Professional Documents
Culture Documents
The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/authorsrights
Author's personal copy
112 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 The Electricity Journal
Author's personal copy
But how do the risks of Independent of their type, we completed utility-scale projects
construction cost overruns conclude that construction costs greater than 1 MW in size. This
compare across different forms of for electricity projects are difficult excluded forms of distributed
electricity infrastructure? How do to predict, as only 39 projects generation such as fuel cells and
utility-scale renewable sources of across the entire sample were Stirling engines, smaller-scale
electricity such as wind farms and completed at or under budget. We generation sources such as diesel
solar facilities perform compared also find that hydroelectric dams, generator sets and microhydro
to thermal plants, nuclear nuclear power plants, wind dams, and projects canceled or
reactors, and hydroelectric dams? farms, and solar facilities each still under construction in late
Where do high-voltage have their own unique set of 2013 and early 2014. As locating
transmission networks fall on this construction risks, which we reliable data for both the original
continuum of overrun risk? elaborate on below. cost estimate and the actual cost of
May 2014, Vol. 27, Issue 4 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 113
Author's personal copy
Its actual construction cost; precise phrase, we also inferred A. Hydroelectric dams
and them if there was a difference
If available, its estimated between the quoted cost at the Sixty-one hydroelectric dams
construction time and actual start of project and quoted cost at made it into our sample
construction time (which we were the completion of project. representing 113,774 MW of
able to find for a smaller installed capacity worth $271.5
subsample of 327 projects). billion of investment. These
We updated all costs and III. Results projects experienced a total of
currencies to US$2012 using $148.6 billion in cost overruns
historical currency conversions Across our entire sample of (38.3 percent of the total overrun
available at Oanda.com and more than 400 projects, the cost across the entire sample) and
adjustments for inflation from the average installed capacity was exhibited a mean cost escalation
Statistical Abstracts of the United 927 MW for a power plant, and of 70.6 percent. Cost overruns also
States. Our final dataset the average length was 193 km afflicted more than three out of
encompassed 401 projects for a transmission line. The mean every four (75.4 percent) of the
(Appendix II). installed cost per kilowatt (kW) hydroelectric projects in the
axis)
means we investigated so-called 2000 80
Months
publicize, we not only detected Figure 2: Mean Construction Times and Cost Overruns for Electricity Infrastructure by
them when documents used that Reference Class. Note: Construction time data are from a smaller sample of projects
114 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 The Electricity Journal
Author's personal copy
World Commission on Dams has cost escalation was 56 percent and the monotonic relationship
also posited that large dams are that about three-quarters of between time overruns and cost
unusually prone to unforeseen projects were afflicted by cost overruns. One explanation for
excavation and construction overruns (World Commission on this mismatch between project
problems given that geotechnical Dams, 2000). In parallel, the delays and cost overrun trends
conditions at the site, such as the World Bank has noted could be that costly attempts were
quality of foundational rocks and meaningful cost overrun trends in made to accelerate schedules so as
composition of construction some of their assessments of large to minimize delays. This could
materials, cannot be precisely hydropower projects (Merrow have resulted in higher wages and
evaluated until after construction and Shangraw, 1990; Bacon et al., overtime costs, used to attract
begins (World Commission on 1996). workers, leading to a decrease in
Dams, 2000). Building dams thus lead times but an increase in
involves a great deal of B. Nuclear power plants expenses (U.S. Energy
(expensive) trial and error. Information Administration,
May 2014, Vol. 27, Issue 4 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 115
Author's personal copy
Hoffman, 1996). Similarly in used throughout industry and not University of California at
France, Charles de Gaulle necessarily only the energy Berkeley, and the Lawrence
promoted nuclear power plants industry): those at ‘‘big plants,’’ Berkeley National Laboratory
as a mechanism to reconstruct those with medium-sized assessed a three decade historical
French national identity. Nuclear ‘‘modules,’’ and those with database of delivered costs from
technology was seen by French smaller-scale systems that were each of 99 conventional nuclear
policymakers as a way to ‘‘continuously’’ operated. They reactors operating in the United
simultaneously rebuild French found that the ‘‘highest learning States (at that time) (Hultman
infrastructure and reestablish its effects’’ were observed for the et al., 2007). Their assessment
role as a world leader (Hecht, smallest facilities and that ‘‘big found a significant group of
1998). In both the French and plants and modules display less plants with extremely high
American cases, government dynamic learning effects’’ historical costs: 16 percent in the
created a market for nuclear (Christiansson, 1995). Indeed, in more than 8¢/kWh category. The
power, rather than the other way authors pointed out two unique
around—meaning cost attributes of reactors that made
considerations came second to Nuclear technology was them prone to unexpected
political ones. seen by French increases in cost: (1) their
116 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 The Electricity Journal
Author's personal copy
[(Figure_4)TD$IG]
Mean size, Right Axis
9000
Mean cost, Le Axis
2000 cost of scaling is justified by
8000 Mean cost overrun, Le Axis 1800 greater energy capture, resulting
7000 1600 in a lower levelized cost of wind
1400 generated electricity. Solar
6000
1200
5000 facilities, by contrast, have
1000
4000 benefited from lower priced
800
3000 inputs such as silicon and faster
600
2000 rates of technological learning
400
1000
(Nemet, 2006, 2009).
200
0 0
Hydro Nuclear Thermal Wind Solar
Figure 4: Mean Size (MW), Cost Overrun (Millions of USD), and Cost (per installed kW) for IV. Discussion
Electricity Infrastructure by Reference Class
What, then, accounts for the
prevalence of cost overruns across
million. Thirty-nine large solar PV escalation. However, the price per these distinct classes of electricity
or CSP power plants, installed kW jumps—suggesting projects? While not an exhaustive
representing 2,374 MW of that one must ‘‘pay’’ for the explanation, this section proposes
installed capacity worth $16.5 mitigation of construction risk at least three fundamental
billion worth of investment, that these smaller projects offer. reasons.
actually came in as a class $4.2
million under budget, or $200,000
less than expected per project,
As to the subtle differences
between wind and solar—both
are considered forms of
F irst, construction cost
overruns are multi-causal;
they cannot be reduced to a single
with a mean cost escalation distributed generation and are factor, which means even having
(overall, when it did occur) of 1.3 built in much smaller increments one thing ‘‘go wrong’’—such as a
percent. These solar systems had than thermal, nuclear, and hydro delay, missing component, or
the highest rate (28.2 percent) of facilities—perhaps wind shortage of labor—can impact
cost underruns, when excluding experienced a greater degree and construction schedules. The U.S.
transmission projects, as well. extent of cost overruns due to Energy Information
May 2014, Vol. 27, Issue 4 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 117
Author's personal copy
118 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 The Electricity Journal
Author's personal copy
Table 1: Mean Cost Escalation for Various Infrastructure Projects. lisation of nuclear power. In: Byrne,
J., Hoffman, S.M. (Eds.), Governing
Technology Mean Cost Escalation (%) (n) for the Sample the Atom: The Politics of Risk.
Nuclear reactors 117 180 Transaction Publishers, London,
pp. 11–46.
Hydroelectric dams 71 61 Christiansson, L., December 1995. Dif-
Railway networks 45 58 fusion and Learning Curves of
Bridges and tunnels 34 33 Renewable Energy Technologies.
IIASA, Luxemburg WP-95-126.
Roads 20 167
Cohen, B.L., 1990. Costs of Nuclear
Mining projects 14 63 Power Plants: What Went Wrong?
Thermal power plants 13 36 Nuclear Energy Option. Plenum
Wind farms 8 35 Press, New York.
De Bondt, W.F.M., Makhija, A.K., 1988.
Transmission projects 8 50 Throwing good money after bad?
Solar farms 1 39 Nuclear power plant investment
Source: Data for electricity infrastructure comes from this study. Data for other items come from (Flyvbjerg et al., 2002, decisions and the relevance of sunk
2004). costs. J. Econ. Behav. Org. 10, 173–
199.
Findlay, T., 2010. The Future of Nuclear
Energy to 2030 and Its Implications
Appendices I and
I n sum, electricity for Safety, Security, and Nonproli-
feration. Center for International
infrastructure seems prone to II. Supplementary data Governance Innovation, Waterloo,
cost overrun issues independent, Ontario.
almost, of technology or location. Supplementary material related Flyvbjerg, B., 2005. Design by decep-
Each of these different types of to this article can be found, in the tion: the politics of megaproject
approval. Harvard Des. Mag.
electricity infrastructure poses online version, at http://dx.doi.
(Spring/Summer) 50–59.
different construction risks. The org/10.1016/j.tej.2014.03.015. Flyvbjerg, B., 2006. From nobel prize to
findings of this study do not bode project management: getting risks
References right. Project Manage. J. (August)
well for climate change mitigation 5–15.
efforts, given that two of the largest Ansar, A., Flyvbjerg, B., Budzier, A., Flyvbjerg, B., Holm, M.S., Buhl, S., 2002.
‘‘wedges’’ (Pacala and Socolow, Lunn, D., 2014. Should we build Underestimating costs in public
more large dams? The actual costs works projects: error or lie? J. Am.
2004) that we have to mitigate of mega-dam development. Energy Plan. Assoc. 68 (Summer (3)) 279–
emissions—hydroelectric dams Policy (forthcoming). 295.
and nuclear reactors—have the Bacon, R.W., Besant-Jones, J.E., 1998. Flyvbjerg, B., Bruzelius, N., Rothengat-
Estimating construction costs and ter, W., 2003. Megaprojects and Risk:
greatest amount and frequency of schedules. Energy Policy 26 (4) An Anatomy of Ambition. Cam-
cost overruns. Yet as systems get 317–333. bridge University Press, New York.
smaller in overall capacity Bacon, R.W., Besant-Jones, J.E., Heidar- Flyvbjerg, B., Holm, M.S., Buhl, S., 2004.
ian, J., August 1996. Estimating Con- What causes cost overrun in trans-
(moving to wind and solar) we see
struction Costs and Schedules port infrastructure projects? Trans-
generally higher per installed kW Experience with Power Generation port Rev. 24 (January (1)) 3–18.
costs but also less risk to overruns Projects in Developing Countries. Grubler, A., 2010. The costs of the
and lower rates of overruns that World Bank, Washington, DC. French nuclear scale-up: a case of
Bolinger, M., Wiser, R., 2009. Wind negative learning by doing. Energy
are less costly when they occur. power price trends in the United Policy 38, 5174–5188.
This may imply that utilities and States: struggling to remain compe- Grubler, A., October 2009. An Assess-
project sponsors face the dilemma titive in the face of strong growth. ment of the Costs of the French
Energy Policy 37 (March (3)) 1061– Nuclear PWR Program, 1970–2000.
of having to ‘‘pay’’ for risk 1071. International Institute for Applied
protection: you pay higher costs Bolinger, M., Wiser, R., 2012. Under- Systems Analysis, Vienna IR-09-376.
per kW to achieve more standing wind turbine price trends Hecht, G., 1998. The Radiance of France:
in the U.S. over the past decade. Nuclear Power and National Identity
predictable risks/investment
Energy Policy 42 (March) 628–641. After World War II. MIT Press.
ratios due to solar and wind’s Byrne, J., Hoffman, S.M., 1996. The Hultman, N.E., Koomey, J.G., Kam-
modularity. ideology of progress and the globa- men, D.M., 2007. What history can
May 2014, Vol. 27, Issue 4 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 119
Author's personal copy
teach us about the future costs of theory and new evidence. RAND J. Sovacool, B.K., 2010. Exploring the
U.S. nuclear power. Environ. Sci. Econ. 22 (Spring (I)) 148–154. hypothetical limits to a nuclear
Technol. 1 (April) . Merrow, E.W., Shangraw, R.F., July and renewable electricity future.
IHS Costs and Strategic Sourcing, 2014. 1990. Understanding the Costs and Int. J. Energy Res. 34 (November)
Power Capital Costs Index and Eur- Schedules of World Bank Supported 1183–1194.
opean Power Capital Costs Index., Hydroelectric Projects. World Bank Sovacool, B.K., Cooper, C.J., 2013. The
http://www.ihs.com/info/cera/ Energy Series Working Paper No. Governance of Energy Megapro-
ihsindexes/index.aspx (accessed 31, Washington, DC. jects: Politics, Hubris, and Energy
February 2014). Nemet, G.F., 2006. Beyond the learning Security. Edward Elgar, London.
Koomey, J., Hultman, N.E., 2007. A curve: factors influencing cost U.S. Energy Information Administra-
reactor-level analysis of busbar reductions in photovoltaics. Energy tion, 1986. An Analysis of Nuclear
costs for U.S. nuclear plants, Policy 34, 3218–3232. Power Plant Construction Costs.
1970–2005. Energy Policy 35, Nemet, G.F., 2009. Interim monitoring Office of Coal, Nuclear, Electric
5630–5642. of cost dynamics for publicly sup- and Alternate Fuels, U.S. Depart-
Levitt, A.C., Kempton, W., Smith, A.P., ported energy technologies. Energy ment of Energy DOE/EIA-0485,
Musial, W., Firestone, J., 2011. Pri- Policy 37, 825–835. Washington, DC.
cing offshore wind power. Energy Pacala, S., Socolow, R., 2004. Stabiliza- World Commission on Dams, Dams
Policy 39, 6408–6421. tion wedges: solving the climate and Development: A New Frame-
Lyon, T.P., Mayo, J.W., 2005. Regula- problem for the next 50 years with work for Decision-making (London:
tory opportunism and investment current technologies. Science 305 Earthscan, 2000).
behavior: evidence from the ILS. (August) 968–972. Zerger, B., Noël, M., 2011. Nuclear
Electric utility industry. RAND J. Severance, C.A., 2009. Business Risks power plant construction: what
Econ. 36 (Autumn (3)) 628–644. and Costs of New Nuclear Power. can be learned from past and on-
Marshall, J.M., Navarro, P., 1991. Costs Climate Progress, Washington, DC, going projects? Nucl. Eng. Des.
of nuclear power plant construction: pp. 13. 241 (August (8)) 2916–2926.
120 1040-6190/$–see front matter # 2014 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2014.03.015 The Electricity Journal