Because the standard approach studies use overnight costs as the dependent variable, identifying economies of scale correctly is a three-step process. The first step is to calculate what effects, if any, scaling up plant size has on overnight costs, “the quantities of land, labor and materials needed to construct a nuclear power plant” (Cantor and Hewlett 1988, p. 318). However, overnight costs, by definition, do not take into account time-related costs like interest and inflation. The second step is to determine how scaling-up plant size affects construction time, and the third step is to determine how an increase in construction time affects costs. Construction time affects real costs in two ways: an increase in construction time results in greater interest and inflation as well as greater overnight costs. Although the relationship between construction time and overnight costs is not immediately clear, Cantor and Hewlett’s evidence of such a relationship is described below. Table 4-4 lists the studies’ estimated scale effects. As a reference point, Komanoff wrote in 1981 that “capital cost projections of the U.S. Atomic Energy Commission (AEC), its successor agencies, and the power industry assume that nuclear costs per kW decline by 20 to 30 percent when reactor size is doubled” (Komanoff 1981, p. 200).
Table 4-4: Economies of Scale
Study
Scale Economies; Results of Doubling Unit Size
Scale Economies with IDC/Construction-
Time Effects
Mooz (1979) Insignificant
Paik and Schriver (1979) N/A; did not use cost per kW as dependent variable
Komanoff (1981) 13 percent reduction in cost per kW
With IDC effects, 10 percent cost reduction Zimmerman (1982) 11.1 percent reduction
In cost per kW
Cantor and Hewlett (1988) 36.2 percent reduction in cost per kW
With construction time- overnight cost relationship, 9 percent increase
McCabe (1996) Insignificant
Canterbery et al. (1996) 51 percent reduction in cost per kW
In Mooz’s study, the size coefficient in his regression was insignificant, implying no economies of scale. However, he did not take construction-time effects into account; doing so may have added costs and implied diseconomies of scale. Paik and Schriver (1979) used plant costs, not costs per kW costs, as the dependent variable, making testing for economies of scale impossible. Komanoff found that doubling reactor size led to a 13 percent reduction in costs per kW when overnight costs were used as the dependent variable. Komanoff recognized that construction-time effects had to be considered: “Doubling reactor size extended construction time by an average of 28 percent… the resulting increase in IDC adds approximately 3 percent to real costs, so that the net effect of doubled size is only a
10 percent cost reduction” (Komanoff 1981, p. 200). Komanoff’s results imply slight economies of scale; however, Cantor and Hewlett’s results (below) indicate that Komanoff should have considered the construction time-overnight cost relationship in addition to the construction time-interest relationship. Zimmerman found that doubling unit size resulted in an 11.1 percent reduction in costs per kW, but the size coefficient had “relatively large standard error” (Zimmerman 1982, p. 302). Moreover, Zimmerman did not take construction-time effects into account.
Cantor and Hewlett’s study departs from previous studies by accounting for the relationship between construction time and overnight costs. This relationship is not intuitive: overnight costs, by definition, contain no time-related costs. However, Cantor and Hewlett hypothesize that construction time might increase overnight costs in three ways.
Construction delays could result in morale problems, a stricter regulatory environment (since regulation increases with time), and increased hiring costs (since workers often must be laid
off and then re-hired). They find overnight costs correlated with construction time beyond the 95 percent significance level. Using just overnight costs and no construction-time effects, Cantor and Hewlett find that a doubling of unit size resulted in a 36 percent decrease in costs per kW. However, when the construction time-overnight cost relationship is taken into account, the longer construction time necessary to build larger plants makes a doubling in plant size 9 percent more expensive, in costs per kW. “Thus, there appears to be some evidence supporting the claims that the industry has attempted to build units that are too large to be efficiently managed by the constructors” (Cantor and Hewlett 1988, p. 318). The construction time-overnight cost relationship excludes the construction time-interest cost relationship which affects LCOE.
Marshall and Navarro (1991) deal with the measurement rather than the concept of overnight costs. They observe that the calculation of overnight costs as the sum of
expenditures over the construction period must be placed on a correct time basis to satisfy the requirements of capital theory. They point out that overnight costs as commonly calculated cannot accurately represent economies of scale because they leave out construction-time effects. The summation of construction expenditures must be dated to the completion date of the plant, which requires the proper correction for price-level change in the calculation of overnight costs. In their test for economies of scale in Japanese nuclear plants using
overnight costs dated to plant completion dates, they find that when overnight costs are used as the dependent variable, economies of scale seem to exist, but when opening-date overnight costs are used, the size coefficient is insignificant, implying no economies of scale exist (Marshall and Navarro 1991, p. 153). Although Marshall and Navarro are critical of the previous studies’ calculation of overnight costs, they come to the same conclusion – if economies of scale exist, they are minimal. McCabe (1996) uses Marshall and Navarro’s opening-date cost and finds the size coefficient insignificant (McCabe 1996, p. 370). Canterbery et al. (1996, p. 558, n. 4) note that Marshall and Navarro’s contention that the definition of overnight costs exclude inflation is incorrect.
The results on size effects are diverse. In Mooz’s regression, the size coefficient is insignificant. Paik and Schriver do not use costs per kW, so their results cannot be used to test for economies of scale. Komanoff finds evidence of slight economies of scale, taking into account construction time-interest effects but not construction time-overnight cost effects. Cantor and Hewlett find evidence of slight diseconomies of scale, taking into account construction time-overnight cost effects, but not construction time interest effects. Zimmerman finds evidence of modest economies of scale, but the coefficient is only marginally significant. Using opening-date overnight costs, both Marshall and Navarro (1991) and McCabe (1996) find no evidence of economies of scale, but Canterbery et al. (1996) find a strong size effect reducing cost. It seems reasonable to conclude that few if any scale economies existed in nuclear plant construction in the 1970s and 1980s to confound the identification of learning effects.
4.7.5. Region
All of the studies included some measure to control for region. Utilities building in different regions face different input costs and different inflation rates. Komanoff, Cantor and Hewlett, and McCabe used the Handy-Whitman index to control for regional differences in cost of living. Other studies do not mention such adjustments. All of the studies include a variable in their regressions to account for different input prices. Studies often note
construction labor as a major input that varies significantly in cost across regions.
Construction wage rates “are generally about 5 percent higher in the Northeast than in the Midwest and 25 percent higher than in the Southeast” (Komanoff 1981, p. 199). Another potential source of regional cost differences is difficulty finding a site. For example, in the Pacific region, builders have had to consider seismic potential, while developers in the densely-populated Northeast have had to find a site sufficiently far from urban centers.
Table 4-5 shows the results of the studies. The first four (Mooz through
Zimmerman), use regional dummy variables. These four studies found that plants built in the Northeast cost more than plants built elsewhere. Komanoff, for example, found that plants built in the Northeast were “28 percent more expensive, on average, than plants in other regions” (Komanoff 1981, p. 199). The last two studies, Cantor and Hewlett’s and McCabe’s, do not use regional dummies, but their regressions confirm that differences in construction labor costs account for a good deal of cost differences. Both Cantor and Hewlett and McCabe found that regional wage rate correlated with plant cost at 95 percent significance (Cantor and Hewlett 1988, p. 326; McCabe 1996, p. 371). Altogether, plants built in the Northeast were more expensive than plants built elsewhere, to a large extent a result of construction wage rates.
Table 4-5: Regional Cost Differences Study Number of Regions Controlled For Independent Variable Findings
Mooz (1979) 5 Dummy Northeast 38.8 percent more expensive than other regions Paik and
Schriver (1979)
4 Dummy South 21.3 percent cheaper, Mountain 23 percent cheaper than Northeast and Central Komanoff
(1981)
3 Dummy Northeast 28 percent more expensive than other regions Zimmerman
(1982)
5 Dummy Midwest 28.1 percent, South 30.2 percent, Mountain/Texas 39.3 percent, Pacific 14.8 percent less expensive than Northeast Cantor and
Hewlett (1988)
50 Construction labor wage rate in state where plant is located
Labor wage rate by state correlated with plant costs per kW at 95 percent
significance McCabe (1996) 4 Average union
wage rate in BLS region
Average wage rate by region correlated with costs per kW plant at 95 percent
significance