Document 11634926

advertisement
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Review
draft, 5 June 2003. Do not cite without author’s permission.
Toward Cost Buydown Via Learning-By-Doing
For Environmental Energy Technologies
Robert H. Williams
Princeton Environmental Institute
Princeton University
Workshop on Learning-by-Doing in Energy Technologies
Resources for the Future
Washington, DC
17-18 June, 2003
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Review
draft, 5 June 2003. Do not cite without author’s permission.
INTRODUCTION
There is growing interest in use of learning and experience curves for understanding better the
prospects for realizing cost reductions for innovative energy technologies (Cody and Tiedje,
1992: Williams and Terzian, 1993; Goldemberg, 1996; Wene, Voß, and Fried, 2000; Lipman and
Sperling, 2000; Wene, 2000; Williams, 2002; McDonald and Schrattenholzer, 2001; Duke, 2002;
Rubin et al., 2003; Riahi et al., 2003).
Learning curves describe the relationship between cumulative production and labor costs for a
given product manufactured by a specific firm. A 1936 study of airplane manufacturing (Wright,
1936) first introduced formal learning curve analysis showing that manufac-turing experience
facilitates worker skill improvements and that benefits increase in a regular manner with
cumulative production—which serves as a proxy for the stock of worker skill improvements
achieved in this process (Argote and Epple, 1990). Arrow called the process “learning by doing”
(LBD) in a seminal paper introducing the concept to economics (Arrow, 1962).
It has been observed that for many products over long periods of time the cost of producing a
unit of a product is related to its cumulative production by the learning-by-doing equation
y = ax-b, where:
x = cost of producing the next unit of a product,
y = cumulative production of the product,
a = constant coefficient,
b = learning rate exponent.
The learning rate exponent is usually presented in a progress ratio PR = (100*2-b) or learning rate
LR = 100*(1 – 2-b). For LR = 20% (PR = 80%), b = 0.3219.
During the 1970s, the Boston Consulting Group (BCG) introduced experience curves that
generalize the labor productivity learning curve (Boston Consulting Group, 1972). The BCG
presented evidence that its clients benefited from a predictable percentage reduction in overall
costs associated with each doubling of cumulative production. That implies learning-by-doing
not only in the narrow sense of labor productivity improvements but from all conceivable
opportunities for cost reduction, including technological improvement via the fruits of R&D,
input substitution, economies of scale, new product design, and changing input prices. Typically,
experience curves are presented in terms of product prices, which are readily observable, instead
of costs, which are not. Dutton and Thomas (1984) compiles over 100 firm-level experience
curve studies from a variety of manufacturing sectors, finding a mean LR ~ 20% (~ PR = 80%).
When innovations spill over among competing firms, an experience curve can be estimated for
cumulative industrial output that represents the weighted average of the experience curves for the
industry’s individual firms. Cody and Tiedje (1992) points out that highly competitive industries
that make major investments in R&D tend to be characterized by high learning rates, while more
mature industries that invest little in R&D tend to be characterized by low learning rates.
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
Experience curves for some energy technologies are shown in Figure 1. More generally, LRs are
in the range 5-25% with each doubling of cumulative output for energy (McDonald and
Schrattenholzer, 2001).
Use of an experience curve model for which cumulative production is the main explanatory
variable for estimating cost reductions for energy technology that also certainly reflect other
important ongoing processes besides labor productivity improvement is appealing in its
simplicity. But this model lacks a sound theoretical basis. However, Duke (2002) suggests why
both R&D and scale economy impacts (two especially important cost reduction determinants that
complement labor productivity improvement) might roughly be captured by the cumulative
production index. Regarding R&D, he points out that a key part of the new technology
deployment process is R&D carried out by the firm, the cost of which tends to be proportional to
sales revenues (Jenen and Black, 1983); and, at least in the United States, government support
for industrial R&D is typically cost-shared. Thus, it is plausible that during the deployment
period for new technologies, cumulative total R&D investments are roughly proportional to
cumulative production. Duke (2002) also argues that production scale cannot be considered
independently of experience:
“there are…sharp limits on the ability of any firm to reduce costs through dramatic scaleup without first working out the kinks in intermediate scale manufacturing facilities.
Learning-by-doing is therefore an integral part of the process of ramping up production
(Wene, 2000).”
Such considerations suggest the importance of new research aimed at better understanding the
causality of cost reduction associated with energy technological innovation, especially the
robustness of the parsimonious experience model in which cumulative production is the key
explanatory variable for predicting cost reductions.
USING EXPERIENCE CURVES
Despite the tight data fits for many experience curves (see, for example, Figures 1 and 2),
caution should be exercised in using industrial experience curves to make forecasts, especially
long-term forecasts, because so many variables are involved, and there can be no assurance that
historical trends will persist. Figure 1 shows, for example, that as gas turbine technology became
more mature, its learning rate fell from the earlier 20% to about half that value after 1963.
One might try to use experience curves to estimate the cost of “buying down” the price of a
system to a target level where the technology would be widely competitive in many energy
markets. But there are dangers involved in such uses of experience curves. Consider for example,
the buy-down cost (shaded area) in Figure 2 for reducing the retail selling price for PV modules
from the 2003 level of $3.9/Wpeak to $1/Wpeak—a level at which PV technology is likely to be
competitive in many mass energy markets. If the historical average LR (19.7%/y) were to persist,
the required cumulative PV production and the associated buydown cost would be 170 GWpeak
(5% of total global electric generating capacity) and $70 billion, respectively. But if the LR were
2
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
reduced to 15%/y or increased1 to 25%/y the buydown cost would be increased by more than 3fold or reduced by more than half, respectively.2 Another way the buydown effort might be
lightened is if the PV market would come to be dominated by thin-film technologies that require
far less photovoltaically active materials than crystalline silicon systems. At present thin-film
technologies are characterized by about the same production cost as the dominant crystalline
silicon PV modules, but the cumulative production level to date is only about 0.1 as large. If thin
film PV were to advance from this point forward at the same historical LR as for crystalline
silicon PV (19.7%/y), which is a plausible outcome (Payne et al., 2000; Duke, 2002), the
required cumulative production and buydown cost would be 17 GW and $7 billion, respectively.
Nevertheless, experience curves can be helpful in improving understanding of the effort required
to reach targeted or long-term price levels for innovative energy technologies. One approach to
improve confidence that near- to mid-term (10 – 20 year) prices are not underestimated is to
combine experience curve analysis with detailed technology assessments of manufacturing costs
under mass production conditions; the effort required to get to such cost levels can then be
explored as a function of the learning rate—see, for example, Lipman and Sperling (2000) in the
case of PEM fuel cells for cars. Or, one might want to take a longer term perspective and
“guesstimate” floor prices that reflect ultimate material price constraints and minimal
manufacturing costs,3 as Duke (2002) has done for PV technology.
A key issue is the influence of energy technology scale on the utility of experience curves. The
energy LBD survey analysis of McDonald and Schrattenholzer (2001) indicates generally higher
LRs characteristic of more modular systems amenable to mass production in factories4 and lower
LRs characteristic of large field-erected power plants5 for which standardized designs are more
difficult to realize. However, our energy system is dominated by large-scale energy systems at
present, so it is desirable to ascertain if impacts of LBD can be economically important despite
the generally low LRs that seem to characterize large-scale technologies.
Recent experience curve analyses (Rubin, 2003; Riahi, 2003) to assess LBD benefits in
mitigating climate change via CO2 capture and storage (CCS) technologies for fossil fuels
suggest the importance of LBD even if LRs are inherently slow for large-scale technologies.
1
The BCG also gave many examples of industry-wide experience curves in which learning rates increase as the
industry matures.
2
For LR = 15%, the required cumulative PV production and the associated buydown cost would be 760 GWpeak and
$225 billion, respectively. For LR = 25%, the required cumulative PV production and the associated buydown cost
to reach a $1/Wpeak price would be 61 GWpeak and $30 billion, respectively.
3
In general such an exercise is frought with uncertainty. However, for thin-film PV technologies [the technologies
explored in Duke (2002)] the uncertainties are probably much less than for many other future technologies because
the photovoltaically active materials represent a tiny fraction of materials costs, which are dominated by costs for
common materials [e.g., substrate (e.g., glass or thin-rolled steel), wires, etc.] from mature industries. A key
parameter is the module efficiency, which determines the amount of electric power extractable from a m2 of module
in sunlight and can be projected largely on the basis of considerations of the physics of thin films.
4
E.g., a learning rate of 20% (R2 = 0.99) for global sales of PV modules, 1968-1998 [module prices ($/Wpeak) vs
cumulative capacity sold (MWpeak] and 16% (R2 = 0.66) for compact fluorescent lamp sales in the United States,
1992-1998 [$/lumen vs. cumulative units sold].
5
E.g., learning rates of 1.4% (R2 = 0.89), 5.8% R2 = 0.95), and 7.6% (R2 = 0.90), for hydroelectric, nuclear, and coal
power plants, respectively, in OECD countries during 1975-1993 [specific investment cost ($/kWe) vs cumulative
capacity (MWe)].
3
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
Consider that if society should eventually seek to stabilize atmospheric CO2 at some level in the
range 450 – 550 ppmv, many zero- and low-CO2 emitting options will have to be pursued
simultaneously, including fossil fuel decarbonization via CO2 capture and storage (CCS) as well
as renewables. Although renewables (esp. wind and PV) hold great promise for climate change
mitigation in the electricity sector (Williams, 2002), their potential for reducing GHG emissions
at relatively low costs in energy markets that use fuels directly (other than for electricity
generation) is limited. If underground storage of CO2 proves to be a viable option in large-scale
applications, CCS will be an important, low-cost climate mitigation option for markets that use
fuels used directly.6
Although most of the component technologies needed for CCS are well established in the
refining and chemical process industries, few costs for integrated systems are available in the
public domain, because hardly any integrated systems have been built. However, Rubin et al.
(2003) presents a detailed LBD analysis for flue gas desulfurization and selective catalytic
reduction, for which LRs for the specific investment costs ($/kWe) were found to be 11% and
12%, respectively. This analysis was followed up with an extrapolation to CCS technologies,
assuming alternative average LRs of 0%/y and 12%/y for CCS technologies deployed for
electricity generation, H2 production, and carbon-based synthetic fuels production in runs of the
IIASA MESSAGE-MACRO integrated assessment model to explore impacts of constraining the
A2 scenario of the IPCC’s Special Report on Emissions Scenarios (Nakicenovic, 2000) so as to
reduce the atmospheric CO2 concentration in 2100 from 783 ppmv for the A2 scenario (existing
base case) to 550 ppmv (new GHG emissions-constrained variant). To accomplish this, CCS
technologies were introduced and their costs were considered both without and with LBD. The
modeling exercise found that the total amount of CO2 stored by 2100 would be twice as much
with LBD as without, and that the required carbon tax would be reduced from $25/tC to $19/tC
by 2020 and from $82/tC to $27/tC by 2050 (Riahi et al., 2003).
Despite this example suggesting the importance of experience in realizing cost reductions for
large-scale technologies, care should be exercised in extending learning curves blindly to all
large-scale energy technologies. Consider nuclear power, for which experience has often led to
cost increases rather than cost reductions. In the early 1970s Fisher (1974) raised concerns about
rising costs of nuclear power with experience,7 pointing out that for many decades plant
construction in the power industry followed a pattern in which part of the construction (mainly
the building and the boiler) was carried out in the field and part (manufacture of the turbine,
generator, and power conditioning equipment) was carried out in large factories serving many
6
Markets that use fuels directly account for about 2/3 of global CO2 emissions—a fraction that is not likely to
decline much in the future, despite the expected continuing trend toward electrification of the global energy
economy (Williams, 2003).
7
“When measured in constant dollars per kilowatt of capacity, the cost of constructing a nuclear power plant
increased by perhaps 50 percent in the last decade…When power plant costs rises an explanation is required, as we
expect all power plant costs to decline through the economies of scale and new technology. The environmental
movement was responsible for part of the rise in nuclear plant costs, by causing various procedural delays and by
requiring additional safeguards to protect against hypothetical accidents. But there appears to be another cause for
increasing construction costs, associated with a growing portion of high-cost field construction and a shrinking
proportion of low-cost factory construction for the very large power plants now being built…The costs associated
with a shift to field from factory can more than offset anticipated economies of scale….”
4
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
power plants. As electric generating capacity doubled every decade, factory capacity also
doubled, as did field construction at each site. Manufacturing and construction costs per kWe
declined both at the factory and in the field along with these scale increases. As long as both
activities grew in proportion, the economies of scale produced similar percentage cost reductions
in each, and therefore overall cost reductions, even though the unit cost of field construction was
higher than that for factory construction. This pattern held until plant size reached about 200
MWe and resulted in a good fit of the US average electricity price to a 75% experience curve
during most of the period up to 1970 (Williams and Terzian, 1993). This trend was upset,
however, with the introduction of nuclear power. Because of the requirement for shielding and
containment and other features specific to nuclear power, nuclear plants were expected to be
more capital-intensive than fossil fuel plants of the same capacity. Because such costs account
for a smaller fraction of total costs at larger plant sizes, it was reasoned that nuclear plants should
be built larger. Accordingly, nuclear plants were built in the early 1970s to sizes of the order of
1000 MWe. But building such large plants requires shifting a large fraction of total construction
from the factory to less economically-efficient field sites, thereby raising costs. Fisher’s
important insight is that economies of scale in power generation are constrained by the empirical
facts that (i) field construction is inherently more costly than factory construction, and (ii) with
field construction, it is much more difficult to advance along experience curves, in contrast to the
situation with factory construction.
Although no nuclear power plants have been ordered in the United States since 1974 as a result
of both widespread public opposition to nuclear power and skepticism about its economic
prospects in the investment community, the nuclear power industry flourished under central
planning conditions throughout much of the last quarter of the 20th century in France. There
standardized nuclear power plant designs were introduced in an attempt to control costs, but
plant sizes grew to ~ 1300 MWe, despite Fisher’s prescient early warnings about the economic
pitfalls of large plants. Installed nuclear capacity grew rapidly in France, from 1.8 GWe in 1977
to 65.2 GWe in 2000. Yet, despite the favorable conditions for nurturing nuclear power, specific
costs ($/kWe) increased in this period (see Figure 3).8 An important research issue is to ascertain
the causes of this “negative learning”—whether it reflects merely misjudgments about scale
economies or whether nuclear technology issues were also to blame.
BUYDOWN SUBSIDIES FOR ENVIRONMENTAL ENERGY TECHNOLOGIES
Perhaps the most important potential use of energy technology experience curves is as an
analytical tool that can be used in conjunction with government programs that help buy down the
costs of new energy technologies offering promise in dealing with environmental and energy
supply insecurity challenges—as an important complement to government support for research
and development (R&D) for such technologies. Experience curves can be used both to help
shape such technology commercialization programs and in monitoring cost performance in the
field.
8
These data for France stand in sharp contrast to the 5.8% LR for OECD countries, 1975-93, reported in McDonald
and Schrattenholzer (2001). The latter is especially difficult to understand in light of the fact that nuclear costs also
rose in the US during this period, the fact that France and the US together accounted for ~ 60% of OECD nuclear
power generation in that period, and the likelihood that nuclear industrial activities in France and the US affected
nuclear construction activities in other OECD countries during that period.
5
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
Addressing effectively the multiple environmental and energy insecurity challenges posed by
conventional energy while providing affordable energy, especially to meet the economic
aspirations of developing countries, will require a major speeding up the rate of technological
innovation worldwide for energy technologies that offer promise in addressing sustainable
development objectives (PCAST Energy R&D Panel, 1997; PCAST Panel on International
Cooperation in ERD3, 1999; UNDP, UN DESA, and WEC, 2000; Williams, 2001a). Climate
change seems to be the most daunting challenge. Avoiding “dangerous anthropogenic
interference” with the climate system might require stabilizing atmospheric CO2 in the range 450
– 550 ppmv (O’Neill and Oppenheimer, 2002). Achieving such a goal would require reducing
CO2 emissions from the energy system relative to a BAU future (the IPCC’s IS92a scenario) 3065% by mid-century and 70-90% by 2100 (Hoffert et al., 1998). However, external costs
associated with air pollution (especially health impacts of small particle air pollution) and oil
supply insecurity might well have comparable values (see, for example, Rabl and Spadaro, 2000;
Ogden, Williams, and Larson, 2003).
The needed technological innovation requires substantial government support for both energy
R&D and for commercialization (technology cost buydown).
The economic case for government support for R&D is familiar: the private firm will invest in
R&D less than the optimal amount from a societal perspective because the firm cannot
appropriate the full benefits of such long-term investments, which are needed to spur
technological innovation, the single most important contributor to long-term economic growth.
In the case of energy, there is compelling evidence that government R&D support should be
increased substantially (PCAST Energy R&D Panel, 1997; Margolis and Kammen, 2000).
There are some scattered government programs promoting energy technology commercialization
but there is no integrated framework or analytical rationale for such programs, which are often
lampooned as “corporate welfare.” Indeed, a widely held view is that the government’s role in
the energy technological innovation process should end with R&D.
One reason for supporting energy technology commercialization is to compensate for the fact
that there are damages associated with climate change, air pollution, and energy supply
insecurity are not internalized in market prices for energy. But even with full internalization of
such damage costs it would be desirable for government to subsidize commercialization of
promising energy technologies because of a significant positive externality: a private firm that
invests in technology cost buydown via LBD runs the risk that the fruits of its investments will
spill over to its competitors (Duke and Kammen, 1999; Duke, 2002).
When new technologies are introduced into markets, their costs tend to be higher than the costs
of the technologies they would displace. Early investments are needed to “buy down” the costs
of new technologies along their experience curves to levels at which the technologies can be
widely competitive. In principle, a firm introducing a new technology should consider
experience effects when deciding how much to produce and consequently to “forward-price”:
that is, it should initially sell at a loss to gain market share and thereby maximize profit over the
entire production period. In the real world, however, the benefits of a firm’s production
6
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
experience spill over to its competitors, so that the producing firm will forward-price less than
the optimal amount from a societal perspective; that phenomenon provides a powerful rationale
for public-sector support of technology cost buy-downs (Duke and Kammen, 1999; Duke, 2002).
Still another rationale for government support of technology cost buydown activities relates to
technology lock-in and lock-out (Cowan, 2000). The LBD phenomenon can impart a powerful
market advantage to technologies that are launched first in the market (lock-in), making it
difficult for alternative technologies to catch up (lock-out), even if they have superior attributes
and better long-term economic prospects. Buy-down support for a portfolio of technologies
would help minimize the risks of technology lock-in/lock-out and in the process enhance the
prospects for good economic returns on earlier government R&D investments in the alternatives.
Duke (2002) develops a methodology for estimating the NPV of the net benefits of government
investments in technology cost buydowns and has shown that in principle, in order to maximize
social welfare, government support for buydowns should persist for qualifying technologies not
just to the point of breakeven with competing technologies but all the way to the (ultimate) price
floor for such technologies. The high costs of commercialization activities compared to R&D
and the general scarcity of public-sector resources dictate that public-sector investments in such
innovative activities be prioritized. Duke (2002) also formulates a set of criteria for technology
choice aimed at maximizing NPV of such investments [see Duke (2002) for details]:
•
•
•
•
•
There is high LBD spillover;
The technology has begun to sell in commercial niche markets and emerging data suggest
a reliably steep experience curve while technology assessment projects a low price floor;
Niche markets have proven insufficient to pull prices down rapidly, but at lower prices
demand is price elastic such that buydown subsidies can open up large markets;
Based on the preceding criteria, no other technology (including both incumbent and
emerging substitutes) has better long-term prospects in the relevant market segments;
The targeted technology produces strong public benefits not related to private underinvestment in LBD—e.g., major environmental benefits.
If government is to subsidize the energy technology cost buydown process, it would necessarily
have to try to “pick winners”—a prospect that makes many policy analysts uncomfortable. But
because of the paucity of experience with truly radical clean technologies, public-sector
decision-makers are unlikely to be less prescient than those in the private sector (Baumol, 1995).
An appropriate strategy for dealing with the winner-picking concern is to support a portfolio of
technologies that satisfy the formal selection criteria and offer great promise in addressing
societal goals relating to environment and energy supply security—thus reducing the overall
buy-down program performance risk through diversification.
The activity would have to be carried out in a manner that minimizes the potential for evolving
political pork barrel projects such as present grain ethanol subsidies. Duke (2002) suggests the
establishment of an Energy Commercialization Policy Agency (ECPA) where teams of
technologists, economists, and scientists would:
• provide energy technology assessment advice to policymakers regarding candidate
technologies for buydown support;
7
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
•
•
monitor and evaluate field performance of technologies being subsidized; and
provide advice on periodic reallocation of subsidy resources in light of such
monitoring/evaluations.
An EPCA should be independent of the DOE or any other agency that is promoting advanced
technologies that might become candidates for government buydown support. And the analysis
carried out by an EPCA should be subject to rigorous external review by academics and
managers with generic expertise in technology commercialization but no particular attachment to
the technologies in question.
Establishment of a criteria-based energy technology cost buydown activity would represent a
major change in US energy technology innovation policy, which has focused on R&D (however
inadequately), and there are many uncertainties as to how to proceed…but it is difficult to
imagine how the major 21st century challenges posed by conventional energy can be addressed
effectively without extending R&D policy to address the formidable technology
commercialization issues as well.
Although the basic thrust of such a policy would represent a major departure from the status quo,
implementation could be advanced with relatively minor modifications of promising policy
instruments that have already been tested.
POLICY INSTRUMENTS FOR BUYDOWNS
Although there has been no strong analytical framework for government policy relating to
subsidies for technology commercialization, much experience has been accumulated as to the
efficacy and efficiency of alternative instruments that can be used to subsidize energy technology
commercialization. To illustrate the utility of this experience, a brief discussion of how a
Renewable Portfolio Standard (RPS) or Green Certificate Market (GCM) as it is known in
Europe, might be modified and used in the power sector as a buydown instrument…only one of
many possible alternative instruments that might be adapted to technology cost buydown
support.
The RPS or GCM is emerging as an instrument of choice in many parts of the world for
encouraging the commercialization of new renewable energy technologies (see Box A). One of
the main attractions of this instrument is its use of market forces to allocate subsidy resources for
qualifying technologies. Another is that by specifying the quantity of supply to be subsidized
over time, renewable-energy developers are given the confidence to launch projects—assuring
them stable revenue streams and facilitating access to low-cost financing. Specifying the quantity
of the technology to be subsidized would also be helpful in a climate-mitigation policy framed
around an atmospheric CO2 stabilization target (e.g., 500 ppmv).
A concern about subsidy programs that fix the quantity of technology to be subsidized is that a
priori the cost is very uncertain—a concern that has led many to favor subsidy programs that
specify instead the magnitude of the subsidy…but with this approach the quantity of the
technology subsidized is very uncertain. In the embryonic Texas RPS, which is regarded as one
of the best formulated RPS programs (Wiser, 2001), the two approaches are combined in that a 5
cents/kWh noncompliance penalty caps the cost of the program.
8
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
Box A: Renewable Portfolio Standard
An RPS or GCM requires that each electricity provider include in its supply mix a small but
growing fraction of renewables. Companies can either produce renewable electricity or purchase
renewable energy credits (RECs or, in Europe, "green certificates") in a credit-trading market.
Several industrialized countries have introduced or are developing national RPS/GCM
initiatives. China intends to explore the implementation of an RPS under its 10th Five-Year Plan
(Jacard, Chen, and Li, 2001). Ten U.S. states have implemented an RPS, often in conjunction
with electricity-market-restructuring initiatives, and, in recent years, a federal RPS has been
considered in various Congressional bills.
If it is to be used as a technology cost buydown instrument it would be desirable to redesign an
RPS to support a portfolio of qualifying technologies—perhaps by establishing separate tranches
for the different qualifying technologies. Those who are apprehensive about entrusting the
government to “pick winners” insist that an RPS should be technologically blind except for the
“renewables” requirement and that the market should determine which technologies win the
largess of the subsidy. But there are serious problems with this approach as a buydown
instrument. At the most fundamental level, different technologies are at different stages on their
development paths at any one time. Less mature technologies can often make significant nearterm contributions in niche markets [rooftop PV systems for new houses will soon be
competitive at the retail level (Payne, Duke, and Williams, 2000), for example]. A
technologically blind policy would tend to neglect such opportunities and give all or most of the
available subsidy to the technology (wind, for example) that has advanced furthest along its
experience curve. In effect, a “technology blind” policy might end up picking a single winner.
Also, some technologies that might be competitive under a technology-blind policy have limited
overall potential (biogas and landfill gas come to mind), so that the leveraging effect of the
subsidy would be modest.
An RPS might eventually be extended to cover technological options other than renewables that
satisfy the selection criteria—e.g., an extension to fossil fuel power generating systems with CO2
capture and storage, after underground CO2 storage has been sufficiently demonstrated to be
viable as a major clean energy option.
TOWARD EARLY ACTION FOR BUYDOWNS
Although radical technological change for energy technology is required to address effectively
the major environmental and energy supply insecurity posed by conventional energy,
laboratory advances are not needed before moving ahead with major new efforts. Much can be
accomplished with commercially ready new technologies that are too costly to make major
inroads in the energy economy today but are strong candidates for technology cost buydown
support—e.g., in the case of renewables: intermittent wind power, baseload wind power (e.g.,
via wind/compressed air energy storage hybrids), and PV for residential rooftop applications
(Cavallo, 1995; Payne, Duke, and Williams, 2001; Williams, 2002); and, in the case of fossil
fuels: integrated gasification combined cycle technology, which is a key stepping stone to lowcost CO2 capture and storage technology for coal (Williams, 2003; Foster Wheeler, 2003).
Also, experience to date has highlighted promising policy instruments that might be used for
9
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
buydown, although experimentation with alternative approaches is desirable to find out what
works best with different classes of technologies and perhaps in different regions as well.
Such buydown activity should go hand in hand with a reinvigoration of the energy R&D
enterprise, and the two activities would be mutually reinforcing: increased R&D would facilitate
sustaining high learning rates, and market growth as a result of buydown success would catalyze
increased R&D investments by the firms involved.
Early buydown action offers the potential for increasing social welfare as a result of speeding up
the approach to the ultimate floor prices for the technologies supported (Duke, 2002), while
generating substantial and energy security and environmental benefits—not the least of which
would be a significant increase in the chances of realizing the goal set forth in the Article 2 of the
UN Framework Convention on Climate Change:
“to…achieve stabilization of greenhouse gas concentrations in the atmosphere at a level
that would prevent dangerous anthropogenic interference with the climate system…”
REFERENCES
Argote, L. and D. Epple (1990). “Learning curves in manufacturing.” Science 247: 920-924.
Arrow, K., 1962: The economic implications of learning by doing, Review of Economic Studies,
29: 166-170.
Baumol, W., 1995: Environmental industries with substantial start-up costs as contributors to
trade competitiveness, Annual Review of Energy and Environment, 20: 71–81, 1995.
Boston Consulting Group, 1972: Perspectives on Experience, Boston.
Cody, G. D. and T. Tiedje, 1992: The potential for utility scale photovoltaic technology in the
developed world: 1990-2010, Energy and the Environment, B. Abeles, A. J. Jacobson and P.
Sheng. Singapore, World Scientific.
Cohen, L. R. and R. G. Noll, 1991: The Technology Pork Barrel, The Brookings Institution,
Washington, DC.
Cowan, R., 1999: Learning curves and technology policy: on technology competitions, lock-in
and entrenchment, in Experience Curves for Policy Making: the case of energy technologies, C.O. Wene, A. VoB and T. Fired, eds., Proceedings of the IEA International Workshop at Stuttgart,
Germany, 10-11 May, 1999, International Energy Agency.
David, J., and H. Herzog, 2000: The cost of carbon capture. In Proceedings of the Fifth
International Conference on Greenhouse Gas Control Technologies, D.J. Williams, R.A. Durie,
P. McMullan, C.A.J. Paulson, and A.Y. Smith, eds., CSIRO Publishing, Collingwood, Victoria,
Australia, pp. 985-990.
10
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
Dooley, J.J., and P.J. Runci, 1999: Adopting A Long View to Energy R&D and Global Climate
Change, PNNL-12115, Battelle Northwest National Laboratory Report Prepared for the Panel on
International Cooperation in Energy Research, Development, Demonstration, and Deployment,
of the President’s Committee of Advisors on Science and Technology, Office of Science and
Technology Policy, The White House, Washington, DC, February.
Foster Wheeler Energy Ltd., 2003: Potential for Improvement in Gasification Combined Cycle
Power Generation with CO2 Capture, IEA Greenhouse Gas R&D Programme, UK.
Duke, R. D. and D. M. Kammen, 1999: The Economics of energy mrket transformation
initiatives, The Energy Journal, 4: 15-64.
Duke, R.D., 2002: Clean Energy Technology Buydowns: Economic Theory, Analytic Tools, and
the Photovoltaics Case, PhD. dissertation, Woodrow Wilson School of Public and International
Affairs, Princeton University, November.
Dutton, J. M. and A. Thomas, 1984: Treating progress functions as a managerial opportunity,
Academy of Management Review, 9(2): 235-247.
Fisher, J.C., 1994: Energy Crises in Perspective, Wiley, New York.
Girard, P. (Commissariat à l’Energie Atomique), Marignac, Y. (WISE-Paris), and Tassart, J.
(La Commission Française du Développment Durable), Le Parc Nucléaire Actuel, Mission
d’évaluation économique de la filière nucléaire, March.
Foster Wheeler, 2003: Potential for improvement in gasification combined cycle power
generation with CO2 capture, IEA Greenhouse Gas R&D Programme.
Agency.
Goldemberg, J., 1996: The evolution of ethanol costs in Brazil, Energy Policy, 24(12): 1127
-1128.
Hoffert, M.I., K. Caldeira, A.K. Jain, E.F. Haites, L.D.D. Harvey, S.D. Potter, M.E.
Schlessinger, S.H. Schneider, R.G. Watts, T.M.L. Wigley, and D.J. Wuebbles, 1998: Energy
implications of future stabilization of atmospheric CO2 content, Nature, 395, 881-884.
Jacard, M., H. Chen, and J. Li, 2001: Renewable portfolio standard: A tool for environmental
policy in the Chinese electricity sector, Energy for Sustainable Development, V (4): 111–119.
Jelen, F. C. and J. H. Black, 1983: Cost and Optimization Engineering. New York, Mc-GrawHill Book Company.
Nakicenovic, N. (Coordinating Lead Author) et al., 2000: Special Report on Emissions
Scenarios, prepared by Working Group III of the Intergovernmental Panel on Climate Change,
Cambridge University Press, Cambridge, UK.
11
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
Lipman,T.E., and D. Sperling, 2000: Forecasting the costs of automotive PEM fuel cell
systems—using bounded manufacturing progress functions, in Experience Curves for Policy
Making: the Case of Energy Technologies, Proceedings of the IEA International Workshop
at Stuttgart, Germany, 10-11 May, 1999, Wene, C.-O., A. Voß, and T. Fried, eds., International
Energy Agency, April, ISSN 0938-1228.
Margolis, R., and D.M. Kammen, 1999: Evidence of under-investment in energy R&D in the
United States and the impact of federal policy, Science, 285: 690-692, 30 July.
Mattson, N. and C.-O. Wene, 1997: Assessing new energy technologies using an energy system
model with endogenized experience curves, International Journal of Energy Research, 21:
385-393.
McDonald, A. and L. Schrattenholzer, 2001: Learning rates for energy technologies, Energy
Policy, 29: 255-261.
Nakicenovic, N., A. Grübler, and A. MacDonald, 1998: Global Energy Perspectives, Cambridge
University Press, Cambridge, U.K.
Ogden, J.M., R.H. Williams, and E.D. Larson, 2003: Societal lifecycle costs of cars with
alternative fuels/engines, Energy Policy (in press).
O’Neill, B.C., and M. Oppenheimer, 2002: Dangerous climate impacts and the Kyoto Protocol.
Science, 296, 1971-1972.
Payne, A., R. Duke, and R.H. Williams, 2001: “Accelerating Residential PV Expansion: Supply
Analysis for Competitive Electricity Markets,” Energy Policy, 29: 787-800.
PCAST Energy R&D Panel, 1997: Federal Energy Research & Development for the Challenges
of the 21st Century Report of the Energy R&D Panel, The President’s Committee of Advisors
on Science and Technology, November 1997. Available on the World-Wide Web at
http://www.whitehouse.gov/WH/EOP/OSTP/html/ISTP_Home.html.
PCAST Panel on International Cooperation in ERD3, 1999: Powerful Partnerships: the Federal
Energy Research & Development for the Challenges of the 21st Century Report of the Panel on
International Cooperation in Energy Research, Development, Demonstration, and Deployment of
the President’s Committee of Advisors on Science and Technology, June 1999. Available on the
World-Wide Web at http://www.whitehouse.gov/WH/EOP/OSTP/html/ISTP_Home.html.
Pope, C., et al., 1995: Particulate air pollution as a predictor of mortality in a prospective study
of U.S. adults, American Journal of Respiratory and Critical Care Medicine, 151(3): 669-674.
Rabl, A. and J. V. Spadaro, 2000: Public health impact of air pollution and implications for the
energy system, Annual Review of Energy and the Environment, 25: 601-627.
12
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
Riahi, K., E.S. Rubin, M. Taylor, L. Schtrattenholzer, and D. Hounshell, 2003: Technological
learning for carbon capture and sequestration technologies, Energy Economics (in press).
Wiser, R., and O. Langniss, 2001: The Renewables Portfolio Standard in Texas: An Early
Assessment, a Lawrence Berkeley Laboratory report prepared for the Office of Power
Technologies of the U.S. Department of Energy, LBNL-49107, November 2001.
Rubin, S.S., M.R. Taylor, S. Yeh, and S.A. Hounshell, 2003: Learning curves for environmental
technologies and their importance for climate policy analysis, Energy, the International Journal
(in press).
UNDP (United Nations Development Programme), UN DESA (United Nations Department of
Economic and Social Affairs), and WEC (World Energy Council), 2000: World Energy
Assessment: Energy the Challenge of Sustainability, (a study sponsored jointly by the United
Nations Development Programme, the United Nations Department of Social and Economic
Affairs, and the World Energy Council), published by the Bureau for Development Policy,
United Nations Development Programme, New York, 508 pp.
Wene, C.-O., 2000: Experience Curves for Technology Policy, International Energy Agency,
OECD, Paris.
Wene, C.-O., A. Voß, and T. Fried, eds., 2000: Experience Curves for Policy Making: the Case
of Energy Technologies, eds., Proceedings of the IEA International Workshop at Stuttgart,
Germany, 10-11 May, 1999, International Energy Agency, April, ISSN 0938-1228.
Williams, R. and G. Terzian, 1993: A benefit/cost analysis of accelerated development of
photovoltaic technology, CEES/PU Report No. 281, Center for Energy and Environmental
Studies, Princeton University, October.
Williams, R.H., 2001a: Addressing challenges to sustainable development with innovative
energy technologies in a competitive electric industry, Energy for Sustainable Development, V
(2): 48-73, June.
Willams, R.H., 2001b: Toward zero emissions from coal in China, Energy for Sustainable
Development, V(4): 39-65, December.
Williams, R.H., 2002: Facilitating widespread deployment of wind and photovoltaic
technologies, pp. 19-30, 2001 Annual Report of the Energy Foundation (available at
www.ef.org).
Williams, R.H., 2003: Decarbonized fossil energy carriers and their energy technological
competitors, pp. 119-135, in Proceedings of the Workshop on Carbon Capture and Storage of
the Intergovernmental Panel on Climate Change, Regina, Saskatchewan, Canada, 18-21
November 2002.
Wright, P., 1936: Factors affecting the cost of airplanes, Journal of Aeronautical Science, 1: 122128, February.
13
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
20000
1983
RD&D phase
1981
Photovoltaics
(learning rate ~ 20%)
Commercialization
phase
10000
US(1990)$/kW
5000
2000
USA
Japan
1992
1995
Windmills (USA)
(learning rate ~ 20%)
1982
1000
1987
500
1963
Gas turbines (USA)
1980
(learning rate ~ 20%, ~10%)
200
100
10
100
1000
10000
100000
Cumulative MW installed
Figure 1: Experience curves for photovoltaics, wind generators, and gas turbines
These curves illustrate the well-established phenomenon that, for new technological products
amenable to the economies of mass production, prices tend to decline with cumulative
production.
Source: Nakicenovic, Grübler, and MacDonald (1998).
14
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
100
All-PV Retail Price 1976-2000
Learning Rate = 20%
2003: 2.3 GWp, $3.90/Wp
$70 billion buydown
2000$/W p
10
Thin-film only
Learning rate = 20%
0.23 GWp & $3.90/Wp in 2003
$7 billion buydown
1
0.1
00.1
1
10
100
1,000
10,000
100,000
1,000,000
cumulative MWp
Figure 2: Costs of buying down crystalline silicon and a-Si:H modules to $1.0/Wp, from
equal price levels in 2003
This figure shows projections of retail PV module prices based on both the historical PV
experience curve for all PV technologies (dominated by experience with crystalline PV
technologies) and a postulated a-Si:H curve assumed to have the same progress ratio and initial
module price in 2003. Buy-down is defined here as the total incremental expenditure required to
reduce module prices from their initial values to the $1.0/Wp target price, without taking credit
for niche market opportunities that would reduce the buy-down cost considerably. The module
prices in the historical PV experience curve are the wholesale prices augmented by a 20 percent
retail mark-up.
Source: Williams (2002), based on analysis in Payne, Duke, and Williams (2001). Other PV
experience curve analyses can be found in Cody and Tiedje (1992), Williams and Terzian
(1993), and Duke (2002).
15
R.H. Williams, “Toward cost buydown via learning-by-doing for environmental energy technologies.” Do not cite
without author’s permission.
S pe cific inv estm ent c ost (19 98 FF /k W)
11000
65.7 GW , 2000
10000
9000
25.5 GW , 1983
1.8 GW , 1977
43.7 G W , 1986
8000
7000
1
10
100
Installed nuclear capacity (G W)
Figure 3: Specific investment cost for nuclear power in France vs cumulative installed
nuclear capacity
Calculations (preliminary) carried out by Arnulf Grübler (International Institute for Applied
Systems Analysis, Laxenburg, Austria) based on data presented in Girard, Marignac, and Tassart
(2000).
Cost data for nuclear power plants are total investment costs at project completion (including
interest during construction. Individual project completion data have been smoothed with 3-year
running averages.
Note that the specific investment cost in 2000 = FF1998 10,570/kWe = $1998 1790/kWe (@
1998 average exchange rate) = $2002 1920/kWe (using US GDP deflator).
Source: Private communication from Arnulf Grübler, June 2003.
16
Download