Labor Statistics’ Consumer Price Index
caused human health damages costing approximately $0.0016/kWh of electricity generated, representing about 2 percent of the average retail price of all electric power sold that year.
Electric power plants also produce 39 percent of all U.S. emissions of greenhouse gases (GHGs) (which trap heat in the earth’s atmosphere)—the largest share of any source (EPA, 2016). Translating GHG emissions into climate-related damages depends on estimates of damages per ton of carbon dioxide (CO2) equivalents. The above NRC (2010b) study estimates the climate-related damages to be 1.0-10.0 cents per kWh of electricity produced by coal-fired plants and 0.5-5.0 cents per kWh for natural gas-fired plants, corresponding to damages of $10-100 per ton of CO2 equivalents.
Reducing emissions further to ameliorate these harms will require a technological shift to increasingly clean—that is low- or nonpolluting—technologies for the generation of electric power. The magnitude of ongoing harms, including those likely due to climate change, makes it imperative to effect this shift as quickly as is efficient. Such increasingly clean technologies rely either on non-fossil fuel sources, such as wind, nuclear, or solar, or on “tailpipe” solutions—technologies that capture or otherwise prevent emission of the pollution from fossil fuels. Effecting this technological shift will in turn require ensuring that newly built generating assets (power plants) are increasingly clean (low- or nonpolluting) compared with those currently operating or recently retired. This means not only building increasingly clean power plants in response to new demand, but also encouraging the retirement of more polluting assets in favor of those running on increasingly clean technologies. The latter strategy is particularly important given that new asset builds in response to demand are likely to remain small. Although electricity demand in the United States continues to grow, the rate of increase has been in secular decline since the 1950s (see Figure 2-1). Consequently, it is reasonable to expect that most new power plants in the United States will be built to replace retiring plants rather than to increase total generating capacity in response to rising demand (EIA, 2015a).
Two factors—inaccurate market prices and the large amount of capital required to build a power plant—have led to a bias in the current mix of power plants in the United States2 in favor of higher-polluting technologies (see Figure 2-2).
2 The same is largely true in other countries around the world as well.
The first factor is that delivered electricity prices do not incorporate the full cost of the harms from the pollution caused by power plants. Because the cost of power plant pollution is not built into the cost of construction, power producers have tended to build more of these plants than they otherwise would have done. And because the delivered price of electricity also does not incorporate the full costs of pollution, end-users consume more electricity from these sources than they otherwise would.
The second factor is that power plants are expensive to construct, requiring large amounts of up-front capital. Such high costs take many decades to fully amortize. Once these costs have been fully amortized, the cost of operating a plant decreases and operating profits increase. Firms may thus have a strong financial incentive to keep a plant operating as long as possible, depending on how the state regulator sets retail rates for electricity (see Chapter 6 for more detail on the ratemaking process). Therefore, retirement of currently operating, higher-polluting plants might be unlikely even if the current price of electricity were to be corrected to include the costs of pollution. Thus power-generating assets are typically kept in operation for 40-50 years, and often even longer. It is important to note this fact when considering the long-term impact of new power plants; choices made today can have pollution consequences for decades.
When the financial incentive to keep power plants operating is combined with low market prices for fossil fuels and other factors, it comes as no surprise that most new plants built over the past 30 years have been powered by fossil fuels. As seen in Figure 2-3, from 1989 to 2011, more fossil fuel plants were built than any other type. This figure also shows the likely impact of policies on new plant builds. The Clean Air Act amendments of 1990 created an SO2 trading system, effectively a price on SO2, to help diminish the impacts of acid rain resulting from power plant pollution. Coal-fired plants produce more SO2 per kWh of generated electricity relative to natural gas-fired plants, so it is not surprising that from 1991 to 2011, most capacity additions were natural gas-fired plants. Figure 2-3 also shows increasing construction of new wind and solar facilities following the increase in tax subsidies for these facilities in 2005.
Looking to the future, most new plants are expected to continue to be predominantly fossil fuel-powered, with these capacity additions being greater than they would be if the market reflected the true costs of pollution. Since the market price does not reflect the full costs of pollution, government policies are required to ensure prices that more accurately reflect actual costs. Given such policies, production and consumption will be closer to its efficient and socially optimal quantities.
In light of the historic low natural gas prices at the time of this writing (2016), most new plants projected to be built through 2040, like those built in recent decades, are expected to be natural gas-fired. For example, according to the Energy Information Administration’s (EIA) projections from its National Energy Modeling System (NEMS)—based on assumptions about future fuel prices and expiring tax subsidies for renewable sources such as wind and solar—new builds may be primarily wind and solar for a few years, but will be predominantly natural gas as the tax subsidies for new wind generating facilities decline through 2019.3 Notably, EIA and other forecasters expect very few new plants to be powered by nuclear fuel, currently the largest source of nearly emissions-free electricity.
In its Annual Energy Outlook 2016, EIA projects total installed electricity generation capacity through 2040 (EIA, 2016a). As seen in Figure 2-4, those projections include an approximately 17 percent increase in total installed capacity between 2016 and 2040, with much of that increase occurring after 2030. The mix of capacity types is expected to change as well. EIA projects that the share of coal will decrease from 29 to 18 percent, mainly before 2020, while that of renewables will increase from 17 to 29 percent. Natural gas is projected to fluctuate slightly until 2020 and then remain stable at 43 percent, and nuclear to decrease from 10 to 8 percent.
As of 2014, emissions from power plants of SO2 and particulate matter 10 microns or less in size had decreased by 80 percent and of NOx by 65 percent relative to their levels at the time of the Clean Air Act amendments of 1970 (EPA, 2015). Even at those decreased levels, however, these pollutants are known to cause harms to human health, as discussed earlier (NRC, 2010b). On the other hand, emissions of GHGs due to electric power generation rose by a bit more than 60 percent during the same period (EIA, 2016a, Table 12.6). Meanwhile, CO2 emissions per megawatt hour (MWh) of electricity produced decreased modestly (Figure 2-5) as a result of improvements in the efficiency of coal plants in the 1950s and 1960s, the growth of nuclear power, and a partial switch from coal to natural gas and wind power in the late 2000s and early 2010s.
3Chapter 7 provides a more detailed discussion of these tax and other subsidies. Schedule available at http://energy.gov/savings/renewable-electricity-production-tax-credit-ptc (DOE, n.d.-c).
To understand better the barriers to greater adoption of increasingly clean electric power generation technologies—that is, to understand why power producers are likely to choose to build fossil fuel-powered plants over plants with carbon capture technology or those powered by wind or solar energy—the committee took an in-depth look at the technology readiness and cost of currently available cleaner technologies.
A key first step in understanding the barriers to market adoption for low- and no-emission technologies is assessing their readiness to be incorporated into existing infrastructures. Technologies that can readily and easily be incorporated into the existing electric power grid and associated infrastructure are much more likely to be adopted and utilized. There currently exist a wide range of increasingly clean electric power generation technologies that can produce lower or no emissions when used. The committee assessed the technology readiness of the most promising of these technologies in each of the following categories:
- Renewable power generation—These technologies focus on the generation of electricity from wind, solar, biomass, geothermal, and hydropower sources. They include, for example, advanced and improved wind turbines, photovoltaic (PV) devices, and enhanced geothermal power generation. The committee also included in its assessment technologies whose deployment would enhance the ability of the grid to host increasing amounts of renewable power production, such as storage technologies (including batteries) since improved storage can support variable power generation from renewables.
- Advanced fossil fuel power generation—These technologies focus on improving the pollution control technologies of coal- and natural gas-fired power plants, such as advanced carbon capture and storage. The committee also included water treatment technologies since treating cooling water is a significant obstacle to the construction of new thermal plants (including nuclear plants).
- Nuclear power generation—This category includes new and next-generation nuclear technologies and the development of cost-effective technologies that can maximize the use of existing nuclear plants.
- Electricity transmission and distribution—This category includes technologies with the potential to reduce losses from and increase the efficiency of the transmission and delivery of electricity to end-users.
As much as 11 percent of all electricity generated is lost during transmission and delivery (Jackson et al., 2015).
- Efficient electrical technologies for buildings and industry—This category includes technologies being deployed and developed to reduce building energy needs and energy used in industrial processes.
The detailed assessment of each of these technology categories in 4 of promising technologies in that category in 2016, 2020, and 2035, if estimates were available; and associated technological and commercialization barriers. Table 2-1 summarizes the 2016 TRLs of these technologies.
To understand impediments to the deployment of increasingly clean energy technologies, the committee reviewed assessments of the economic competitiveness of such technologies in a technology-neutral policy environment.5 The results of this assessment provide a baseline for evaluating the competitiveness of these technologies and reveal the need for further technological advances.
Developing a levelized cost of electricity (LCOE) is a commonly used method for estimating current and future costs of producing electricity from different generating technologies. While any LCOE must be viewed in light of its assumptions, LCOE estimates can provide a convenient indicator of the relative costs of different technologies and thus a basis for comparing each technology’s ability to compete on the basis of its underlying economic performance. Several sources develop estimates of recent and projected future LCOEs for specific technologies.
EIA has a long history of developing LCOE estimates for different electric power generation technologies (EIA, 2014h).6 These empirical estimates take the location of wind and solar resources into account and incorporate the cost of transmission. EIA also adjusts the LCOE estimates to reflect the relative value
4 A well-accepted method for identifying the readiness of a technology for ultimate dissemination in the marketplace is the TRL taxonomy developed by the National Aeronautics and Space Administration (NASA) as a means of managing its space-related research and development. Further information is provided in Appendix D.
5 That is, comparing the market prices of technologies absent technology-specific policies that, inter alia, lower prices through subsidies or prevent costs from being incorporated into the market price.
6 EIA’s LCOE estimates include capital costs, fuel costs, fixed and variable operations and maintenance (O&M) costs, financing costs, and an assumed utilization rate for each type of generation technology. EIA also provides information on regional variations in the LCOE for different technologies.
of variable wind and solar generation that cannot follow dispatch instructions easily or at all. The EIA estimates show that the value of wind, which blows more at night, when energy prices are low, can be 12 percent below the unweighted average price of electricity; and the value of solar power, with the sun shining when energy prices are higher, can be 16 percent greater than the unweighted average price of electricity (Schmalensee, 2013). EIA also adjusts its LCOE estimates for wind and solar power based on differences in the avoided costs of electricity derived from sources displaced by each. The agency reports its LCOE estimates based on the factors used in modeling for its Annual Energy Outlook publication. In its Annual Energy Outlook 2016, EIA provides detailed information on the estimated LCOEs for different electric power generation capacity additions anticipated to enter service in 2020 (EIA, 2016a).7Appendix B provides a more complete description of the essential elements used for consistent cost estimates in Annual Energy Outlook 2016.
In developing the Annual Energy Outlook, EIA must make assumptions regarding future policies. As a result, its LCOE estimates reflect two important policy assumptions that are not technology-neutral. First, EIA adds 3 percent to the weighted average cost of capital for new coal plants as a proxy for anticipated carbon reduction policies (EIA, 2014h). Second, certain technologies, including nonhydro renewables and combined heat and power, are allowed to use a modified accelerated tax depreciation that is not available to other technologies,8 resulting in substantially lower fixed-charge rates for renewable capital costs (see Appendix B). The committee adjusted EIA’s LCOE estimates to eliminate the impact of these two assumptions and enable comparison of supply options on the basis of technology-neutral policies.
The committee then compared the economic competitiveness of different technologies, first incorporating only an estimate of including the harms from criteria pollutants (see Greenstone and Looney’s (2012) values for impacts of criteria pollutants based on estimated emission rates from Muller and colleagues (2011). Figure 2-6 compares the projected costs of various electric power generation technologies against the cost of an advanced combined-cycle natural gas plant from this perspective where firms do not directly bear the costs of GHG pollution.
7 EIA also provides LCOE estimates for 2040 that include an assumed learning rate for newer technologies. This assumption contributes to renewables being somewhat more competitive in EIA’s forecasts for 2040 relative to those for 2019 (EIA, 2013d).
8 Section 125 of the Tax Increase Prevention Act of 2014 extended the placed-in-service date for the modified accelerated cost-recovery system (Tax Increase Prevention Act of 2014, Public Law 113-295, 113th Congress [December 19, 2014]).
|Technology Category||Technology Readiness Levela|
|Renewable Power Generation|
|1: Electric energy storage|
|2: Hydro and marine hydrokinetic powerb|
|3: Advanced solar photovoltaic powerc|
|4: Advanced concentrating solar power|
|5: Advanced solar thermal heating|
|6: Advanced biomass power|
|7: Engineered/enhanced geothermal systems|
|8: Advanced wind turbine technologies|
|9: Advanced integration of distributed resources at high percent|
|Advanced Fossil Fuel Power Generation|
|10: Carbon capture, transport, and storage|
|11: Advanced natural gas power and combined heat and power (CHP)c|
|12: Water and wastewater treatment|
|Nuclear Power Generation|
|13: Advanced nuclear reactors|
|14: Small modular nuclear reactors|
|15: Long-term operation of existing nuclear plants|
|Technology Category||Technology Readiness Levela|
|Electricity Transmission and Distribution|
|16: Advanced high-voltage direct current (HVDC) technologies|
|17: Reducing electricity use in power systems|
|18: Smart-grid technologies (grid modernization)|
|19: Increased power flow in transmission systems|
|20: Advanced power electronics|
|21: Efficient electrical technologies for buildings and industry|
aTechnology readiness levels are shown on a scale of 1 to 9, where 1 is the least ready. Most of the technology categories shown include technologies with varying readiness levels. A shaded box below a TRL number indicates there is at least one technology at that TRL. See Appendix D for more detail.
bThe committee identified barriers at lower TRLs for hydropower technologies but was unable to make specific level assignments.
cFor concepts beyond three junctions.
These estimates suggest that most increasingly clean power technologies are uncompetitive in the market compared with advanced combined-cycle natural gas power generation unless supported by a technology-specific policy. For example, EIA’s benchmark LCOE for onshore wind generation is 43 percent higher than that for an advanced combined-cycle natural gas unit. Without accounting for GHG externalities, wind and solar energy also are often not competitive with new IGCC coal plants (see 9
Including a cost of $15/ton for GHG pollution still shows a similar picture. While the relative costs for fossil fuel electric power generation technology without carbon capture increase, the costs of increasingly clean technologies such as wind, solar, and carbon capture still remain significantly higher than that of advanced combined-cycle natural gas generation. Figure 2-7 compares the projected costs of various electric power generation technologies against the cost of an advanced combined-cycle natural gas plant where firms account for the costs of pollution from GHGs when installing power generation technologies, using EIA’s assumption that future carbon abatement policies will add roughly $15/ton to the cost of capital for carbon-intensive technologies.10Figure 2-7 shows that wind is still 32 percent more expensive than advanced combined-cycle natural gas generation when the costs of pollution are taken into account. Even if the price charged for the carbon pollution were doubled from $15 to $30/ton of CO2, which would approximate EIA’s estimate of the possible future cost of carbon assuming a 2.5 percent rather than a 3.0 percent discount rate, onshore wind would still remain 23 percent more expensive on average than an advanced combined-cycle natural gas unit.
In addition to EIA, several other groups have begun to produce LCOE estimates for recent historical prices, as well as projections. The National Renewable Energy Laboratory (NREL) has developed its own model that produces a range of high, middle, and low LCOE estimates for essentially the same technologies as those to which the EIA estimates apply. Because of its focus on renewable energy sources, NREL’s model characterizes renewable fuel technologies in greater detail relative to NEMS. Like EIA, NREL expresses the caveat that any method of estimating LCOE is subject to high levels of uncertainty and is dependent on modeling assumptions.11 Also like EIA, NREL generates a range of scenarios when developing its projections. Importantly, current and future cost reduction trajectories are not estimated but are defined as inputs to NREL’s model (Sullivan et al., 2015).
NREL’s available scenarios do not explicitly include a price on pollution but do account for technology-specific tax policies, making it difficult to
10 By comparison, the Interagency Working Group on Social Cost of Carbon (2015) estimates the social cost of carbon pollution to be $36/ton.
11 NREL developed and uses the Regional Energy Deployment System (ReEDS) model. Assumptions in this model are intended to account for, inter alia, transmission infrastructure expansion costs, electric system operation costs, cost of capital, busbar costs at the plant gate, costs of transmission spur lines, site-specific construction costs, and projected changes in capacity factor (Sullivan et al., 2015).
develop a useful comparison of unsubsidized technologies at full cost. As with EIA’s estimates, the committee adjusted the NREL estimates to eliminate the tax policies for all technologies in order to compare LCOE estimates in a more technology-neutral policy environment. Figure 2-8 compares the projected costs of various electric power generation technologies against the cost of conventional combined-cycle natural gas power generation from the market perspective where firms do not directly bear the costs of pollution.
Like EIA’s estimates, these estimates suggest that while the costs of renewable technologies have declined significantly in recent years, absent subsidies or an appropriate price on pollution, increasingly clean technologies often cost more in the marketplace. Based on NREL’s estimates, for example,
solar costs may range from as low as 16 percent less than those of natural gas to as great as 251 percent more, while onshore wind may range from 38 percent less to 56 percent more costly.
Financial advisory and asset management firm Lazard has produced a set of LCOE estimates since 2007 using its own parameters. The most recently available set, from November 2015, provides a range of scenarios, including one that is labeled “unsubsidized” (Lazard, 2015). The assumptions stated, though, are less clear than the assumptions and model parameters specified by EIA and NREL. For example, Lazard notes that its estimates do not include factors that could have a potentially significant effect on its estimates, including “capacity value vs. energy value; stranded costs related to distributed generation or otherwise; network upgrade, transmission or congestion costs; integration costs; and costs of complying with various environmental regulations (e.g., carbon emissions offsets, emissions control systems)” (Lazard, 2015, p. 19). Lazard also states that in its LCOE estimates, it does not attempt to account directly for such externalities as the cost to society of pollution. Instead, a cost of carbon abatement is calculated separately. The analysis makes no mention of the lead time for construction or the year the assets are expected to enter service.
Nonetheless, Lazard’s estimates show ranges similar to those from EIA and NREL. For example, the estimates for utility scale solar PV range from 49 percent less expensive than conventional gas to 56 percent more costly. Onshore wind is estimated to be as much as 62 percent less to as high as 71 percent more costly. Advanced coal with CO2 capture and storage is projected to cost 311 percent more than conventional gas. Again, these ranges suggest that progress has been made in improving the cost-competitiveness of increasingly clean technologies, but continued cost declines are still needed. This is especially true once grid upgrade costs, such as the cost of new transmission assets to accommodate additional wind and solar plants, are incorporated into cost estimates.
Data and news provider Bloomberg New Energy Finance (BNEF) publishes an annual Sustainable Energy in America Factbook that includes retrospective LCOE estimates for a wide range of electric power generation technologies (BNEF, 2016). One advantage of the BNEF estimates is that the analysts attempt to use data from actual constructed power plants whenever possible. A full analysis of BNEF’s estimates, however, was complicated by the opacity of their presentation; estimates are presented as a graphic without underlying figures. BNEF also does not provide details of its model or the assumptions used in any particular scenario except to state that “EIA is source for capex ranges for nuclear and conventional plants” (BNEF, 2016). For example, the estimates are identified as being for “unsubsidized…power generation technologies” (BNEF, 2016, p. 35), but the notes provided do not describe the methodology used to adjust for and eliminate the impact of various subsidies.
Despite these difficulties with its use in the present analysis, the BNEF graphic is illustrative of current costs. For example, in the 2016 factbook presenting estimates for plants entering service in 2015 around the world (BNEF, 2016), it appears that approximately 20 percent of wind plants and fewer than 2 percent of solar PV plants constructed cost less to build than gas- or coal-fired plants. Apparent costs compared with a combined-cycle natural gas plant range tremendously. Onshore wind ranges from 34 percent less to as much as more than 240 percent more expensive. The range for solar PV12 is 26 percent less to nearly 400 percent more expensive. And for offshore wind, the range is 54 percent to approximately 430 percent more expensive. BNEF’s factbook does not provide LCOE estimates for carbon capture-equipped fossil fuel plants.
In addition to estimating a range of costs for various technologies, BNEF estimates “central values” using actual data in its model13 (BNEF, 2016). Looking at these average estimates, it is clear that, except for small hydropower stations, they all either held steady from the first to the second half of 2015 or declined; small hydro costs apparently increased slightly. The data points reveal that the global average costs for onshore wind-generated electric power are greater than the costs of conventional gas- or coal-generated power in the United States and China, and lower than those in Europe and Australia.14 Costs for thin-film and stationary crystalline silicon solar PV appear to be close to competitive with those for coal-fired power generation in Australia but higher than the costs for coal-fired generation for the rest of the world and lower than the costs for gas-fired generation anywhere. Tracking crystalline silicon solar PV is estimated to be more expensive than any fossil fuel generation source. Average costs for large-scale hydropower plants are estimated to be greater than those for natural gas-sourced power in the United States, and greater than those for coal-sourced power in China but lower than those for fossil fuel-fired plants in the rest of the world.15 Offshore wind and concentrating solar power are both estimated to be more expensive than fossil fuel-fired plants anywhere.
Reviewing this evidence and the salient recent literature, it becomes clear that the higher average cost of key increasingly clean electric power generation technologies remains a barrier to their broad deployment (Aldy, 2011).16
12 Includes three technologies: crystal silicon without tracking, crystal silicon with tracking, and thin film.
13 The accompanying notes refer to these as “global central scenarios,” explaining that “these central scenarios are made up of a blend of inputs from competitive projects in mature markets” (BNEF, 2016, Slide 35).
14 BNEF provides estimates for the cost of gas-fired generation only for the United States, China, Europe, and Australia.
15 The factbook does not define the difference between large and small hydropower plants.
16 Uncertainty regarding climate change and the possibility of its very large negative impacts have raised questions about the application of cost-benefit analysis and the social
Finding 2-1: On average, unsubsidized increasingly clean electric power generation technologies are estimated to cost between 43 percent and 391 percent more than a new combined-cycle natural gas facility when prices do not account for the costs of pollution.17
EIA’s Annual Energy Outlook 2016 No Clean Power Plan (CPP) case assumes that the performance of increasingly clean power generation technologies will continue to improve and that governments will continue other policies favoring those technologies.18 Applying these assumptions, EIA projects increases in renewable electric power generation through 2040. Renewables start from a current cost disadvantage and power generation market share of approximately 13 percent, but they (including hydroelectric resources) are projected to provide 23 percent of electric power generation in 2040. That penetration rate for renewables is still far below the projected market shares for natural gas and coal of 32 percent and 29 percent of U.S. electric power generation, respectively (EIA, 2016a).19 The projected increase in renewable electric power generation has only a limited impact on the overall mix of power generation technologies, as reflected in Figure 2-9. In the Annual Energy Outlook 2016 No CPP case, nuclear power’s approximately 20 percent market share is projected to fall to about 15 percent by 2040. As a consequence, with projected growth in demand, EIA’s No CPP case projects that CO2 emissions from the production of electricity could be nearly 4 percent higher in 2040 than they were in 2015.
EIA also conducted one sensitivity analysis that assumes the extension of some policies through 2040 and expansion of other policies meant to decrease CO2 emissions.20 Under those assumptions relative to the No CPP case,
17 Geothermal and hydroelectric power generation costs are exclusive of constraints on capacity increases. The smaller number is associated with wind and the higher number with solar thermal generation.
18 EIA’s No CPP case “assumes that the final CPP rule is permanently voided and is not replaced by other controls on power sector CO2 emissions.” The committee used these projections given the U.S. Supreme Court’s stay of the Clean Power Plan in February 2016 (see Martin and Jones, 2016).
20 Specifically, this sensitivity case assumes that tax policies such as the production tax credit extend beyond their current sunset dates and remain in force, while corporate average fuel economy standards, appliance standards, and building codes are expanded beyond current provisions, and the Clean Power Plan is reinstated with tightening regulation of CO2 emissions starting in 2030.
renewables are projected to grow to supply 34 percent of electric power generation, the same as natural gas, while coal and nuclear both shrink to 16 percent. These results are shown in Figure 2-10. This case projects CO2 emissions from the electricity sector to be roughly 30 percent lower in 2040 than in 2015.
Comparison with an earlier, alternative sensitivity analysis is helpful to consider how extending current policies compares with enacting a policy that would incorporate the cost of pollution into the market price of electricity. In its Annual Energy Outlook 2014, EIA projected that a significant price on carbon starting at $25 per ton of CO2 could increase the market share for nuclear power to more than 37 percent and reduce the electricity sector’s CO2 emissions by nearly 80 percent compared with its 2012 emissions.21 This assumed price on CO2 emissions was projected to increase the average electricity price for 2040
21 In sensitivity cases, EIA examined policies that would favor increasingly clean technologies and made modestly more favorable assumptions regarding the cost of renewable electric power generation. The additional cases included those in which the capital cost of nonhydroelectric renewables was assumed to be 20 percent below reference case levels, and a case in which the carbon price is initially set at $25 per ton of CO2 and increases at a rate of 5 percent per year. In both of these cases, renewable generation was forecast to increase but remain at below a 25 percent market share in 2040 (EIA, 2014a, Appendix B).
by 23 percent compared with the reference case and (in constant 2012 dollars) by 39 percent relative to the average 2012 price. Other studies assuming limited improvements in the cost of low-carbon resources have reached similar conclusions: achieving large reductions in the U.S. electricity sector’s carbon emissions by incorporating the full costs of pollution into electricity prices would lead to significant increases in prices to ultimate consumers.22
These results suggest that major improvements in the cost-competitiveness of low-carbon increasingly clean technologies—improvements that go beyond those assumed in EIA’s or NREL’s analyses—will be required if those technologies are to be market-competitive globally to a degree that encourages significant displacement of incumbent technologies. These improvements will be essential to achieving long-term reductions in GHGs, such as the reduction called for in the COP21 agreement,23 without significantly increasing electricity prices.24
22 Other assessments of the costs of reducing carbon emissions can be found in Clarke et al. (2009, 2014), Fawcett et al. (2009, 2013), CBO (2009), EIA (2009a), Paltsev et al. (2009), Fischer and Newell (2008), and CCSP (2007).
23 Under that agreement, the “United States intends to achieve an economy-wide target of reducing its greenhouse gas emissions by 26%-28% below its 2005 level in 2025 and to make best efforts to reduce its emissions by 28%” (United States, 2015).
Finding 2-2: Achieving long-term targets for reducing GHG emissions from the electricity sector by 80 percent or more without significantly increasing electricity prices would require significant improvements in the performance of low-carbon increasingly clean technologies.
Given the above finding that currently available increasingly clean electric power generation technologies are not yet economically competitive compared with conventional, higher-polluting technologies, the committee considered the extent to which policies designed to expand the deployment of cleaner technologies produce meaningful performance improvements and associated cost declines as a result of “learning by doing” (LBD),25 and the extent to which LBD benefits might offset the difference in the societal costs of low-carbon and conventional resources. The committee found no evidence that increasingly clean technologies could become economically competitive in the near term based primarily on performance improvements achieved through expanded deployment and LBD alone (Gallagher et al., 2012), even though LBD is often assumed to have a material effect on costs. This leads to the conclusion that improving the cost-competitiveness of increasingly clean technologies will require that attention be paid to the larger innovation system.
“Experience curves” are a common component of innovation system models. Simple experience curves have been developed for various technologies and industries in which an historical doubling of technology deployment is associated with coincident reductions in costs or improvements in performance. For newer alternative energy sources, single-factor experience curves typically estimate a 15-20 percent improvement in costs with each doubling of a technology’s adoption (McDonald and Schrattenholzer, 2000). This relationship of improvements in cost or performance to deployment is often presented as a “learning rate.”26 However, experience curves simply document historical associations and by themselves provide limited information on the impacts of increased deployment or appropriate public policy choices.
There are significant limits on the inferences that should be drawn from such experience curves. First, an historical “learning rate” does not necessarily imply that a given technology will continue to improve along its historical trend line. Emerging technologies are quite complex and often include both improving
25 For purposes of this discussion, the phenomenon of customers “learning by using” new technologies is included as a subset of LBD.
and relatively mature components. Improving cost or performance often requires solving “problems” across a wide range of TRLs. At some point, progress in improving components is likely to diminish and be offset by other factors, such as increased input prices. Second, long-term forecasts of technology costs are sensitive to small variations in the choice of the underlying historical data (Nemet, 2006; see also NRC, 2010d; Weisenthal et al., 2012). Third, simple experience curves do not reveal what factors led to observed performance improvements. Proponents of the experience curve method acknowledge that it treats the mechanisms of performance improvement as a “black box” (Junginger et al., 2008). Thus, single-factor experience curves cannot help answer the policy question of the optimal balance between public investment in research and development (R&D) and direct support for increased market adoption, such as expenditures for deployment. Fourth, the associations documented in experience curves do not imply a causal relationship (Clarke et al., 2006; Popp et al., 2010). An increase in the adoption of a technology, for example, may reflect cost reductions that were the result of an independent government research program, innovations developed in a different industry, or other external factors. Thus, the “learning ratios” in experience curves would not, by themselves, demonstrate that programs subsidizing larger-scale deployment of increasingly clean technologies led to material performance improvements.
Experience curves often reflect the impact of multiple factors in addition to LBD (Clarke et al., 2006). This point has important public policy implications. The basic mechanisms by which performance improvements may occur have been widely documented (Junginger et al., 2010). Broadly defined, they include the following:
- Learning by (re)searching (LBS)—This is R&D broadly defined. LBS is an intentional and often costly effort to seek out and develop innovations. Its goal is to develop an innovation until it is at or near the stage of large-scale deployment. Typically, R&D is risky and can have large spillover benefits that are not fully captured by the organization sponsoring the research. These knowledge spillovers justify public support, as private entrepreneurs would otherwise underinvest in R&D activities.
- LBD—This is the creation of new information that reduces the cost of future production. LBD is passive (Thompson, 2010)—it is a free byproduct of deployment rather than an explicit undertaking with its own costs. LBD can produce spillover effects that, depending on the cost of additional deployment and the rate of learning, may justify some amount of public support.
- Economies of scale—Economies of scale reflect decreasing unit production costs as production at a plant or firm reaches an efficient size. In the energy sector, economies of scale may be relevant at the unit, plant, firm, or industry level (Gillingham and Sweeney, 2010).
According to Borenstein (2012, p. 83), “The distinction between learning-by-doing and economies of scale may seem minor, but the implications for public policy are immense. If one firm can drive down its costs by producing at large scale in its factory or its installation operation, those benefits are highly appropriable by that large firm.…Thus, significant economies of scale in any industry, short of creating a natural monopoly, are not generally seen as a basis for government intervention.” Private entrepreneurs can be expected to invest in the realization of economies of scale when doing so will produce an economically competitive product.
- Learning by waiting (LBW)—The spillover effects from other industries, technologies, or countries are essentially exogenous—that is, developed on the outside, from the perspective of the firm (Thompson, 2010). The resulting innovations will appear over time and can be exploited by waiting. In some cases, government support may have played a role in the development of the borrowed technologies for another industry; in other cases, government may be able to accelerate technology transfer and the adaptation of technologies developed in other fields, often with limited intervention. However, LBW is the result primarily of innovation that occurs elsewhere and not of accelerating technology deployment. Separating LBD from the impacts of external technological change is difficult, such that estimates of learning rates based on experience curves can easily be biased upwards (Nordhaus, 2014).
Studies examining the factors contributing to performance improvements that coincide with policy-driven deployments of increasingly clean technologies indicate that single-factor experience curves should be viewed with caution. They may overstate the extent to which significant innovation and performance improvements can be achieved through policies focused primarily on expanding deployment. Söderholm and Sundqvist (2007) developed two-factor models examining the impacts of both R&D and LBD on improvements in wind generation in four European countries from 1986 through 2000. They concluded that the problem of omitting such variables as LBS must be taken seriously. After accounting for R&D, their models estimated that LBD was associated with improvements of about 5 percent with every doubling of deployments (Söderholm and Sundqvist, 2007). Other two-factor studies have reached similar conclusions and indicated that learning rates associated with LBS may be higher than those associated with LBD (Jamasb, 2007; Kahouli-Brahmi, 2008; Kobos et al., 2006). In a widely cited analysis, Nemet (2006) further disaggregated the factors contributing to reductions in U.S. PV energy costs from 1975 to 2001. He considered seven different factors, including economies of scale, efficiency improvements, and reductions in material costs. According to Nemet, “Overall, the ‘learning’ and ‘experience’ aspects of cumulative production do not appear
to have been major factors in enabling firms to reduce the cost of PV” (Nemet, 2006, p. 3226). He goes on to state that “a much broader set of influences than experience alone contributed to the rapid cost reductions” (Nemet, 2006, p. 3230).
Other studies have examined the impacts of deployment subsidies on patent filings and other evidence of innovation. A 2010 study by Swiss researchers, for example, analyzed the effectiveness of “demand-pull” (i.e., deployment) and “technology-push” (i.e., R&D) policies for PV across 15 OECD countries. Their analysis found that “demand pull policies only foster incremental innovation,” and the authors cite “anecdotal evidence that in phases of rapid induced market growth such policies even disincentivize non-incremental innovation.” They conclude that “only technology-push support is able to incentivize non-incremental innovation” (Peters et al., 2011, p. 2).
A recent analysis of the impact of the German Renewable Energy Sources Act, the so-called Erneuerbare-Energien-Gesetz (EEG), on patents for innovation in renewable energy technologies questions whether German feed-in tariffs for PV, wind, and geothermal energy have led to innovation in these technologies. The authors found statistically significant negative correlations between feed-in tariffs for hydroelectric and biomass generation and innovation in these technologies. They conclude that “empirical data of the German feed-in regulation over the last two decades…do not lend support to the proposition that German feed-in tariffs under the EEG spur innovation.” The study found that in the case of PV, which received very high incentives for deployment, “the EEG does not engender innovative output” (Böhringer et al., 2014, p. 15).
An analysis by Nemet (2012) of $1 billion in public investments leading to the deployment of $2 billion in wind generation in California and in contemporaneous performance improvements between 1985 and 2005 found evidence of LBD. However, the LBD benefits were found to diminish with additional deployments (Nemet, 2012). The finding of diminishing benefits from LBD are similar to results from other industries (Arrow, 1962; see also Argote et al., 1990; Benkard, 2000; Darr et al., 1995). They suggest that increasing the scale at which new technologies are deployed cannot be expected to produce a proportionate improvement in performance from LBD because learning rates and their benefits also may moderate as a technology begins to mature. Moreover, Nemet (2012) found that the benefits from LBD also may diminish over time. This may occur because some of the knowledge acquired during deployments may be retained by employees as tacit knowledge and be lost to the firm when they leave, and other lessons learned may become less relevant with changes in technology, demand, or industry structure. Qiu and Anadon (2012) analyzed improvements associated with the Chinese government’s wind power concessions during 2003-2007. Chinese wind prices saw reductions during this time. These reductions, however, can be explained largely by economies of scale and other factors. Taking such factors into account, Qiu and Anadon estimated an LBD rate of only 4 percent for each doubling of production.
These results do not suggest that LBD should be discounted entirely as a mechanism of progress (Arrow, 1962). However, caution is necessary in attributing observed performance improvements to deployment and LBD alone. While LBD appears to have a positive impact, learning rates for LBD may be lower than those for LBS. Importantly, when other factors are taken into account, reasonable estimates for LBD learning rates may be in the single digits. Moreover, evidence of diminishing returns suggests that in some cases, most LBD benefits may be gained through more limited deployment initiatives.
Studies examining patent filings and other evidence of innovation suggest that LBD may play a larger role in incremental improvements in technology, while LBS may be more important for fundamental improvements. Thus, the importance of LBD may depend on the stage of a given technology’s development. And in some cases, given the iterative nature of the innovation process, LBD may complement LBS investments.
27 This appendix contains a quantitative analysis illustrating the analytical model with specified assumptions. While the analysis suggests that LBD is relevant and that a material learning benefit can be associated with deployment, it also suggests that this learning benefit may be too small to offset much of the cost of large-scale deployment of increasingly clean electric power technologies. The analysis indicates that LBS (i.e., investments farther upstream in the innovation system, such as at the R&D stage) is more important in the near term (Nemet and Baker, 2008; NRC, 2010c).
Some increasingly clean electric power technologies will be economically competitive, either generally or in specific applications, independent of any LBD benefits that might result from their deployment. In the near term, it will be advisable to continue to deploy energy-efficient (see Acemoglu et al., 2012). Reducing pollution to socially optimal levels by implementing only a pollution price would likely cost more than doing so in tandem with complementary innovation-focused policies (Parry et al., 2015). Thus while pollution pricing is a critical complement to innovation policies, achieving the level of desired pollution abatement will require tailoring policies to promote innovation in energy technologies, to be comprehensive, to address undue barriers to
innovation in each stage of the innovation process, and to provide significant support for research, development, and deployment (RD&D).
Finding 2-3: Evidence suggests that policies focused disproportionately on subsidizing deployments of increasingly clean technologies will not produce the large, timely, cost-effective improvements in the cost and performance of these technologies required to address pollution problems. Rather, what is required to achieve these improvements in currently available technologies and to create new, as yet unknown breakthrough technologies is a major investment in innovation.
The development of affordable low-carbon increasingly clean electric power generation technologies could position the United States to take more effective measures to address the risks and uncertainties of climate change. In its report Limiting the Magnitude of Future Climate Change, the NRC (2010c) identifies “an urgent need for U.S. action to reduce greenhouse gas emissions.” Recent assessments of risks and uncertainties associated with climate change are consistent with that conclusion (IPCC, 2013, 2014a,b; Walsh et al., 2014). Multiple reviews and expert panels likewise have concluded that global GHG emissions pose clear risks to U.S. economic prosperity (CEIR, 2007; CBO, 2009; Dell et al., 2012, 2014; Interagency Working Group on Social Cost of Carbon, 2013; Nordhaus, 2013) and national security (CNA Military Advisory Board, 2007, 2014; Defense Science Board Task Force, 2011; DoD, 2014).
The committee both agrees with these prior conclusions and acknowledges the uncertainty inherent in making forecasts for complex climate systems. Yet the existence of uncertainty does not mean that the United States should eschew mitigation measures. Avoiding the potential negative consequences of significant climate change is critical to protecting the nation’s economic and security interests. Effective mitigation of climate risks may require a transition to low-carbon energy technologies on a global scale and possibly within a compressed time frame. Significantly reducing the cost and improving the performance of low-carbon energy resources appears both the most efficient and the most likely path to providing options for making an affordable transition to a low-carbon global economy. There is an urgent need for the development of energy technology options that could make the global transition to a low-carbon economy practical, affordable, and timely.
The federal government has taken a number of recent actions to support innovation in electric power generation technologies. For example, the Department of Energy made clear in 2015 that it planned to increase its focus on “crosscutting R&D,” including electric power grid modernization, with a primary goal of continuing to decrease the costs of increasingly clean energy technologies (DOE, 2015c). DOE expects it will need “partnerships with
university scientists and engineers, researchers at both established and entrepreneurial companies, federal and state agencies, and others” to induce the level and kind of transformational innovation needed (DOE, 2015c, p. iii). The committee finds this a positive development, as it is expected that an increase in support for innovation activities such as R&D will be cost-effective in reducing the costs of increasingly clean electric power generation technologies (Baker et al., 2015). The remainder of this report is aimed at providing guidance on how the Department of Energy and other federal entities, along with state governments and other stakeholders, can take action to support and encourage breakthrough innovation to meet the energy challenge.
The committee’s review of currently available increasingly clean electric power generation technologies suggests that they are not yet capable of meeting the challenge of supplying reliable electric power at socially acceptable pollution levels at prices that make them competitive in current electric power markets. Policies therefore need to focus on both the improvement of currently available and the development of new increasingly clean energy technologies. The approach of increasing deployment in the hope that LBD will drive down costs and increase performance appears unlikely to succeed at the scale needed to address the pollution challenge adequately. The gains from LBD are too small to expect that expanded deployment will yield the level of innovation needed. While adequately pricing pollution would also help—both to induce additional innovation and to create a level playing field so that prices reflect the full costs of technologies—it also would likely be insufficient absent other policies.
The implication of these findings for increasingly clean energy innovation policy is that the most important priorities are identifying and creating new options, demonstrating the efficacy of these options, and setting the stage for early adoption of those that are most promising. Although policies could be instituted that would enhance the conditions for eventual large-scale take-up and improvements in use, these policies are likely to be expensive and ineffective without a substantial investment in the earlier stages of the innovation process. The emphasis needs to be on developing technologies that can truly compete with incumbent energy sources. Such technologies are not available today, and efforts to create these future technologies need to be expanded and accelerated. A major investment to this end is warranted, with a clear view of the challenges ahead. These challenges create an opportunity and a need for action by governments at all levels, keeping an eye on the prize of expanding the innovation machine.
Recommendation 2-1: The U.S. federal government and state governments should significantly increase their emphasis on supporting innovation in increasingly clean electric power generation technologies.
Recommendation 2-2: Congress should consider an appropriate price on pollution from power production to level the playing field; create consistent market pull; and expand research, development, and commercialization of increasingly clean energy resources and technologies.