Probability Distributions of Net Present Values for U.S. Nuclear Power Plants by Chris Trayhorn, Publisher of mThink Blue Book, January 15, 2002 Introduction From the dawn of electricity deregulation in the United States until mid-1998, a pessimistic cloud hung over the nuclear power industry. This was expressed in early 1998 by the CEO of Commonwealth Edison, James O’Connor, after three other plants (Haddam Neck, Maine Yankee, and Big Rock Point) had closed before the end of their licensed lifetimes in 1996 and 1997: “Our decision to close Zion Station is based on economics. A thorough analysis of the projected costs to produce power at the station and the expected price of electricity in a deregulated market led us to one conclusion: Zion Station will not be able to produce competitively priced power in a deregulated marketplace over the remaining useful life of the plant.” On June 25 of 1998, Commonwealth Edison paid up to $6,000 per megawatt-hour (MWh) in the fiercely competitive marketplace. The pessimistic cloud lifted. The last nuclear power plant to retire was Millstone 1 in July 1998 after being inoperable since December 1995.Two nuclear power plant operators began to aggressively purchase plants in 1998 and 1999. AmerGen, a joint-venture between PECO (originally Philadelphia Electric Company, operator of Limerick 1 & 2 and Peach Bottom 1 & 2) and British Energy (operator of 15 Advanced Gas Reactors in the United Kingdom) won bids to acquire Three Mile Island 1, Clinton, and Oyster Creek. Entergy (operator of Arkansas Nuclear One 1&2, Grand Gulf in Mississippi, and River Bend and Waterford in Louisiana) bought Pilgrim. Announced prices per kilowatt (kW) ranged from $128 to $16 (Table 1). Similar offers were made for Nine Mile Point 1 & 2 ($117 for Unit 1 and $136 for Unit 2) and for Vermont Yankee ($44), but these sales were suspended in the belief that these plants were worth more than what was being offered. Soon after AmerGen concluded these purchases, PECO announced its intended merger with Commonwealth Edison (operator of 11 nuclear units in Illinois) to become Exelon. The following year Carolina Power & Light Company (operator of Brunswick 1&2, Robinson, and Shearon Harris) merged with Florida Power Corporation (operator of Crystal River) to form Progress Energy. Table 1 – Nuclear Power Plant Sales, 1998-2001. see larger image Also in 2000, Entergy acquired Fitzpatrick and Indian Point 3 (jointly) at an announced price of $450/kW and Indian Point 1 & 2 at an announced price of $372/kW. In 2001 two more nuclear power operators acquired plants. Dominion (originally Virginia Electric & Power Company, operator of North Anna 1 & 2 and Surry 1 & 2) paid $640/kW for Millstone 1, 2, & 3. Constellation (originally Baltimore Gas & Electric Company, operator of Calvert Cliffs 1 & 2) paid $524/kW for Nine Mile Point 1 & 2. Further, Entergy concluded the purchase of Vermont Yankee for $290/kW. Announced prices should be adjusted for: • Fuel Inventory Payments, • Purchase Power Agreements, and • Decommissioning Costs and Trust Funds1. Once these adjustments are made, the sale prices for plants during 1998 and 1999 ranged from $46/kW for Clinton to -$74/kW for Oyster Creek (Table 1). (A negative sales price implies that the seller paid the buyer to accept ownership of the plant’s assets and liabilities.) Were these prices reasonable? Did they reflect the Net Present Values of these plants? What is the probability that these NPVs were negative? The Net Present Value of Nuclear Power Plants The Net Present Value of any project is total revenues, TR, minus total costs, TC, from each year of the project, discounted to the present and summed2. For the electricity generators (TR — TC) is equal to the price of electricity, P, minus Average Variable Cost, AVC, in dollars per megawatt-hour ($/MWh) times the quantity of electricity sold, Q (in MWh). Under deregulation, the price of electricity is determined in the market. AVC is equal to the sum of (1) fuel costs per MWh plus (2) Operation and Maintenance expenses per MWh plus (3) Capital Additions per MWh. (Capital Additions include all “going-forward,” depreciable investments.) The quantity of electricity is equal to the Capacity Factor times the size of the plant (in MW) times the number of hours in the year (usually 8,760). As discussed in Rothwell3, we can transform AVC into a “minimum” AVC/MWh (this is AVC/MWh at a 100 percent Capacity Factor) divided by the Capacity Factor. For example, with an average variable cost of $30/MWh at an 80 percent capacity factor, the “minimum” value of average variable cost would be $24/MWh at a 100 percent capacity factor. Therefore, to forecast the NPV of an NPP we need for each year (t) 1. An appropriate discount rate, r (percent), 2. Projected prices of electricity, Pt ($/MWh), 3. The size of the plant, W (MW) 4. Projected Capacity Factors, CFt (percent), 5. Projected “minimum” Fuel costs per MWh, minFt ($/MWh), 6. Projected “minimum” O&M expenses per MWh, minO&Mt ($/MWh) 7. Projected “minimum” Capital Additions per MWh, minKt ($/MWh). For example, assuming constant values, if the P is $35/MWh, minAVC is $24/MWh, CF is 80 percent, and W is 1000MW, then profit (without considering annual payments for plant construction and financing) is $35 million per year. Assuming a discount rate of 9 percent (real, i.e., not including inflation) and a remaining lifetime of 15 years, the Present Worth Factor would be about 8.0; so the NPV would be $280 million Further, dividing by the number of kW (i.e., 1 million), the NPV/kW is $280. If the market price of the plant is less than $280/kW, the buyer gains. If the book value of the plant is greater than the market price, the difference is either written off as a loss (to the shareholders) or recovered as a stranded asset (from the customers), or distributed between the shareholders and the customers. To estimate the NPV/kW for each NPP: • I forecast the Capacity Factor, “minimum” Fuel, “minimum” O&M expenses, and “minimum” Capital Additions with a simple Ordinary Least Squares estimate of a plant or unit-specific constant and a time trend, controlling for plant or unit type, age, and size4, and • I use U.S. Energy Information Administration forecasts of deregulated electricity prices for each state5. These projections are given for a discount rate of nine percent (real) in the last column of Table 1. Table 1 shows that both the adjusted price per kW and the estimated NPV/kW for Pilgrim and Oyster Creek were negative, the adjusted price per kW for Pilgrim and Clinton are similar, but for Three Mile Island 1 and Oyster Creek differ by factors of more than 10, and the adjusted price per kW for Fitzpatrick plus Indian Point 3 falls between the forecast NPVs of each of these units. To investigate the robustness of these forecasts, the next section discusses the probability distributions of the estimated NPVs. NPV Probability Distributions Of course, any estimate involves forecasting error. Fortunately, with econometrics the forecasting error can be calculated. Given the simple statistical structure of the forecasting equations, I can calculate the variance associated with the forecast of Capacity Factor, Fuel, O&M, and Capital Additions. It is equal to the variance of the plant or unit-specific constant plus the variance of the residual (Rothwell, 20004). Also, there is variance in electricity prices. To proxy this variance I use the variance in the average electricity revenue from residential customers for each state, available in EIA Form 826 for 1990-1997. With the forecasts of annual net revenue and these variances, I simulate the probability distribution of the NPV for each NPP unit using Monte Carlo techniques. This is done 1,000 times by sampling from the five probability distributions for electricity price, Capacity Factor, Fuel, O&M, and Capital Additions for each remaining year of the NPP licensed life, calculating the net revenue, and discounting the net revenue to the present. When discounting probabilistic outcomes, the recommended discount rate is the risk-free rate of interest. According to Brealey and Myers:6 “First, we should explain what is meant by a distribution of NPVs. The cash flows for each iteration of the simulation model are trans lated into a net present value by discounting at the risk-free rate. Why are they not dis counted at the opportunity cost of capital? Because, if you know what that is, you don’t need a simulation model, except perhaps to help forecast cash flows. The risk-free rate is used to avoid prejudging risk. The ‘expected NPV’ includes no allowance for risk. Risk is reflected in the dispersion of the NPV distribution.” Figures 1, 2, and 3 present NPV distributions for Indian Point 3, Fitzpatrick, and Fitzpatrick plus Indian Point 3 using a risk-free rate of 3 percent. The expected NPV at a discount rate of three percent for Indian Point 3 was $58/kW with a standard deviation (SD, the square root of variance) of $201/kW. In the 1,000 simulations, the NPV for Indian Point 3 was negative 37 percent of the time: P(NPV<0)=37 percent. On the other hand, the expected NPV for Fitzpatrick was $335/kW with a standard deviation of $175/kW. In the 1,000 simulations, the NPV for Fitzpatrick was negative in only 20 simulations (two percent of the time). With these plants operating together and facing the same random price of electricity: • The expected NPV is a weighted average of the expected values of each unit (where weights are equal to unit capacity divided by plant capacity). • The standard deviation is lower than the standard deviation for either unit, because randomly higher costs at one unit can be balanced by randomly lower costs at the other unit. • The probability of a negative NPV, nine percent, is closer to the value for Fitzpatrick, because of the lower standard deviation. Figure 1 – The NPV Probability Density for Indian Point 3 at a three percent Discount Rate. Figure 2 – The NPV Probability Density for Fitzpatrick at a three percent Discount Rate. This is the portfolio effect: portfolios of generating units reduce the risk associated with random cash flows from a single unit. A well-known result of modern finance theory is that for each level of return from a portfolio of assets, diversifying the portfolio can reduce risk. As nuclear power plant owners attempt to maximize profit in a deregulated environment, they will seek ways to diversify their portfolios. One way is to buy other nuclear power plants. Which plants should they buy? Figure 3 – The NPV Probability Density for Fitzpatrick plus Indian Point 3 at a three percent Discount Rate. Implications for Future Nuclear Power Plant Sales Table 2 presents the results of probability analysis described above applied to every nuclear power plant in the United States. Three-quarters (77/103) of all U.S. nuclear power plants have less than a 10 percent chance of having a negative NPV. This is true for nuclear power plants in states that have not passed deregulation legislation (32/43) and states that have passed deregulation legislation (45/60). Of the 12 units that have been sold, 60 percent (7/12) had a less than 10 percent chance of having a negative NPV, but the other 40 percent had a higher chance of having a negative NPV (two of them were Pilgrim and Oyster Creek). Sales of these poorly performing units depressed prices in the nuclear power plant purchasing market in 1998 and 1999. Although 12 operating units have been sold in states that have started to deregulate, there are 48 units that have not been sold. Most of them have less than a 10 percent chance of having a negative NPV, but eight of them have greater than a 50 percent chance of having a negative NPV. The owners of these units should either sell them or merge with better managed nuclear power plant owners. Tab. 2 Probability of Negative NPV Further, because of portfolio effects and because it appears to be difficult to receive full value on nuclear power plant sales, owners who can profitably manage one unit or one plant should either buy other units or merge with owners of well-managed plants. As more states deregulate and either require their utilities to sell generating assets or allow their utilities to merge, the nuclear power plant industry will become more concentrated. Although this might increase market power in some generating markets, it will surely increase the market power of the utilities in the nuclear services market. As the nuclear power industry becomes more profitable, the probability of a new nuclear power plant order will increase. But until prices of “used” nuclear power plants (until now about $500/kW) approach the cost of constructing and financing a new nuclear power plant (about $2000/kW), new orders are unlikely. Therefore, sales prices in the deregulated nuclear power plant market signal the future of the nuclear power industry. Acknowledgments I thank R. Hagen and E. O’Donnell and participants at my presentations to the Society for Risk Analysis Annual Meeting, Arlington, Virginia (December 3-8, 2000), the Western Economics Association Annual Meetings, San Francisco, California (July 4-8, 2001), and the Department of Nuclear Engineering, University of California, Berkeley, for helpful suggestions. Of course, any errors are my own. Footnotes 1 D. Donoghue, D. Haarmeyer, and R.T. McWinney, “The US Nuclear Plant Market: The Search for Value,” Nuclear News, June 2000. 2 T. Gomez, and G.S. Rothwell, “Principles of Electricity Economics: Regulation and Deregulation,” IEEE Press, forthcoming. 3 G.S. Rothwell, “Profitability Risk Assessment at Nuclear Power Plants under Electricity Deregulation,” The Utilities Project, sponsored by PricewaterhouseCoopers, www.UtilitiesProject.com, 2000. 4 G.S. Rothwell, “The Risk of Early Retirement of US Nuclear Power Plants under Electricity Deregulation and CO2 Emission Reductions.” The Energy Journal 21(3), 2000. 5 U.S. Energy Information Administration, “Competitive Electricity Prices: An Update,” Issues in Midterm Analysis and Forecasting, 1998. 6 R. Brealey and S. Myers, “Principles of Corporate Finance,” McGraw-Hill, 1981. Filed under: White Papers Tagged under: Utilities About the Author Chris Trayhorn, Publisher of mThink Blue Book Chris Trayhorn is the Chairman of the Performance Marketing Industry Blue Ribbon Panel and the CEO of mThink.com, a leading online and content marketing agency. He has founded four successful marketing companies in London and San Francisco in the last 15 years, and is currently the founder and publisher of Revenue+Performance magazine, the magazine of the performance marketing industry since 2002.