¹ International Atomic Energy Agency, Nuclear Power Experience, Proceedings of a Conference on Nuclear Power Experience (Vienna, 1983), vol. I. p. 51.Google Scholar

² For other case studies of competing technologies, see David, Paul,“CLIO and the Economics of QWERTY,” American Economic Review, 75 (05 1985)Google Scholar; Teubal, Morris and Steinmueller, Edward,“Government Policy, Innovation and Economic Growth: Lessons From a Study Satellite Communications,” Research Policy, II (10 1981)Google Scholar; and Karlson, S. H.,“Adoption Competing Inventions by United States Steel Producers”, Review of Economics and Statistics, (08 1986).CrossRefGoogle Scholar

³ See David, Paul with Bunn, Julie, “The Economics of Gateway Technologies and Network Evolution: Lessons from Electricity Supply History,” Information Economics and Policy, 3 (No. 2, 1988).CrossRefGoogle Scholar

⁴ See Arthur, Brian, “Competing Technologies and Economic Prediction,” Options, International Institute for Applied Systems Analysis (Laxenburg, Austria, 1984)Google Scholar; and McLaughlin, Charles, “The Stanley Steamer: A Study in Unsuccessful Innovation,” Explorations in Entrepreneurial History, 7 (10 1954).Google Scholar

⁵ For a survey of the recent competing technologies literature, see Arthur, Brian, “Competing Technologies: An Overview,” in Dosi, G. et al. ,, eds., Technical Change and Economic Theory (London, 1988). “Superior” here means “inherently superior.” Theoretical results indicate that under a variety of conditions, only one technology will survive in the market. Given this result, the superior technology is that which, if it were to be the surviving one, would maximize net benefits from the technology choice process. This is an ex post definition of “superior.”Google Scholar

⁶ “Technology” here is a generic term. Following Kenneth Arrow: “At any moment of time, the new capital goods incorporate all the knowledge then available, but once built their productive efficiency cannot be altered by subsequent learning.” Arrow, Kenneth, “The Economic Implications of Learning by Doing,” Review of Econo, nic Studies, 29 (06 1962), p. 157. Technologies improve but particular instances of them do not.Google Scholar

⁷ This model is presented in Brian Arthur, “On Competing Technologies and Historical Small Events: The Dynamics of Choice Under Increasing Returns” (International Institute for Applied Systems Analysis Working Paper WP-83–90, 1983); and in Arthur, Brian, “Competing Technologies, Increasing Returns, and Lock-In by Historical Events,” Economic Journal, 99 (03 1989). It is possible to find sets of parameters and functions such that only one technology is ever used, and so these results will be violated. The results are very general, however.CrossRefGoogle Scholar

⁸ Many other models make the same assumption. See, for example, Farrell, Joe and Saloner, Garth,“Standardization, Compatibility and Innovation,” Rand Journal of Economics, 16 (Spring 1985)CrossRefGoogle Scholar; Farrell, Joe and Saloner, Garth,“Installed Base and Compatibility,” American Economic Review, 76 (12 1986)Google Scholar; Katz, Michael and Shapiro, Carl,“Network Externalities, Competition and Compatibility,” American Economic Review, 75 (05 1985)Google Scholar; and Katz, Michael and Shapiro, Carl,“Technology Adoption in the Presence of Network Externalities,” Journal of Political Economy, 94 (08 1986).CrossRefGoogle Scholar

⁹ See Cowan, Robin, “Backing the Wrong Horse: Sequential Technology Choice Under Increasing Returns” (Ph.D. diss., Stanford University, 1987). The multiarmed bandit is a problem studied in probability theory, characterized as a slot machine with several arms. The arms are assumed to have different probabilities of paying out, and the object is to play the arms one at a time in any order so as to maximize the expected present value of the winnings.Google Scholar

¹⁰ When an atom is split, neutrons are released which bombard other atoms, causing them to split and so creating a chain reaction. The chain reaction generates considerable heat which is used to turn turbines which generate electricity. To sustain a chain reaction there is an optimal speed, or energy level, for the neutrons. By causing the neutrons to travel through particular substances in the reactor core (moderators), this optimal energy level can be obtained. For an easily accessible account of the technology of nuclear power reactors, see Bupp, Irwin and Derian, Jean-Claude, Light Water: How the Nuclear Dream Dissolved (New York, 1968), chap. 1, fn. 3.Google Scholar

¹¹ Deuterium is a naturally found isotope of hydrogen. D₂O is found in nature, in the ratio of approximately I part in 5,000 parts H₂O.Google Scholar

¹² The sources for the following paragraphs are Nucleonics, 15 (June 1957)Google Scholar; Marshall, W., ed., “Reactor Technology,” Nuclear Power Technology, vol. I (Oxford, 1983)Google Scholar; Cottrell, Alan, “The Pressure on Nuclear Safety,” New Scientist (Mar. 25, 1982)Google Scholar; Green, David, “AGR v PWR: The Debate Continues,” Energy Policy, 14 (02 1986)CrossRefGoogle Scholar; Piran, M. and Murgatroyd, W., “Fuelling Costs of Nuclear Reactors,” Energy Policy, 12 (03 1984)Google Scholar; and International Atomic Energy Agency, Nuclear Power Reactors in the World (Vienna, 1987).Google Scholar

¹³ Nucleonics, 15 (June 1957), p. 71.Google Scholar

¹⁴ The source for these figures is International Atomic Energy Agency, Nuclear Power Reactors in the World, table 17. A reactor is included in the average for the entire time it is connected to the electricity grid up to 1987. These figures do not control for things such as different regulatory regimes. If the average performance within a country is used as an observation point, light water looks much better, largely due to extremely good performance in Belgium and Sweden, though still not as good as heavy water.Google Scholar

¹⁵ Mcintyre, Hugh,“Natural-Uranium Heavy-Water Reactors,” Scientific American, 233 (10 1975).CrossRefGoogle Scholar

¹⁶ Fanjoy, G. R.,“Generating Costs From Candu,” European Symposium on the Candu Reactor, London (03 1982)Google Scholar; and Hydro, Ontario,“Ontario Hydro CANDU Operating Experience,” NGD-9 (1987). The higher estimate uses the world average load factor; the lower estimate assumes that the annual load factor would be higher under Ontario Hydro management policies.Google Scholar

¹⁷ For more discussion on the merits of other technologies, see Agnew, Harold, “Gas-Cooled Nuclear Power Reactors,” Scientflc American, 244 (06 1981)Google Scholar; Weinberg, Alvin and Spiewak, Irving, “Inherently Safe Reactors and a Second Nuclear Era,” Science, 29 (06 1984)Google Scholar; and Marshall, Eliot,“The Gas Reactor Makes a Comeback,” Science, 29 (05 1984).Google Scholar

¹⁸ See Mullenbach, Phillip, Civilian Nuclear Power: Economic Issues and Policy Formation (New York, 1964), p. 39.Google Scholar

¹⁹ See Bupp and Derian, Light Water, p. 6.Google Scholar

²⁰ Rosenberg, Nathan, Inside the Black Box: Technology and Economics (Cambridge, 1982), p. 122.Google Scholar

²¹ Hinton also added: “Let us remember that the first movement onward from the Bolton and Watt engine was really made by Trevithick when he built his high-pressure steam engine on the Thames, with its cast iron boiler which blew up, killed eight men, nearly ruined him and set back the development of the steam engine by a great many years. We must make certain that we do not do that sort of thing…” United Nations, Proceedings of the Second International Conference on Peaceful Uses of the Atom (Geneva, 1955), p. 368. See also the proceedings of the First Conference.Google Scholar

²² See Bupp and Denan, Light Water, p. 45.Google Scholar

²³ Ibid., p. 46.

²⁴ DeLeon, Peter, Development and Diffusion of the Nuclear Power Reactor: A Comparative Analysis (Cambridge, MA, 1978), p. 200.Google Scholar

²⁵ A turnkey contract is one in which a price is fixed before construction begins, and any unforeseen costs are borne by the designers, in this case Westinghouse and General Electric.Google Scholar

²⁶ Joskow, Paul and Rozanski, G. A.,“The Effects of Learning by Doing on Nuclear Plant Operating Reliability,” Review of Economics and Statistics, 61 (05 1979).Google Scholar

²⁷ Ibid., p. 167.

²⁸ Mooz, W. E., Cost Analysis of Light Water Reactor Power Plants (Prepared for the Department of Energy, Rand Corporation, R-2304-DOE, Santa Monica, 1978).Google Scholar

²⁹ Zimmerman, Martin, “Learning Effects and the Commercialization of New Energy Technologies: The Case of Nuclear Power,” Bell Journal of Economics, 13 (Autumn 1982). He estimates that the completion of the first plant reduces the cost of future plants by 12 percent. Completing the second plant reduces costs further by 4 percent.CrossRefGoogle Scholar

³⁰ International Atomic Energy Agency, Nuclear Power Experience, vol. 1, pp. 137, 170.Google Scholar

³¹ Strong static increasing returns are present as well but are of less interest from the point of view of this article.Google Scholar

³² At the time, there was a single European reactor technology, namely the French gas graphite. The French saw Euratom, in part, as a way to get their technology adopted throughout Europe.Google Scholar

³³ This report was jointly authored by Franz Estel, German vice president of the European Coal and Steel Community; Francesco Giordiani, former president of the Italian Atomic Energy Commission; and Louis Armand, president of the French National Railroad Company.Google Scholar

³⁴ Bupp and Derian, Light Water, p. 37.Google Scholar

³⁵ Particularly important was the Oyster Creek station, announced in 1963. This was an early turnkey plant built by GE which promised power at 4 mills per kWh. This was a decrease of 60 percent from the costs quoted by the Atomic Energy Commission in 1962. The “bandwagon market” refers to the time 1962 to 1965 during which U.S. utilities ordered 13 generating stations.Google Scholar

³⁶ There is a distinct similarity between the actions of the U.S. government in this role and those of General Electric and Westinghouse in offering turnkey contracts. U.S. government subsidies applied only to reactor varieties tested in the United States. (Hewlett, R. G. and Duncan, F., Nuclear Navy, 1946–1962 [Chicago, 1974], p. 135.) This policy gave considerable assistance to Westinghouse and General Electric and their European subsidiaries.Google Scholar

³⁷ The faith in the light water technology displayed by European willingness to abandon their own gas graphite technology could only encourage utilities in the United States to believe that light water was good. The apparent breakthrough in light water, evidenced by the rash of orders in the United States, in turn encouraged the Europeans to continue to use the U.S. technology.Google Scholar

³⁸ This section draws heavily on Williams, R., The Nuclear Power Decisions: British Policies, 1953–78 (London, 1980).Google Scholar

³⁹ Ibid., p. 234.

⁴⁰ Interestingly, in 1962 this reactor was considered by the Atomic Energy Agency to have better development potential than light water. (Ibid., p. 197.) In 1971 it was referred to as “the best of American BWR [a light water technology], Canadian and British technologies.” See p. 213.

⁴¹ Ironically, recent experience with the AGR has been very good. In terms of reliability and availability, it has looked better than other technologies since the mid-1980s.Google Scholar

⁴² This decision was reconsidered in the 1980s but was not in the end changed.Google Scholar

⁴³ In 1964 Sweden completed a 10 MW reactor which was shut down 10 years later. In France in August 1967 a 70 MW heavy water moderated, gas-cooled reactor was brought on line. Later in 1967 the United Kingdom brought on line a 92 MW heavy water prototype similar in design to the Candu.Google Scholar

⁴⁴ Hafstad, Lawrence,“Reactors,” Scientific American, 184 (04 1951), p. 43. Quoted in Bupp and Derian, Light Water, p. 32.CrossRefGoogle Scholar

⁴⁵ The technologies were being studied under the GE project; in naval work on gas and light water coolants; and in AEC work on graphite reactors and its work on light water.Google Scholar

⁴⁶ These were light water; liquid-metal-cooled, graphite-moderated; aqueous homogeneous; and fastbreeder reactors. See Mullenbach, Civilian Nuclear Power, p. 131.Google Scholar

⁴⁷ Stern, Theodore,“Appraisal of Reactor Systems for Central-Station Power Plants,” Chemical Engineering Progress Symposium Series, part I, 54 (11 1954), summarized these studies. The cheapest electricity, 6.4 mills per kWh, would be generated by boiling water, a light water technology. The next cheapest would be a fast breeder, 6.5 mills per kwh, followed by pressurized water, another light water technology, 6.8 mills per kWh. The most expensive was the sodium- graphite technology at 10.3 mills per kWh. Stern noted, though, that in analyses of this sort the difference between 6.4 and 10.3 “may not be out of the margin of uncertainty.”Google Scholar

⁴⁸ Lane, J. A., “An Evaluation of Geneva and Post-Geneva Nuclear-Power Economic Data,” The Economics of Nuclear Power, series 8 (New York, 1957).Google Scholar

⁴⁹ In defense of water reactors, their cost estimates assumed plants with relatively small generating capacity, which, given the faith in increasing returns to scale, would appear to put them at a disadvantage.Google Scholar

⁵⁰ The types were liquid-metal-cooled, heavy water moderated; gas graphite; graphite moderated, liquid-metal fuel; homogeneous; two variants of light water; and an organic hydrocarbon cooled and moderated reactor. Lane, “An Evaluation of Geneva.”Google Scholar

⁵¹ The participants in the debate were J. R. Menke, president of the Nuclear Development Corp., and W. B. Lewis, vice president of Atomic Energy Canada Limited, both of whom spoke in defense of natural uranium reactors;and Chauncey Starr of North American Aviation and W. E.Shoupp of Westinghouse, both of whom spoke in defense of enriched uranium reactors. See Nucleonics, 15 (June 1957), p. 68.Google Scholar

⁵² Ibid., p. 70.

⁵³ Weinberg, Alvin,“Power Reactors,” Scientific American, 191 (12 1954), p. 36.CrossRefGoogle Scholar

⁵⁴ There are two other technologies that overcome this problem. One is the closed cycle submarine, in which diesel exhaust gas is recycled and mixed with oxygen which has been stored in cylinders, and then re-used. The second is a snorkel submarine, in which air for combustion while the submarine is submerged is obtained from a snorkel arrangement which trails the submarine on the surface. After the war, the U.S. Navy was working on all three of these technologies, only one of which has survived.Google Scholar

⁵⁵ Hughes, Jonathan, The Vital Few: The Entrepreneur and American Economic Progress (New York, 1986).Google Scholar

⁵⁶ His other studies were not completed before Rickover had made his decision.Google Scholar

⁵⁷ Combustion Engineering had also been drawn into light water by the navy and developed a reactor for the hunter-killer submarine, Tullibee, which was launched in 1960.Google Scholar

⁵⁸ Hewlett, R. G. and Duncan, F., A History of the United States Atomic Energy Commission, vol. 2: Atomic Shield, 1947/1952 (University Park, PA, 1969), p. 226 ff.Google Scholar

⁵⁹ Rickover did not want to abandon the carrier project in favor of Shippingport. His proposal was a fall-back position in which he could work on a reactor which would provide valuable information for any future carrier project if the current one was put on the shelf.Google Scholar

⁶⁰ See Weinberg, “Power Reactors.”Google Scholar

⁶¹ Hewlett and Duncan, A History of the United States Atomic Energy Commission, p. 231.Google Scholar

⁶² See Cowan, “Backing the Wrong Horse,” chap. 4; and Rosenberg, Inside the Black Box.Google Scholar

⁶³ Hertsgaard, M., Nuclear mc: The Men and Money Behind Nuclear Energy (New York, 1983), p. 25.Google Scholar

⁶⁴ Ibid., p. 27.

⁶⁵ Ibid., p. 28.

⁶⁶ Recall that the Atomic Energy Commission wanted to do more research before building Shippingport but was overridden.Google Scholar

⁶⁷ By the end of 1960, 13 nuclear ships had been launched, and a further 33 were under construction. The two firms had completed or begun construction on eight power reactors in Europe and the United States.Google Scholar

⁶⁸ McKittenck, John. General Electric vice president for corporate planning, quoted in Hertsgaard, Nuclear Inc, p. 42.Google Scholar

⁶⁹ Significantly, three of the six utilities involved had, at the same time, plans to build other generating stations using light water. None of these plans were cancelled, and by 1974 construction had begun on all of them.Google Scholar

⁷⁰ Bupp and Derian, Light Water, are also convinced that the current dominance by light water is due to its early acceptance in the United States, and subsequent rapid spread into Europe. They emphasize, however, the way in which reactors were built largely on expectations of future performance. Many reactors were ordered based on the claims of manufacturers that the next, bigger generation had achieved enormous cost reductions. These claims turned out to be far from true. The degree to which orders (both in the United States and Europe) preceded experience is astounding. In 1968, for example, the largest light water reactor that had been operating for a year or more was 200 MW. In contrast the mean size of reactor ordered that year was 926 MW (see figure 4–1, p. 73). This sort of advertising (and to be fair, the willingness to accept it) generated enough light water orders, that, again, other technologies were left behind.Google Scholar