【正文】 Huettner and Landon (hereafter H and L) make an important theoretic contribution to the literature in recognizing that researchers, in addressing scale (and performance) issues with electric utilities, must deal with either one of two potential speci?cation issues. For example, if a researcher chooses to use the typical translog cost function for estimating scale and performance effects, restrictions on the production function and elasticities of substitution among inputs are minimized but at a cost which does not control for the important structuraldifference among ?rms. On the other hand, if a researcher were to use a more structurespeci?c cost function, differences between ?rms are controlled, but speci?cation errors may arise on the elasticity of substitution among inputs. H and L argue that the structural based approach to modelling utility performance and scale issues are more appropriate for the utility industry given its peak load, obligation to serve requirement. H and L construct an empirical model based upon 1971 utility industry data for each sector of the The transmission sector model runs transmission O and M costs against a number of structural industry characteristics including total capacity, wage costs, underground circuit miles, overhead structural miles, mercial and industrial sales, regional dummy variables, and a number of relatively ‘a(chǎn)d hoc’ dummy variables for certain utility holding panies. Results from the H and L model do not support the economies of scale argument for transmission and the authors posit that the industry must have a Ushaped average cost curve. The only variables which have any statistical signi?cance at traditionally accepted levels include overhead structure miles and one regional dummy variable. The model explained only 43 per cent of the variation in transmission costs. There are a number of problems associated with the H and L model, most of which are speci?cation oriented. First, the authors mix measurements associated with transmission lines. One variable, for instance, has overhead transmission miles measured in structure miles, while the other has underground transmission measured in circuit Second, the authors use the percentage of total generating capacity as a measure of transmission capacity. We believe this is a serious error associated with a transmission cost model. Utility transmission systems must be operated in a manner to support net power transfers and wheeling transactions, which could extend transmission capacity needs well beyond any system’s internal generation capacity. Thus, a generation capacity measure is an inaccurate indicator of transmission system capacity. Gilsdorf (1994) estimates a multiproduct translog cost function, and therefore indirectly single product cost functions for electric utility operations, including transmission. The primary purpose of this research was to analyse the effect of vertical integration on electric utility cost structures. Economies of integration were not supported. However, the important conclusion of this work is the presence of economies of scale in the provision of transmission services. Pollitt (1994, 1995) presents another recent estimate of transmission costs. His model takes the H and L approach of developing a structural, rather than translog, empirical speci?cation. The independent variables used by Pollitt differ somewhat from those presented by H and L. First, Pollitt excludes regional indicator variables. Second, and more importantly, Pollitt uses the product of circuit miles and voltage levels as a measure of transmission capacity. He also includes transformer capacity as a variable. In addition, Pollitt includes an indicator variable representing transmission ownership characteristics (public versus private). The empirical results from the Pollitt investigations support the notion of economies of scale in transmission service. Capacity (circuit miles * voltage) is found todecrease transmission costs signi?cantly, while increased underground transmission lines and wage costs have signi?cant positive impacts on average transmission costs. None of the remaining variables were of any statistical signi?cance. The model explained only 21 per cent of the variation in transmission Pollitt’s model corrects a number of the problems associated with the earlier H and L work. First, overhead and underground transmission lines are represented in mon terms (circuit miles). Second, Pollitt recognizes that generation capacity is a very poor proxy for capacity in the transmission system. Pollitt attempts to remedy this problem by including a transmission capacity variable de?ned as the product of transmission circuit miles and voltage. The construction and use of this variable, in addition to the use of total system transformer capacity, raises a number of questions regarding the strength of its results. From a technical perspective, an aggregated voltage level/circuit miles product (v *cm) of transmission lines is not a good measure of transmission system capacity. First, distance has nothing to do with capacity. Second, transmission system design begins with sizing a transformer to link a particular load and source of power given some voltage level. Transformers convert usable and unusable power between different circuits。 typically circuits of different voltages. Once appropriate voltage level and transformer size (in megavoltamperes or MVA) is determined, circuits and conductor sizes are determined in order to connect the load (or generation) to the In a modern utility system, energy from a generator may under go four or ?ve different transformations between generator and end user. Thus, any transmission system is likely to have several times more MVA of transformer capacity than generation capacity (General Electric Company, 1983, p. 146). These transformers serve to connect different circuits and can be viewed as a facilitator, or bottleneck, fo