Statistical Factors Affecting the Success of Nuclear Operations
New York: June 13, 1999
By Sankar Sunder, John R. Stephenson and David Hochma

Multivariate regression analysis linked economically viable operations to several factors: higher equity stakes by the operator; larger electrical generation operations from all fuel sources; and an owner operating relatively few nuclear plants.

I. Introduction

Nuclear power is an issue that raises passions on all sides. Proponents of nuclear power herald it as a marvel of human ingenuity, while opponents view it as the epitome of all that is wrong with technological innovation. While the moral and ethical debates on nuclear power have raged on almost since the start of the nuclear era, nuclear power is today an undeniable reality, with over 400 nuclear power plants operating in 32 countries, and producing 17 percent of the world’s electricity in 1996. The 100 or so nuclear power plants currently operating in the United States account for 14 percent of the nation’s total electricity generation capacity (about 100 GW) and about 20 percent of the nation’s total electricity generation (about 675 GWh in 1997).

Proponents of nuclear power point to the low unit costs of nuclear fuel, the relative abundance of the fuel source, and the relatively low quantities of waste products as the advantages of nuclear plants. Opponents point to the high capital costs of nuclear plants and the potential for environmental disaster as reasons to curtail the growth of nuclear power. In the past 20 years, safety standards mandated by the Nuclear Regulatory Council (NRC) and enhanced by the utility industry have made nuclear plants very safe. The high investment costs of nuclear power, however, have meant that plants that do not perform near their planned capacity turn out to be economic basket cases. Indeed, a significant portion of the “stranded costs” in America’s utilities are associated with sub-economic nuclear plants. Conversely, nuclear plants that run near capacity, and that are well managed, produce power that is very inexpensive and clean. The operations of successful nuclear plants offer valuable insights into the factors governing success in the nuclear industry, while failed or poorly performing nuclear plants provide equally important insights into managerial and operational pitfalls.

In this article, we present a statistical analysis to determine the operational, financial, technical, and managerial factors that most significantly affect the success of nuclear operations. Our study analyzes data for over 70 nuclear plants and 40 operating companies over a period of five years in order to draw conclusions that we hope will be of interest to utility commissions as they seek ways to improve rates of success in nuclear operations. Some of these conclusions will not be surprising – for example, that older plants have heavier maintenance requirements – but others are less intuitive. For instance, our observation that operators of fewer plants have lower costs suggests that any experience curve benefits associated with managing multiple nuclear facilities is overshadowed by the logistic problems of multiple facilities.

After presenting a brief history of nuclear power in America, we outline the motivations of our study and the methodology of our analysis. We end the article with the results of our study and discuss some of the managerial implications of these findings.

II. Nuclear Power in America

Commercial nuclear reactors in America owe their origins to the Atomic Energy Act of 1954 and President Eisenhower’s Atoms for Peace speech of 1953. While the first usable electricity from nuclear fission was produced in 1951 at the Idaho National Engineering Laboratory, the first large commercial nuclear reactor in America did not come about until 1957, when Duquesne Light Company began operating its Shippingport plant in Pennsylvania. By the end of the 1960s, there were 15 operating commercial reactors in the U.S., indicative of the early optimism surrounding peaceful uses of nuclear energy. By the 1970’s, nuclear power had truly come of age with the commissioning of over 50 new nuclear reactors. During this decade, nuclear power increasingly came to be viewed by electric utilities and public utility commissions as economically beneficial, thanks in large part to the rising costs of oil. The oil embargo of the 1970s further reinforced policymakers’ faith in nuclear power, since America was (and continues to be) largely self-sufficient in nuclear fuel.

The year 1979, however, brought about a sea change in the public’s view of nuclear power when Unit 2 of the Three Mile Island (TMI) power plant in Harrisburg, PA, suffered a partial meltdown and released limited radiation into the environment. For the first time, the potential for environmental disaster from civilian uses of nuclear energy seemed very real. While the nuclear industry and the NRC quickly responded to the events at TMI with better employee education, new safety procedures, public disclosure requirements, and better planning for emergencies, the damage to the reputation of nuclear power was irredeemable. During the 1980s and 1990s, as nuclear reactors that were ordered in the 1970s finally came to be commissioned, America reached a peak nuclear generation capacity, some 100GW from 110 reactors in 1996. Public opposition to new nuclear plants, however, meant that no new nuclear plans were ordered in America after the 1970s. Indeed, the newest of America’s nuclear power plants, the Watts Bar 1 unit commissioned in 1996, had been ordered in the late 1970s! The Energy Information Administration’s current models have the country’s nuclear generation capacity beginning steadily to fall off through the 1990s and 2000s unless new plants are ordered, with even world nuclear generating capacity (which has thus far been steadily rising) starting to fall off starting around 2010.

While some of the decline in the planning for new nuclear plants is attributable to environmentalist opposition to nuclear energy, the rest of the decline is due to the lack of economic viability of many existing nuclear plants. In 1989, the Fort St. Vrain nuclear plant operated by the Public Service of Colorado (PSC) in Plattville, CO, become the first nuclear reactor to be decommissioned in America after only sixteen years of operation, compared to the 40-year operating license typically granted to nuclear plants. While the Fort St. Vrain plant was a large investment, and the costs of decommissioning are estimated at over $300 million, PSC and its regulators in Colorado determined that operating the plant would in the long run be even more expensive for Colorado’s ratepayers. By some estimates, almost 30 of America’s reactors are in the same boat, and could be shut down if regulators allow full cost recovery of their associated stranded costs. That number will probably be revised downwards due to the resale of many of these plants to new owners, with regulators reimbursing the original owners for stranded costs. Recent nuclear power plants that have changed ownership are the Pilgrim Nuclear plant in Massachusetts (bought by Entergy) and Three Mile Island Unit I (bought by Peco Energy and British Energy)

III. Motivation and Methodology of the Study

At the heart of the recent repurchases of existing nuclear plants is a re-affirmation of the belief that nuclear plants, when well-managed, can deliver clean power economically, and that managerial decisions lie at the core of the factors contributing to the success or failure of nuclear operations. This study seeks to test that premise by a statistical analysis of factors that affect the success of nuclear projects. Towards this end, the study uses publicly available data for over 70 commercial nuclear plants in the U.S. for the five years from 1993 through 1997.

The study uses non-fuel operations and maintenance (O&M) costs per unit of energy produced (MWh) as the measure of success in nuclear operations. Plants with low non-fuel O&M costs are unlikely to have suffered many expensive unplanned outages and/or technical problems, while measurement on a per MWh basis reduces any biases that may be introduced by plant size. Other commonly used measures of nuclear generation success are nuclear capacity factor, and the unit cost of nuclear power, as follows:

A. Nuclear Capacity Factor

Since nuclear plants are expensive capital investments, economic recovery of capital costs necessitates maximized utilization of these plants. The capacity factor of a nuclear plant is the ratio of the actual electrical energy output of a plant in a given year to its designed capability. Thus, a plant with a nuclear capacity factor of 50 percent is one that produces 50 percent of the electrical energy it was designed for over the course of the year. Plants with high capacity factors are plants with high utilization rates, since their shut-down times are low, and power generation capacity utilization is high. A consistently high capacity factor is also an indicator that a nuclear plant is functioning with no major technical difficulties and is therefore an indicator of managerial competence.

B. Unit Cost of Nuclear Power

The unit cost of power for a nuclear plant is computed as the sum of fuel costs and non-fuel costs for the plant on a per MWh basis. A high unit cost of power is indicative of a plant where fuel and O&M expenses are high relative to the power produced by the plant, and therefore indicative of technical problems and unplanned outages in the plant. Our study chose non-fuel O&M costs per MWh as a measure of nuclear success, since nuclear plant managers do not have a significant level of control over fuel costs.

Figures 1 and 2 demonstrate that all three measures of nuclear success are largely analogous, and that plants with high capacity factors have low non-fuel O&M costs and low unit costs of power.

C. Regression Analysis

We performed multivariate linear regressions to determine the statistical factors that explain variations in non-fuel O&M costs per MWh. Our final regressions used the logarithm of non-fuel O&M costs per MWh as the dependent variable, since this was the form that gave the strongest relationships. Other forms of the regression used non-fuel O&M costs per MWh, total costs per MWh, logarithm of total costs per MWh, capacity factor, and logarithm of capacity factor as the dependent variable. The results of the regression did not vary qualitatively, irrespective of the form of the dependent variable used. Independent variables used for the regression fell into four broad categories; technical factors, operational factors, managerial factors, and financial factors. Independent variables that were strongly correlated to each other were omitted from the regressions, since correlated independent variables could skew regression results. Regressions were also tested for other statistical problems, such as non-random standard errors and auto-correlated data.

IV. Results of the Study

Our final regression analysis used seven explanatory variables, and had a goodness of fit (adjusted R-squared of 15 percent). While the R-squared was low, the F-statistic of the regression was 10.0, indicating a greater than 99 percent confidence level in the relevance of the regression’s findings. Each independent variable used in the study had a confidence level of over 90 percent associated with it. The results of the regression are discussed in greater detail in this section.

A. Technical Factors

The two technical factors considered for the study were design of the reactor cooling system – pressurized water reactors (PWR) versus boiling water reactors (BWR) – and heat rate of the plant. The design of the cooling system was included to determine if there was a statistically significant impact of reactor design on non-fuel O&M costs. Heat-rate (measured in Btu/MWh, and defined as the ratio of heat content of one unit of fuel to the electrical energy produced by the use of fuel) was included in the study to determine if aggressive engineering designs had a material impact on nuclear)&M costs. Regression analysis revealed limited statistical significance in favor of PWRs, while any effect of heat rate was found to be statistically insignificant. For this reason, the effect of heat rate was dropped from the final regression.

B. Operational Factors

Operational factors examined were plant age, the effect of whether plants were pre-or post-Three Mile Island (1979), and the size of the plant. The analysis of plant age was done to test the hypothesis that older plants would have heavier maintenance requirements. This, in fact, was verified by our analysis, which showed a statistically significant relationship between the age of a plant and O&M costs. We used a dummy variable to check if plants that were commissioned before the Three Mile Island incident performed differently from those constructed afterwards. Our analysis showed that post-TMI plants did in fact have lower O&M costs than pre-TMI. While this finding mirrors the previous one on the negative effects of increasing age, it could also signify better compliance with post-TMI regulations at nuclear plants that were commissioned after the incident. Since the age of a plant, and the dummy variable capturing whether a plant is pre- or post-TMI are heavily correlated, only one of these two variables was used in any single regression.

The final operational factor analyzed was the size of the plant, as measured by its installed generation capacity. Size was included as a metric to determine whether large plants have scale advantages that lead to their functioning more efficiently than small plants. The regression analysis showed that there was indeed a statistically significant relationship between large plant size and lowered O&M costs.

B. Managerial Factors

In examining the managerial factors, the following were analyzed: the first factor was the number of plants under management; the second was the number of nuclear MWs in equity; the third was the size of the managing utility; and the fourth and final factor was the nuclear operator’s percentage ownership in the plant. Our analysis revealed no statistical relationship between the number of nuclear MWs in equity and the dependent variables, causing us to drop this variable from our final regression. There was, however, a statistically significant relationship for the other three variables analyzed in this category. Our study concluded that there was a statistically significant relationship between the number of plants managed by an operator and operational costs. We observed that the smaller the number of plants being operated, the lower the costs, which suggests that any experience curve benefits associated with managing multiple nuclear facilities is overshadowed by the logistic problems of multiple facilities.

We also examined the size of the parent company’s generation operations versus the costs associated with running their nuclear facilities. It was observed that large companies (as measured on a large, absolute MWh sold basis) had lower costs associated with their nuclear facilities than small companies, indicating that managerial experience in generation translates into more efficient operation of nuclear assets. The final managerial variable analyzed was the percentage ownership of the operator in the plant. Our analysis revealed that nuclear costs decreased as the percentage ownership in the nuclear assets increased, thereby indicating that larger equity stakes incentivize the management of nuclear operating companies. In cases where there is joint ownership of the plants, a high proportion of equity interest by one party reduces the likelihood that management decisions are paralyzed by the need for consensus.

D. Financial Factors

The financial variables examined in our study were the number of employees in a plant, and the degree of capital investment. The number of employees showed no significant impact on costs and was therefore dropped from the analysis. In examining the capital investment we wanted to test the hypothesis that expensive plants were better. What we found, in fact, was that as the investment in plant increased, so did costs. Since bigger and newer plants tend on average to be more expensive, and this measure seemed highly correlated with other variables, we removed this variable from our regression analysis to determine if it would have any material effect on other independent variables. We found that there was no significant impact on other regressions, suggesting any effects of cross-correlation are minimal.

E. Factors That Were Not Included in the Survey

There were other independent variables that we wanted to include in our survey, but had to discard due to inadequate availability of data. Two of these variables deal with the regulatory climate, while the third deals with managerial compensation. We had to drop a proxy variable that dealt with the issue of incentive-based ratemaking for nuclear plants due to the limited nature of such regulation. We also wanted to test if certain North American Electric Reliability Council (NERC) regions had an effect on nuclear operation, since there is anecdotal evidence that nuclear plants in the South are more efficient than their counterparts in the North. We didn’t test this in our database due to the small number of nuclear power plants in certain regions. Finally, we wanted to test if the nature of managerial compensation for nuclear managers affected the efficiency of their plants. We were unable to include this in our analysis due to the relative sparseness of public disclosure on the compensation systems for nuclear managers.

V. Conclusions

Our study pointed to several statistical factors which help explain success in nuclear operations, as follows:

• PWRs have lower nuclear operating costs associated with them than BWRs, with a statistical significance of over 90 percent.
• Larger plants (those with more installed generation capacity) have lower operating costs than smaller plants, with a statistical significance of over 99 percent
• Newer plants are more efficient than older plants (statistical significance over 99 percent).
• The operating costs of nuclear plants operated by owners with a small number of nuclear plants are lower than those operated by owners with a large number of nuclear plants (statistical significance over 98 percent)
• Operating costs in nuclear plants drop with an increasing equity stake of the nuclear operator (statistical significance over 99 percent)
• Companies with large electric generation operations (from all fuel sources) had lower costs than companies with smaller generation operations (statistical significance over 98 percent); and
• Plants with large capital investments were found to have higher nuclear costs than plants with smaller capital investments (statistical significance of 95 percent)

Other factors that might help explain nuclear success, but were not included in the study, are the regulatory environment and the structure of managerial compensation. It must be noted that the results of this study are only statistical and are not meant to be interpreted as casual in nature. Still, the findings may be of interest to utilities that are seeking to invest in nuclear power or to divest their stakes in existing plants. Regulators and environmental activists would be well advised to note that the idea of nuclear power is not necessarily economically non-viable, and that there are common factors that explain the economic success of nuclear power.