Energy is the lifeblood of modern civilisation. Humans would still be hunter-gatherers if not for our ability to extract and use energy. Major advances in human wellbeing have been driven by transitions to cheaper and more plentiful energy. Examples include: the harnessing of fire, animals, wind and water power, and transitions from wood to coal, and from coal to oil and to gas [1
]. A transition to cheaper, cleaner electricity globally would improve human wellbeing and reduce the environmental impacts of electricity generation [3
People and businesses want cheap, reliable and secure energy. Globally, 1.2 billion people are still living without access to electricity [9
]. According to the World Health Organisation (WHO) [8
], “around 3 billion people cook and heat their homes using open fires and simple stoves burning biomass (wood, animal dung and crop waste) and coal”. WHO [7
] estimated that 4.3 million deaths annually are attributable to indoor air pollution and 3.7 million to ambient (outdoor) air pollution. Gohlke et al. [6
] found that increased electricity consumption per capita correlates with better health outcomes because of better access to clean water and sanitation, and reduced indoor and outdoor air pollution. They also found that access to a centralised power source is necessary to gain many of the benefits of clean power. Many of the deaths caused by indoor air pollution could be avoided if electricity replaced the burning of biomass and coal in homes, and many of the deaths attributable to outdoor air pollution could be avoided if clean technologies replaced fossil fuel for electricity generation.
Nuclear power produces comparatively little air or water pollution. Substituting nuclear for fossil fuel in electricity generation could prevent most of the deaths attributable to electricity generation. Cheap electricity increases productivity and economic growth, drives electrification for people without any electricity or with insufficient or unreliable electricity, and thereby more quickly raises living standards and human wellbeing. As the cost of electricity decreases, deployment rate increases. Transition takes place faster and the benefits are delivered sooner.
History is replete with examples of one technology replacing another [2
]. Large infrastructure transitions have commonly taken around a century [10
]. Examples are transitions to canals, railways, highways, oil and gas pipelines, telegraph, and electricity grids. Transitions typically follow an S-curve from 0 to 100% complete, with three phases: accelerating to about 20%, near-linear to about 80%, and decelerating to 100% [3
]. Electricity grids reached 50% of world population in 1960 and 80% in 2010 [11
The transition to nuclear power began in 1954 with the first reactor connected to the grid. Until the 1970s, it was envisaged that nuclear would emulate earlier energy transitions. For example, Wilson [12
] projected that nuclear power would supply 14 to 21% of world primary energy by 2000. However, the transition to nuclear reached 4% by 1970, then stalled [3
]. The deployment rate of nuclear capacity is currently less than in 1972; the transition has been stalled for 44 years.
The rate that technology transitions take place depends, in part, on the technologies being ‘fit-for-purpose’ and on the learning rates that occur during the transition period. To accelerate the transition to reliable, cheap, clean, safe and comparatively environmentally benign electricity generation, policies need to focus on ways to improve the learning rates and deployment rates of technologies that meet requirements. Historical learning rates provide insight into what rates may be achievable and what could be done to return to rapid rates.
The concept of learning rates, or cost experience curves [I]
, is widely used to quantify the rate at which costs reduce as experience is gained. Learning rate is the fractional reduction in cost per doubling of cumulative capacity or production. Rubin et al. [13
] explain how to calculate learning rates, and summarise learning rates for selected electricity generation technologies. However, their paper has limited information on nuclear power learning rates, and none before 1972 or after 1996.
Lovering et al. [14
provide a comprehensive analysis of nuclear power construction cost experience of early and recent reactors in seven countries; their analysis covers 58% of the reactors ever constructed for electricity generation, between 1954 and 2015. While there have been many studies of the cost escalation of nuclear power plants (e.g., [15
], and others cited in Lovering et al. [14
]), most are for the US and France only, and cover only periods since the 1970s. To the author’s knowledge, there are no comprehensive studies, other than Lovering et al. [14
], that cover the full period of global commercial nuclear power reactor operation, nor any studies that provide the learning rates over the full period, and that highlight their reversal, which began in the late-1960s.
This study extends the literature by providing learning rates of nuclear power reactors for the seven countries analysed by Lovering et al. for the full period from 1954 through 2015. The aim is to answer two questions. What were the global benefits forgone as a consequence of the reversal of learning rates and the stalled deployment rates? What are the policy implications?
Using counterfactual analysis, Kharecha and Hansen [17
] estimated that electricity generated by nuclear power avoided 1.84 million air-pollution-related deaths and 64 Gt of CO2
emissions between 1971 and 2009. The current analysis also uses a counterfactual approach. Lovering et al. data were re-analysed to calculate the historical learning rates and deployment rates of nuclear power, and to project the early rates to 2015. Evidence of disruption to the learning and deployment rates is presented and some of the benefits forgone are quantified. Estimates are presented for:
what the Overnight Construction Cost (OCC) [III],[IV]
of nuclear power could have been in 2015 if the early learning and deployment rates had continued
the additional electricity that could have been generated by nuclear power if the early deployment rate trends had continued
the number of deaths and quantity of CO2 emissions that could thereby have been avoided.
The analysis finds that the benefits forgone because of the disruption, and the resulting stalled transition from fossil fuels to nuclear power, are substantial.
The purpose of this paper is to publish the evidence and the consequences of the disruption, and to suggest an approach to removing the impediments that are delaying progress. It does not explore the causes of the disruption and cost escalations thereafter; that would require extensive studies beyond the scope of this paper.
To summarise, this study provides learning rates for a full set of reactors in seven countries, covering builds from 1954 through to projects that had been completed by the end of 2015, covering 58% of all power reactors ever built globally. It also provides global deployment rates for that period. It estimates the extra electricity that could have been generated by nuclear power since 1980 and what OCC would have been in 2015 if the learning and deployment rates had not been disrupted. It compares the projections to the actuals to estimate forgone benefits of the disruption. It suggests an approach to removing the impediments that are retarding the transition to nuclear power.
3. Results and Discussion
3.1. Learning Rates
has a chart for each of the seven countries and one for all seven combined; trendlines were fitted to the data points before and after the trend reversal points. The equation for each trendline is shown on the charts.
OCC (2010 US $/kW) plotted against cumulative global capacity (GW) of nuclear power reactors, based on construction start dates; regression lines fitted to points before and after trend reversals.
OCC (2010 US $/kW) plotted against cumulative global capacity (GW) of nuclear power reactors, based on construction start dates; regression lines fitted to points before and after trend reversals.
To compare trends for the seven countries, Figure 3
shows all the regression lines. Japan and France had the fastest pre-reversal learning rate; South Korea had a similar rate since it started building reactors in 1972, although it started from a high OCC after the reversal and initial rapid cost escalation in the other countries.
lists the learning rates for both periods in each country for both the cumulative global and the cumulative country capacity. The sixth and seventh columns are the selected reversal point (cumulative global capacity of construction starts, and approximate year it occurred) for each country. The last column is the projected OCC at 497 GW cumulative global capacity if the pre-reversal learning rates had continued.
Learning rates are affected by the growth of cumulative capacity both globally and locally. Following Lovering et al., cumulative global capacity was used as the reference. Figure 4
plots the learning rates against the time span of the construction starts for each period in each country.
and Figure 4
show that, before the reversal, OCC learning rates were 23% in the US, 27% to 35% in the other countries except India (where it was 7%), and 24% for all countries combined. At the reversal, learning rates changed abruptly and became negative (−94% in the US, −82% in Germany, −23 to −56% in the other countries, except in South Korea, and −23% for all seven countries); South Korea started building nuclear power plants after the initial rapid cost-escalation period, achieving a 33% learning rate since 1972. The fact that fast learning rates existed up to about 1970, and in South Korea since, suggests they could be achieved again [IX]
The US’s post-reversal learning rate was the worst of the seven countries. The reversal occurred one to five years later in the other countries and the real cost increase was not as severe as in the US. This suggests the US may have negatively influenced the development of nuclear power in all seven countries (and probably all countries). It also shows that technology learning and transition rates can change quickly and disrupt progress, in this case delaying progress for about half a century so far.
3.2. Deployment Rates and Projections to 2015
shows the annual global capacity of construction starts [X]
and commercial operation starts from 1954 to 2015 [18
]. The capacity of construction starts was accelerating until about 1970, peaked in 1976, then stalled. The annual capacity of commercial operation starts peaked in 1985, averaged 30 GW per year from 1984 to 1986, then declined rapidly and has not recovered. IAEA [24
] shows grid connections peaked at 31 GW per year in 1984 and 1985 and declined rapidly thereafter.
shows cumulative global capacity of construction starts and commercial operations starts plotted against time (top panel), and projections of what they would have been in 2015 if the early deployment rates had continued (bottom panel).
summarises the cumulative global capacity of actual and projected construction starts and the capacity in commercial operation at the end of 2015 for each scenario.
The Linear and Accelerating projections of cumulative global capacity by 2015 in Table 2
represent scenarios calculated on the basis of the stated deployment rate assumptions. The increases in projected cumulative global capacity by 2015 compared with Actual are large. It is useful to compare these scenarios with projections made in the 1970s. For example, the Accelerating deployment rate projects a global nuclear capacity of 1152 GW by 2000. The Workshop on Alternative Energy Strategies (WAES) [12
], projected global nuclear capacity in 2000 at between 913 GW and 1722 GW [XI]
. So the present projection is quite consistent with the outlook of 40 years ago.
3.3. Projected Overnight Construction Costs in 2015
lists the projected OCC in 2015 and the percentage reduction from the actual OCC [20
] for the six countries that were constructing reactors before the learning rate reversals. Actual OCC for Canada, Germany and India are approximate, as noted in Section 2.3
If the pre-reversal learning rates had continued, with the Actual deployment unchanged, until 2015 when cumulative global capacity of construction starts was 497 GW, the OCC of nuclear power would be 5 to 15% of what it was in 2015 (except in India); for example, the OCC would be $
349/kW in the US, $
257/kW in France, and $
484/kW in Japan (Table 3
). These are much lower than the OCC of fossil fuel and other alternative electricity generation technologies [20
If the pre-reversal learning rates and the Linear and Accelerating deployment rates had continued, the OCC would be approximately 2% to 10% of what it was in 2015 (except in India where it would be 31 to 33%) (Table 3
These are striking cost reductions that, to be achieved, would have required pre-reversal learning rates and deployment rates to continue. If the rapid learning and deployment rates that prevailed pre-reversal could be achieved again, nuclear power would become much cheaper than fossil fuel technologies in the future. Some may regard this as too optimistic. However, there is no apparent physical or technical reason why they could not have continued and cannot prevail again. They have prevailed in South Korea over the past 40 years (Figure 4
), and there are examples in other complex technologies and industries of cost reductions at similar rates that persisted over the past 50 years at the same time as the OCC of nuclear power was increasing rapidly [XII]
3.4. Extra Nuclear Electricity, Avoided Deaths and CO2
shows the annual electricity generated by fuel type for the three deployment rate scenarios: Actual, Linear and Accelerating.
shows that, at the Linear deployment rate, the extra nuclear generation from 1985 to 2015 could have substituted for 69,000 TWh of mostly coal-generated electricity globally and avoided approximately 4.5 million deaths (from outdoor air pollution and all other causes in the respective energy chains, but not including deaths that could have been avoided by increasing access to clean water and sanitation services) and 69 Gt CO2
At the Accelerating rate, the extra nuclear generation could have exceeded the actual generation from coal by year 2000 (assuming electricity demand did not change). If the extra nuclear generated electricity had substituted for coal and gas generation, about 9.5 million deaths and 174 Gt CO2 may have been avoided.
In 2015 alone, if the extra nuclear generation had replaced coal and gas generation, and electricity demand was unchanged, nuclear could have:
3.5. Other Benefits Forgone
If the pre-reversal learning rates had continued, OCC, and consequently the cost of electricity, would undoubtedly have declined. Arguably this would have led to other benefits not estimated in this analysis, such as increased productivity, faster economic growth, improved standard of living, and better health and education outcomes.
The declining cost of electricity would probably have caused increasing demand and consumption. Substitution of electricity for fossil fuels for heat and transport may have proceeded faster. With declining costs and increasing demand, electricity grids may have expanded faster with more people being connected. Alstone et al. [11
] charts the world population and the number of people who were connected to an electricity grid for the period 1830 to 2013. If grid connections had continued to accelerate at the rate that prevailed between 1950 and 1975, many of the 1.2 billion people who were not grid-connected in 2015 could have been.
With increased consumption, electricity could have substituted for some combustion of fuels by the 3 billion people who cook and heat their homes using open fires and simple stoves burning wood, animal dung, crop waste and coal, thereby reducing the 4.3 million deaths per year attributable to indoor air pollution [7
]. And clean water and sanitation systems could have been provided to more people, reducing deaths from contaminated water [6
The benefits forgone may have been substantially greater than estimated in the present counterfactual analyses. World energy consumption slowed in the 1970s [3
] and GDP growth rate slowed too [25
]. If the transition from fossil fuels to nuclear power had not been disrupted, world GDP growth may not have slowed as much. The global economy could have been significantly different from what it is now.
3.6. Policy Implications
Policies that increase the real cost of energy would be damaging economically, and are unlikely to be politically sustainable and, therefore, unlikely to succeed in the long term. To reduce the emissions that are detrimental to health and the environment, countries will need access to low-emissions technologies that are cheaper than high-emissions technologies. In this case, carbon pricing and command-and-control policies, such as incentives for low emissions and penalties for high emissions technologies, would not be required.
Cheap electricity increases productivity and economic growth, drives faster electrification for the people without access to electricity or with insufficient and/or unreliable electricity, and thus more quickly lifts the world’s population to higher living standards. As electricity costs decrease, the deployment rate increases and capacity doublings occur faster. Consequently, costs reduce faster; i.e., we progress more quickly down the learning curve [XV]
]. Technology transition takes place faster and the benefits are delivered sooner.
These benefits could be achieved in the future if the impediments that disrupted the transition to nuclear power are removed. While this paper does not attempt to discuss the causes of the disruption and cost escalations thereafter, many others have (e.g., Cohen [15
], Grubler [16
], and Lovering et al. [14
] cites a number of studies). A likely root-cause of many of the causes discussed in the literature was the growing concern about the safety of nuclear power, fanned by the anti-nuclear protest movement, which began in the mid-1960s (Daubert and Moran [27
]; Wyatt [28
]), and the ongoing political, legislative and regulatory responses to the concerns.
The fact that rapid learning and deployment rates prevailed in the past suggests they could be achieved again. To achieve them, it is suggested four steps are needed:
First, recognise that the disruption to the transition occurred and the impediments to progress continue to this day.
Second, recognise the consequences of the disruption for the global economy, human wellbeing and the environment, and the ongoing delays to progress.
Third, identify the root causes of the disruption and cost escalations since, and the solution options.
Fourth, implement policies to remove impediments that are retarding the transition.
The benefits forgone cannot be recovered, but future benefits can be increased by amending the policies that caused the cost increases and slowed the deployment of nuclear power. Human wellbeing could improve faster if the impediments that are slowing the development and deployment of nuclear power are removed.
From 1954 to the late-1960s, learning rates of nuclear power OCC were positive (i.e., OCC decreased as capacity increased). In the late-1960s, learning rates turned negative (i.e., OCC increased as capacity increased) and have remained negative ever since in all the seven countries analysed, except South Korea.
The disruption to learning rates was followed by stalled deployment rates.
If the pre-1970s learning rates had continued, and assuming the actual deployment did not change, OCC of nuclear power in 2015 could have been around 5 to 15% of what it actually was.
If both the pre-1970s learning rates and the Linear or Accelerating deployment rates had continued, OCC in 2015 could have been around 2 to 10% of actual.
If deployment had continued to add 30 GW to global capacity per year since 1985, 69,000 TWh of extra nuclear electricity could have been generated. Assuming this replaced coal-fired electricity generation, 4.2 million deaths and 69 Gt CO2 may have been avoided.
If deployment had continued from 1976 at the Accelerating rate that prevailed from 1960 to 1976, 186,000 TWh of extra nuclear electricity could have been generated. Assuming this extra nuclear generation replaced coal- and gas-powered electricity generation, 9.5 million deaths and 174 Gt CO2 may have been avoided.
In 2015 alone, assuming electricity demand was unchanged, nuclear could have replaced between 49% of coal-powered generation (at the Linear deployment rate) and 100% of coal-powered plus 76% of gas-powered generation (at the Accelerating deployment rate), thereby avoiding between 273,000 and 540,000 deaths and between 4.5 and 11 Gt CO2.
The policy implications are substantial. Benefits would be available in the future by returning to the pre-disruption learning and deployment rates. To achieve this requires a recognition of the disruption and its consequences, identification of its causes, and amelioration of the impediments that are slowing progress.