# Vertical integration efficiencies and electric utilities: a cost complementarity perspective.

During the 1980's, public attitudes and policy toward regulation changed dramatically, emphasizing greater reliance on market forces in place of government intervention. Substantial deregulation has occurred in once heavily regulated industries as natural gas, telecommunications, trucking, and airlines. Pressure for deregulation in the electric utility industry has also been increasing recently as evidenced by the Federal Energy Regulatory Commission's (FERC) proposed rulemaking, Regulations Governing Independent Power Producers (IPPs), which would reduce the degree of Federal regulatory oversight on wholesale generation transactions. (FERC 1988). The FERC developed the proposed rule based on the belief that electric generation is not a natural monopoly in many markets. According to FERC, recent empirical studies raise doubts about the natural monopoly assumption, but even where scale economies at the plant level are significant, technological advancements in transmission have expanded the market size sufficiently to allow a relatively large number of competing sellers to participate. (FERC 1988, pp. 23-25).According to the proposal, rates negotiated between sellers that "lack significant market power," called independent power producers (IPP's), and utilities would be presumed just and reasonable. In order to qualify as an IPP, the seller cannot provide power to a wholesale customer located within a retail franchise service territory possessed by the seller or its affiliates nor be served by essential transmission facilities controlled by the seller or its affiliates.

The FERC tentatively defines essential transmission facilities as those facilities which supply 50 percent or more of the customer's electric needs. The commission defends the 50 percent threshold by arguing that the customer would have access to at least one other alternative supplier at least as large as the IPP. (FERC 1988, pp. 39-40).

These provisions are designed to protect consumers against abusive arrangements between utilities and their affiliates. The concern over self-dealing and potential monopoly behavior arises from the industry's vertically integrated structure and the belief that transmission and distribution stages are natural monopolies due to economies of scale (Weiss 1975; Joskow and Schmalensee 1983). Utilities may be able to circumvent rate-of-return regulation by transferring their monopoly power to the unregulated generation affiliate. Concerns over self-dealing have caused some deregulation advocates to argue for common carriage status in transmission and, possibly, distribution. This position implicitly assumes that cost efficiencies from integration are negligible or non-existent. If cost savings are important, however, Joskow and Schmalensee (1983) state that, "The economies of vertical integration link the secondary and primary markets together, and single firm production of the primary product by a vertically integrated firm is the least cost outcome." (p. 30). Sufficient economics of integration make both stages a natural monopoly even though one stage is not characterized by economies of scale.

This article investigates whether economies of integration exist between generation, transmission and distribution within a multiproduct cost framework. The multiproduct approach is based on the idea that each stage produces an output that, when bundled together, results in electric sales service. From this perspective, economies of vertical integration exist if the three stage mix exhibits global economics of scope.

To determine if the electric industry is characterized by scope economies, the study estimates a multiproduct translog cost function (TMCF). The estimates provide information about cost complementarities within the three stage set. Cost complementarities among all three stages would imply economies of scope and, thus, vertical integration efficiencies.

The study's findings provide no evidence of cost complementarity, but do show increasing product-specific scale economies for each stage at the sample mean. These findings suggest that: (1) continued regulation of transmission and distribution is appropriate; and (2) complete deregulation of generation is premature without substantial alteration in transmission regulation. Product-specific scale economies in generation would necessitate an expansion of generation markets beyond a utility's retail franchise service territory. Given utility ownership of transmission facilities, regulatory policy must ensure non-discriminatory access to the transmission grid for all non-affiliated sellers.

The study begins with a review of previous work on vertical integration in the electric industry. The next section presents the model used to test for cost complementarity, including definitions of variables and specification of the cost function. The subsequent section discusses the empirical results obtained from the estimated model followed by a discussion regarding their regulatory policy implications.

REVIEW OF THE LITERATURE ON VERTICAL INTEGRATION IN THE ELECTRIC INDUSTRY

Virtually all utilities are engaged in the three phases of electric supply. For a sample of 105 private electric utilities, the average ratio of own-generation to total supply (i.e. firm purchases plus own-generation) was about 94 percent in 1985.(1) On the distribution side, sales to ultimate customers represent, on average, 88 percent of total sales for the same sample. These figures suggest a high degree of backward and forward integration.

Landon (1983) discusses various cost efficiency-related incentives for vertical integration and applies them to the electric power industry. He maintains that the different stages share a high degree of technological interdependency, requiring advance knowledge about the design plans for all stages.

Landon also discusses the relationship between information needs, uncertainty, transaction-specific investments, and transaction cost considerations. Proper planning and operation necessitates a free flow of information between stages because the interdependent nature of the production process requires planning input from all other stages at the same time. Centralized decision-making would internalize these informational externalities.

Moreover, the nature of the investments imply long-term contracts. Yet, Landon argues that uncertainty over potential changes in economic conditions, environmental standards, tax laws, technologies, and the availability and relative cost of resources increases with the length of the contract. The greater the uncertainty, the harder it becomes to specify important contract terms. Asset specificity, on the other hand, implies that the asset owner has few alternatives once the investment is made, locking the owner into a small numbers bargaining relationship. Given the high degrees of uncertainty, asset specificity, technological interdependency and complexity, internal organization may represent the least costly means of coordinating the three stages.

Joskow and Schmalensee (1983) also discuss the nature of electric investments within the transaction cost framework. They agree that the industry is characterized by uncertainty, complexity, asset specificity, and infrequency of recurring transactions. Substantial financial commitments coupled with a useful life of forty to fifty years create considerable uncertainty over cost recovery, future operating costs, demand, and regulatory requirements. In addition, generation, transmission, and distribution plant investments are sunk costs and have site-specific characteristics. (p. 123). These characteristics are associated with high appropriable rents and, absent integration, entail long-term contracts to carry out most business activity.

Several other studies have attempted to directly incorporate the effect of vertical integration on utility cost structure. Henderson (1985) and Roberts (1986) conducted cost studies in which they conclude that transmission-distribution systems can not be analyzed independent of generation. However, the underlying methodology in both studies has a very important shortcoming: it fails to account for any possible effect that transmission or distribution may have on generation costs. The models only examine how generation affects the transmission-distribution stages and do not address whether significant cost economies arise from vertical integration. Eftekhari (1989), however, does test for economies of scope between electric generation, transmission, and distribution. His results indicate global diseconomies of scope for all three stages, although he does find stage-specific economies of scope between generation and transmission.

Eftekhari's methodology improves upon Henderson's and Roberts' approaches by explicitly allowing for symmetrical interactions among all stages, but the study has several shortcomings. First, the output measure for transmission, an index of interconnected activity, incorporates only interchanges and wheeling activity and ignores the significant amount of power transmitted on behalf of the firm's own system sales. This causes his measure to substantially underestimate the actual output of the transmission function. Second, his study does not include a distribution density variable, although density could significantly affect utility distribution costs.

Kaserman and Mayo (K-M) (1991) also attempted to directly estimate the degree of integration economies for electric utilities. They estimate a multistage cost function that is quadratic in the outputs and separable from input prices and other hedonic variables. Moreover, the input price and hedonic variable functions are linear, noninteractive expressions. Roller (1990a; 1990b) and K-M argue that a quadratic function provides more robust estimates concerning scope economies than the translog function.

In the K-M methodology, the electric process involves two stages-generation and distribution. The authors treat transmission and distribution as one stage because of their similar operating characteristics (p. 483). Their results show cost complementarities between the two stages and that integration economies exist over much of the sample.

However, an important limitation to the K-M function is that it assumes marginal costs are independent of input prices and other hedonic characteristics. In addition, no mention is made regarding satisfaction of the linear homogeneity requirement of a well-behaved costfunction. Thus, their function loses much of the flexibility inherent in the translog function and, thus, may provide inaccurate estimates of cost coefficients.

The next section discusses the methodology to be used in this analysis. It follows an approach similar to Eftekhari, but includes: (1) a broader, more accurate measure of transmission service; and (2) a distribution density variable.

METHODOLOGY OF COST STUDY

To test for economies of integration, the analysis adopts a multiproduct cost function framework. The multiproduct cost function can indicate if, and the extent, in-house production of generation, transmission, and distribution services is less costly than independent firm specialization.

Definition of Variables

Variables adopted in this study include three outputs, three input prices, and a measure of distribution density. Each is discussed more fully below.

Each stage (i.e., generation, transmission, and distribution) produces a service that culminates in the delivery of electric power. Megawatthours (MWH's) generated (G) by fossil-fuel fired steam plants represent the output of the generation function. The study focuses on fossil-fuel generation because nuclear and hydro-electric production technologies differ substantially from conventional steam production. (Nerlove 1963; Christensen and Greene 1976). Moreover, fossil fuel generation is the dominant generation type, accounting for approximately 70 percent of total utility generation.

The transmission function's basic purpose is to connect generators to load centers. The path provided by the transmission function represents the means by which that purpose is fulfilled. Electric power flows through a set of conductors called a circuit, which represents the route or path electricity must follow. The quantity of power flowing through a circuit depends on both voltage and current levels. Increases in either voltage or current results in more power flow. In addition, voltage levels and their variation depend on the generators' speed. Therefore, the actual flow of electric power is not independent of generation, but rather, represents a "bundled" commodity composed of outputs from both the generation and transmission functions. Transmission output (T) is simply the path of the transmission system, which is a function of circuit mileage and voltage capacity. The U.S. Department of Energy publishes data on the total number of circuit-miles of transmission lines within the following six voltage ranges:

(1) 132,000-143,000;

(2) 144,000-188,000;

(3) 189,000-253,000;

(4) 254,000-400,000;

(5) 401,000-600,000; and

(6) 601,000-850,000.(2)

The study measures transmission output, called circuit-voltage miles (CVM's), by taking the product of each voltage range's midpoint and its corresponding number of circuit-miles. In effect, transmission service is "packaged" in CVM's.

The distribution system connects the utility's final customers to the electrical system. Neuberg (1977) identifies several important activities which underlie the distribution function, including load dispatching, customer installation, meter reading, advertising, and administration. They comprise what may be called the merchant function for retail electric sales. Implicit in this merchant function is a brokerage function where the utility ensures an adequate supply of electric power through negotiating and securing contracts with other generation companies.

Given this conceptual definition of distribution, distribution costs should depend directly on the number of ultimate customers (UC) served. Neuberg (1977) also includes kilowatt-hours (kwh) sold as another factor affecting distribution costs. Retail kwh sold, however, represents a "bundled" commodity composed of outputs from all three production stages, implying that generation and transmission also influence distribution costs. The multiproduct cost function adopted in this study recognizes the potential for this interdependency by allowing for interaction between all three stages.

In addition, the distribution of ultimate customers within a service territory is often thought to have an important effect on cost structure. It is commonly observed that substantial economies of density characterize the distribution function. Thus, utilities with different degrees of density may exhibit systematic differences in cost structure.

A usual measure of density (X) is customers per square mile of service territory. However, this study uses the number of distribution meters per square mile of service territory because it more accurately recognizes differences in customer makeup, since some customers may have more than one distribution meter. The higher the ratio, the greater the distribution density.

The cost study includes three input prices--the prices of labor ([P.sub.L]), fuel ([P.sub.F]), and capital services ([P.sub.C]). The price of labor is calculated by summing total salaries, wages, and employee pensions and benefits and dividing by the number of full-time employees plus one-half of the part-time employees. Fuel price equals fuel expenditures divided by the MMBTU equivalent of the fuel sources. The capital service price variable ([P.sub.C]) is a Divisia price index constructed from individual capital service prices calculated for each asset class (i.e., generation, transmission, distribution, and general plant). The study adopts the Divisia price index for two major reasons. First, each utility hires capital inputs that differ substantially by production stage and the input price variable must reflect this heterogeneity. Second, adopting one capital service price substantially reduces the number of parameters to estimate.

In calculating the capital service Divisia price index ([P.sub.C]), a separate capital service price is calculated for each asset class according to the following formula:

(1) [P.sub.a] = [(1-UZ-K) / (1-U)] (*) [[q.sub.a]r + [q.sub.a][d.sub.a]]

where:

a = the [a.sup.th] asset class and

[P.sub.a] = implicit rental service price for asset class a;

[q.sub.a] = acquisition cost of a unit of stock in class a;

r = weighted average cost of capital;

[d.sub.a] = rate of depreciation;

U = marginal corporate tax rate;

Z = present value of depreciation per dollar of investment; and

K = investment tax credit rate. (Christensen, Gollop, and Stevenson 1980).

The formula in Equation 1 is based on the principle that the market value of an asset equals the discounted value of its capital services. The implicit capital service price depends on the price of the asset, rate of replacement, and cost of capital.

The Handy-Whitman Index of Public Utility Construction Costs provides the acquisition cost measure ([q.sub.a]) for generation, transmission, and distribution assets. The Producer Price Index for capital equipment is used for general plant.(3) The statutory marginal federal corporate tax rate is adopted for (U). Calculation of the present value of depreciation per dollar of investment (Z) assumes utilities depreciate equipment according to property classified as 15 year public utility property.(4) The depreciation rates are provided in statute. The investment tax credit is set at 10 percent. The rate of depreciation ([d.sub.a]) is based on rates developed by Christensen, Gollop, and Stevenson (1980). The depreciation rate is calculated using the 1.5 declining balance method:

(2) [d.sub.a] = 1.5/[L.sub.a]

where: [L.sub.a] = the asset's predicted useful service life.

Service lives for each asset class are as follows:

1. steam production plant - 33 years

2. transmission plant - 37 years

3. distribution plant - 37years

4. general plant - 25 years.

Finally, the weighted average cost of capital (r) is calculated according to the conventional formula. The yield on public utility bonds and preferred stock by rating represent the cost of long-term debt and preferred stock.(5) A discounted cash flow model is used to estimate the cost of common equity.(6)

The cost of long-term debt, preferred stock and common equity are weighted according to their relative share of the firm's total capitalization. Total capitalization equals the sum of the market values of debt, preferred stock, and common equity. Common equity is valued at its market price. The market value of long-term debt and preferred stock is derived by dividing long-term debt interest payments and preferred dividends by the yield on public utility bonds and preferred stock, respectively. (Henderson 1985).

The company's total cost (C) is the sum of operation and maintenance expenses and imputed capital expenditures, less the costs of purchased power.(7) Capital expenditures are obtained by multiplying the imputed capital service price per asset class ([P.sub.a]) by the corresponding estimated level of capital stock. The capital stock measure is derived by using a "triangularized" weighted average of the Handy-Whitman Index:

(3) [KS.sub.a] =S/[summation] t (([Y.sub.t]/210)(*)[PI.sub.at])

where:

[KS.sub.a] = derived capital stock for asset class a;

S = book value of plant in service in 1985;

[Y.subt] = the [t.sup.th] year from the beginning of 1966 through 1985;

[PI.sub.at] = Price Index applicable to asset class "a" in the [t.sup.th] year.(8)

Data on operation and maintenance expenses, book value of electric plant, circuit miles, voltage levels, number of distribution meters, and number of employees are provided in the U.S. Department of Energy publication, Financial Statistics of Privately Owned Electric Utilities In the United States, 1985. Standard and Poor's (1987) Compustat Services Inc. contained data on fuel prices, number of ultimate customers, miles of service territory, megawatthours generated, annualized dividend yields, and other stock market information used to calculate the cost of common equity.(9) Finally, Moody's Public Utility Manual, 1986 includes information on public utility bond and preferred stock yields and bond and preferred stock ratings for each utility.

Sample Design

The study includes seventy-two (72) privately owned electric utilities for the year 1985, using the holding company as the appropriate unit of observation. Several criteria were used to define the final sample. First, the sample includes only those utilities reported in both Financial Statistics of Privately Owned Electric Utilities In the United States, 1985 and Standard and Poor's Compustat Services Inc. Second, the utilities must be engaged in all three stages of production because the cost function employed in the analysis does not allow for zero values. Third, recognizing the significant differences between generation technologies, each utility must generate at least 65 percent of their electricity from non-nuclear steam processes. Conventional steam generation is the dominant technology, accounting for 73 percent of total electric utility generation in 1985, with investor-owned utilities producing approximately 83 percent of the conventional steam output. (Edison Electric Institute 1986, pp. 20, 23). The seventy-two sample utilities accounted for over 72 percent of the total conventional steam production by investor-owned utilities.(10)

Specification of Cost Function

A multiproduct perspective is used here to investigate efficiencies in vertical integration. Two important assumptions underlie direct estimation of a cost function: (1) exogenous output levels and input prices and (2) cost minimizing behavior. Electric utilities are required by regulators to meet all demand at regulated rates. Thus, output levels are exogenous to the firm. Competitive input markets are commonly assumed in the empirical literature and this assumption will be maintained here.(11)

Cost minimization implies no Averch-Johnson (A-J) effect. Joskow and Noll (1981) discuss several conceptual problems associated with the A-J model and review studies which find no evidence of an A-J effect. I assume cost minimizing behavior as do Christensen and Greene (1976), Stevenson (1980), Henderson (1985), Roberts (1986), Sing (1987), and Eftekhari (1989).

The cost function for a utility providing generation, transmission, and distribution services can be expressed as follows:

(4) C = f(G, CVM, UC, [P.sub.L], [P.sub.F], [P.sub.C], X).

The Translog Multiproduct Cost Function (TMCF) has been selected for this study. (Brown, Caves, and Christensen 1979). The translog function is a quadratic function with its elements expressed in terms of their natural logarithm and imposes no constraints on: (1) the first or second derivatives of the cost function; (2) the input elasticities of substitution; nor (3) separability between outputs and input prices.

The TMCF for privately owned electric utilities with m outputs, n inputs, and a technological variable (distribution density (X)) can be expressed as follows:

(5) [Mathematical Expression Omitted]

Since Equation 4 indicates that m = 3 and n = 3, the cost function contains thirty-six parameters but, to ensure linear homogeneity in factor prices, the following restrictions must be imposed:

(6) [Mathematical Expression Omitted]

These restrictions reduce the number of free parameters to twenty-eight. One important limitation of the TMCF is that concavity and monotonicity conditions are not global. (Chambers 1988, pp. 178-179). Here we shall consider the translog function as a local approximation around the sample mean, implying that the result should only be interpreted locally (p. 180).

Furthermore, since concavity and monotonicity cannot be ensured at every observation, concerns regarding the accuracy of the translog function may arise when violations occur. Wales (1977) conducted a simulation study testing the accuracy of flexible forms when certain violations exist. His results indicate that the translog function provides good estimates even with a small number of violations.

Economies of vertical integration can be tested using the TMCF by examining whether or not the cost function exhibits economies of scope. Direct calculation of scope economics is impossible within the TMCF, however, since the cost function does not allow zero values for any of its variables. Nevertheless, economies of scope can be indirectly tested by examining whether or not cost complementarities exist. Baumol, Panzar, and Willig (B-P-W) (1982) define weak cost complementarity as follows:

"A twice-differentiable multiproduct cost function exhibits weak cost complementarities

over product set N, up to y, if [Mathematical Expression Omitted!, for all

y with [Mathematical Expression Omitted], with the inequality holding strictly over a set of nonzero

measure." (p. 74).

Weak cost complementarity implies that the marginal cost of producing one good decreases when production of the other goods within the product set N increases, which implies economies of scope at a given output [gamma]. (Baumol, Panzar, and Willig 1982, p. 75). Cost complementarity in the study's three output case exists when the following conditions hold:

(7) [[differential].sup.2]C/[differential]CVM[differential]G < 0 [[differential].sup.2]C/[differential]UC[differential]G < 0; [[differential].sup.2]C/[differential]UC[differential]CVM < 0.

Equation 7 can be derived within the TMCF as follows. First, let [Y.sub.i] measure the quantity of output i. Then, from Equation 5:

(8) [differential]1nC/[differential]1n[Y.sub.i] = ([differential]C/[differential][Y.sub.i])([Y.sub.i]/C). Thus:

(9) [differential]C/[differential][Y.sub.i] = (C/[Y.sub.i])([differential]1nC/[differential]1n[Y.sub.i]).

Taking the derivative of Equation 9 with respect to [Y.sub.j] yields:

(10) [[differential].sup.2]C/[differential][Y.sub.j][differential][Y.s ub.i! = [1/[Y.sub.i] ([differential]C/[differential][Y.sub.j] [differential]1nC/[differential]1n[Y.sub.i]) + ([differential]/[Y.sub.j] C/[Y.sub.i] [[differential].sup.2]1nC/[different

= [1/[Y.sub.i] (([differential]1nC/[differential]1n[Y.sub.j] C/[Y.sub.j]) [differential]1nC/[differential]1n[Y.sub.i]) + (1/[Y.sub.j] C/[Y.sub.i] [[differential].sup.2]1nC/[differential]1n[Y.sub.j][differential] 1n[Y.sub.i])]

= C/[Y.sub.i][Y.sub.j] [([differential]1nC/[differential]1n[Y.sub.j] [differential]1nC/[differential]1n[Y.sub.i]) + [differential]1nC/[differential]1n[Y.sub.j][differential][Y.sub.i]] Within the TMCF framework:

(11) [Mathematical Expression Omitted] and:

(12) [differential]1nC/[differential]1n[Y.sub.j][differential]1n[Y.sub .i! = [[beta].sub.ij] At the point where the normalized values for the three output variables and PL, PF, PC, and X equal unity, Equation 11 reduces to [[beta].sub.i]. Thus:

(13) [differential]1nC/[differential]1n[Y.sub.j] x [differential]1nC/[differential]1n[Y.sub.i] = [[beta].sub.j] x [[beta].sub.i] at that point. Substituting Equation 13 into Equation 10 implies that {([[beta].sub.j] x [[beta].sub.i]) + [[beta].sub.ij]} < 0 if cost complementarities exist. Therefore, cost complementarity implies that [[beta].sub.ij] < 0 and its absolute value is greater than the product of [[beta].sub.i] and [[beta].sub.j] Cost complementarity conditions for the three partial derivatives at the point of approximation are as follows:

(14a) Generation - Transmission [right arrow] ([[beta].sub.G] * [[beta].sub.GCVM] < 0;

(14b) Generation - Distributon [right arrow ([[beta].sub.G]* [[beta].sub.UC]) + [[beta].sub.GUC] < 0; and

(14c) Transmission - Distribution [right arrow] (beta]CVM* [[beta].sub.UC]) + [[beta].sub.CVMUC] < 0.

It must be noted, however, that economies of scope may still exist for product set N even in the absence of cost complementarity. For example, economies of scope may still arise due to a shared fixed cost among the three outputs. Thus, this study's cost complementarity test is not an exhaustive analysis for economies of scope, which is an important limitation of the TMCF framework (Fuss and Waverman 1981).

Estimation Procedure

The estimation procedure adopted in this study follows that used by Christensen and Greene (1976). They jointly estimated the main translog cost function with the corresponding input cost share equations using Zellner's (1962) iterative estimating procedure for seemingly unrelated regressions. Furthermore, the density, output, and input price variables are scaled such that, at their sample means, they equal unity. The input cost share equations are derived from Shephard's Lemma where the derived demand for some input j ([X(*).sub.j]) is the partial derivative of the total cost function with respect to input price:

(15) [X(*).sub.j] = [differential]C/[differential][P.sub.j]. Applying Shephard's Lemma to the TMCF yields:

(16) [differential]1nC/[differential]1n[P.sub.j] = [differential]C/[differential][P.sub.j] x [P.sub.j]/C. Since [differential]C/[differential][P.sub.j] = Xj, then Equation 16 can be written as:

(17) [differential]1nC/[differential]1n[P.sub.j] = ([X(*).sub.j] x [P.sub.j])/C = [S.sub.j], where [S.sub.j] represents the proportion of total cost attributable to the jth input. For total cost Equation 5, the corresponding input cost share equation for input is:

(18) [Mathematical Expression Omitted]

Kmenta and Gilbert (1968) indicate that iteration of the system of equations yields maximum likelihood estimates. In order to make estimation of the system possible, however, the labor cost share equation is dropped. Barten (1969) has shown that maximum likelihood estimates are not affected by which cost share equation is deleted.

EMIPIRICAL RESULTS

Table 1 contains cost function parameter estimates for electric utilities, including a goodness of fit measure for each equation. The predicted values for all cost share equationswere positive over all observations and inputprice concavity conditions were violated in sixteen cases, not including the point of approximation.

Table 1. PARAMETER ESTIMATES FOR TRANSLOG MULTIPRODUCT COST FUNCTION Variable Coefficient "T" Ratio Intercept 20.923 610.72(*) Wage .18454 30.617(*) Fuel .28014 34.737(*) Pkap .53532 55.27(*) G .48076 7.7707(*) CVM .10736 2.997(*) UC .40669 5.5887(*) DN .026718 1.2095 SQWage .02522 1.3851 SQFuel .16710 15.915(*) SQPkap .12907 5.1086(*) Wage-Fuel -.031623 3.2788(*) Wage-Pkap .0064027 .3337 Fuel-Pkap -.13548 -10.427(*) SQG .091863 .97975 SQCVM .014032 .67548 SQUC -.077517 -.50291 SQDN .015016 1.0851 GCVM -.030452 -.57391 GUC -.014057 -.13251 CVMUC .030891 .47508 GDN -.05854 2.1926(*) CVMDN .012667 .47508 UCDN .04957 1.4135 G-Wage -.040145 4.8664(*) G-Fuel .098713 8.9079(*) G-Pkap -.058568 4.4258(*) CVM-Wage .0099 -1.9281((*)(*)) CVM-Fuel -.016825 -2.516(*) CVM-Pkap .026725 3.2631(*) UC-Wage .058862 5.5066(*) UC-Fuel -.071558 5.1055(*) UC-Pkap .012696 .74025 DN-Wage .00042575 .12926 DN-Fuel -.00049921 -.11157 DN-Pkap .000073454 .013793 Notes:(*) signficant at .05 critical level (**) sigificant at .1 critical level Equation [R.sup.2] Cost Function .98 Capital Share .69 Fuel Share .77

Cost Complementarity

Table 2 provides cost complementarity estimates based on the expressions in Equation 7 using the formula from Equations 14a-c and their associated standard errors.(12) The point estimates for each output combination are positive, indicating that marginal cost in each stage increases as production expands in the other stages. Although point estimates are positive for all combinations, only the Generation-Distribution is statistically significant at a. 1 critical level.(13) Since Equation 7 implies that all three estimates must be negative, however, the findings provide evidence against cost complementarity for the three good set.

[TABULAR DATA 2 OMITTED]

Product-Specific Economies of Scale

Product-specific economies of scale for product (i) in output vector (Y) is given by the following expression:

(19) Si (Y) = ICi (Y)/[Y.sub.i]Ci where:

Si = coefficient for product-specific scale economies;

Ici = incremental cost of product [Y.sub.i]; and

Ci = [differential]C/[differential][Y.sub.i]. (Baumol, Panzar, and Willig 1982, p. 68).

The incremental cost of product (i) in output set Nis:

(20) ICi(Y) = C(Y) - C(YN-i) where YN-i is an output vector with positive output quantities for all goods except [Y.sub.i]. (Baumol, Panzar, and Willig 1982, p. 67). Average incremental cost for product i (AICi) is defined as:

(21) AIC1(Y) = ICi(Y)/[Y.sub.i] Substituting Equation 21 into Equation 19 yields:

(22) Si(Y) = AICi(Y)/Ci.

Product-specific returns to scale are increasing when Si(Y) > 1. Now, if Cii(Y) < 0, marginal costs (Ci) must be decreasing up to Y, implying that incremental costs are increasing at a decreasing rate. Thus, average incremental costs at Y must exceed marginal cost at that output level.

The equation for Cii derived from the TMCF is written as follows:

(23) [Mathematical Expression Omitted] Thus, Cii < 0 if the bracketed expression in Equation 23 is negative. At the point of approximation, marginal costs are decreasing if the following condition holds:

(24) [[beta].sub.ii] + [[beta].sub.i]([[beta].sub.i] - 1)] < 0 Table 3 presents estimates of Equation 24 and associated standard errors for generation, transmission, and distribution.(14) The estimates are all statistically significant at a .05 critical level, indicating that marginal costs are decreasing at the point of approximation for generation, transmission, and distribution, respectively.(15)

[TABULAR DATA 3 OMITTED]

REGULATORY POLICY IMPLICATIONS

The study's findings impact two important regulatory issues. First, some degree of regulation is still required in the industry. Second, wholesale power markets need to be expanded to promote competition in electric generation.

Degree of Regulation Required

The central issue in the deregulation debate involves whether electricity production is a natural monopoly. If integrated utilities are natural monopolies, this implies that the multi-stage cost function is subadditive over the relevant market area. A necessary condition for multi-stage natural monopoly is economies of scope between generation, transmission, and distribution. A sufficient condition for subadditivity would be stage-specific scale economies combined with economies of scope. (Baumol, Panzar, and Willig 1982, p. 176). Thus, cost complementarities and decreasing marginal costs for each stage would suggest subadditivity at the point of approximation.

Nevertheless, natural monopoly may not necessarily result in allocative inefficiency if the market is contestable. Contestable markets have two importaht characteristics: (1) potential entrants incur no costs greater than the incumbent firms; and (2) potential entrants base their entry decision on pre-entry prices. (Baumol et al. 1982, p. 5). The critical factor is that there be no sunk costs; that is, entry must be costlessly reversible. If electric markets are contestable, rate-of-return regulation is unnecessary to correct for allocative inefficiencies associated vdth pure monopoly.

The study's findings indicate stage-specific scale economies in all three stages, but provide no evidence of cost complementarity and, thus, subadditivity for integrated utilities. However, electric markets do not match the characteristics associated with contestability. Landon (1983) and Joskow and Schmalensee (1983) discuss the site-specific and irreversible nature of electric plant investments. In addition, capital costs account for a significant share of the total costs within each stage.(16)

The existence of stage-specific scale economies and high sunk costs suggest that complete deregulation of any phase in electricity production is premature. Transmission and distribution facilities are essential in the provision of electric service since no alternative technology exists which can move electric power from the generator to the final consumer. Regulatory agencies should continue to regulate rates, entry, and access conditions for transmission and distribution services.

Electric generation, on the other hand, has received the greatest attention by deregulation advocates. With increasing returns to scale for the "average" utility, defining average to mean the point of approximation, and assuming the generation market is the utility's franchise service territory, monopoly problems could arise. If competition is possible at all in electric generation, generation markets must be enlarged beyond the utility's service territory.

Expansion of Wholesale Generation Markets

Expanding wholesale market areas may enable a sufficiently large number of firms to enter and compete. Increasing market size, however, means that generation firms must obtain access to transmission and, ultimately, distribution facilities. The ability to wheel depends on transmission capacity constraints and the utility's willingness to transmit power, however. Schmalensee and Golub (1984) indicate that little information is available about existing capacity constraints. Thus, the potential geographical extent of generation markets and the resultant level of concentration is highly uncertain.

The utility's willingness to wheel power is also very important. Vertical integration is extensive and the potential for anticompetitive conduct exists. Utilities would have incentives to prevent third-party generators from obtaining access to essential distribution or transmission facilities. Regulated utilities could, theoretically, evade rate-of-return regulation through integration into the deregulated generation sector. The utility would purchase its own generation at the "monopoly" price and pass it through retail rates.

It may be argued that statere gulators could prevent pass-through of excessive costs. However, state jurisdiction does not extend directly to wholesale purchases. Moreover, regulators must show the utility was imprudent in its purchase; i.e., it could have obtained power elsewhere at a lower cost. Therefore, regulators must obtain information about alternative prices and sources of electric energy that were available to the utility at the time of purchase. This suggests that regulators must replicate the purchasing utility's activities, which is likely to be costly.

APPROPRIATE CARRIIER STATUS

This leads to a third policy issue: the appropriate carrier status for transmission and distribution services. Carriers fall into three general classes: (1) voluntary contract carriers; (2) mandatory contract carriers; or (3) common carriers. A private contract carrier has the legal right to select its customers while a common carrier must serve the public at large. (Harper 1978, p. 121). Moreover, the common carrier does not take possession of the goods transported. (p. 113). The obligations of a mandatory contract carrier are not well-defined and depend on the discretion of the regulatory commission or other government body. The mandatory carrier would be required to provide service to the general public, but would be allowed to own some or all of the goods it transports.

Currently, utilities are voluntary carriers with respect to transmission service. Furthermore, FERC authority over wheeling service is limited. (Burns 1987, pp. 51-69). Nevertheless, the FERC may be able to use its conditioning authority and tie wheeling requirements to proposals which lessen regulatory oversight of generation. FERC has used this authority extensively in the natural gas industry.

Common carrier status would sacrifice possible vertical integration efficiencies, but eliminate the incentive for self-dealing. This study focused on cost complementarity between the production stages and our findings indicate a lack of complementarity. However, economies of scope over the product set or between two products may still exist. Additional research is needed to test directly for scope economies before any definite conclusions can be reached about the nature of the appropriate carrier status.

One possible extension would be to estimate a "hybrid" translog cost function where the output variables are expressed in a Box-Cox transformation. (Caves, Christensen, and Tretheway 1980; Sing 1987). This would allow for zero values and, thus, permit direct calculation of scope economies. Another possibility would be to adopt a quadratic cost function approach, such as in Rolle (1990b) and K-M (1991), but allow for linear homogeneity and interaction between output and other vafiables, including input prices. For example, the "composite" cost function developed by Pulley and Braunstein (1992) may prove useful in this regard.

Finally, an extension the author is currently undertaking involves applying the subadditivity test developed by Evans and Heckman (E&H) (1984).(17) E&H estimate various degrees of subadditivity over an admissible range of observed output levels in their sample. For any given output level Y, E&H divide it between two hypothetical firms and estimate their respective costs. If the single firm's total cost exceeds the sum of the two hypothetical firms' costs, local subadditivity can be rejected. Moreover, if the cost function is not subadditive over this region, then global subadditivity can be rejected. On the other hand, the test is local, such that a finding of subadditivity over the admissible region need not imply global subadditivity. Morevices, Inc. provides the company's annualized dividend yield. Based on the 48 company sample, a 14.29 percent nominal cost of common equity is obtained. I also divided the 48 companies into two beta groups (.55-.6 and .65-.7) and calculated a cost of equity for each. Subsequently, I tested for equality between sample estimates and found that equality of cost of equity estimates can not be rejected at a 5 percent level of significance. (7.) Autility's cost of purchased power equals the product of the seller's regulated wholesale rate and the quantity sold. Thus, purchased power costs reflect the production costs of the seller, not the purchaser. The purchased power is simply a transfer from the producer to the consumer; it does not reflect any productive activity by the purchasing utility in and of itself. The productive activity of the purchasing utility involves establishing the purchase agreement; that is, making arrangements for the transfer of power from producer to consumer The purchaser undertakes this brokerage function in order to serve its ultimate consumers. Thus, the purchase agreement is one aspect of the utility's distribution function. The costs of obtaining the agreement are reflected in the utility's operating expenses and general plant. (8.) See Stevenson (1980). Stevenson uses a 20 year interval to calculate a weighted average price index. The denominator, 210, is the sum of the years digits. (9.) EEI graciously supplied fuel cost data published in the Uniform Statistical Reports for PacifiCorp, and Puget Sound Power and Light. (10.) The average ratios of non-nuclear steam generation to total net generation and total steam generation are 91 percent and 93 percent, respectively, for the sample utilities. (11.) See Christensen and Green (1976), Stevenson (1980), and Sing (1987) for examples. This view is not universal, however. See Atkinson and Halvorsen (1984). (12.) The estimated standard errors for each cost complementarity condition is based on the approximation formula given in Kmenta (1971, p. 444). (13.) Specifically, there is a 94.4 percent, 63 percent, and 86.4 percent probability that the Generation-Distribution, Generation-Transmission, and Transmission-Distribution estimates, respectively, are greater than zero. (14.) Calculation of the standard errors follows the formula as noted in footnote 12. (15.) The value of the output variables at the point of approximation are as follows: Generation-15,106,086 MWE; Transmission456,187,645 CVM; and Distribution-660 ultimate customers. (16.) Combining transmission and distribution, capital costs may account for as much as 35 percent and 75 percent of the total cost associated with generation and transmission-distribution, respectively. See Gilsdorf (1990, p.57). (17.) Shin and [Y.sub.i]ng (1992) recently applied the E&H subadditivity test on local exchange carriers and addressed the concerns raised by Roller (1990a).

REFERENCES

[1] Atkinson, Scott E. and Robert Halvorsen. 1984. "Parametric Efficiency Tests, Economies of Scale, and Input Demand in U.S. Electric Power Generation." Review of Economics and Statistics 25: 647-662. [2] Barten, A. R 1969. "Maximum Likelihood Estimation of a Complete System of Demand Equations." Eurapean Economic Review 1: 7-73. [3] Baumol, William J., John C. Panzar, and Robert D. Willig. 1982. Contestabe Markets and the Theory of Industry Structure San Diego: Harcourt Brace Jovanovich. [4] Brown, Randall S., Douglas W Caves, and Laurits R. Christensen. 1979. "Modelling the Structure of Cost and Production for Multiproduct Firms." Southern Economic Journal 46: 256-273. [5] Burns, Robert E. 1987. "Legal Impediments to Power Transfers." Pp. 51-99 in Non-Technical Impedimmts topower Transfers, edited by Kevin Kelly. Columbus: National Regulatory Research Institute. [6] Caves, Douglas W., Laurits R- Christensen, and Michael W. Tretheway. 1980. "Flexible Cost Functions for Multiproduct Firms." Review of Economics and Statistics 62: 477-481. [7] Chambers, Robert G. 1vices, Inc. provides the company's annualized dividend yield. Based on the 48 company sample, a 14.29 percent nominal cost of common equity is obtained. I also divided the 48 companies into two beta groups (.55-.6 and .65-.7) and calculated a cost of equity for each. Subsequently, I tested for equality between sample estimates and found that equality of cost of equity estimates can not be rejected at a 5 percent level of significance. (7.) Autility's cost of purchased power equals the product of the seller's regulated wholesale rate and the quantity sold. Thus, purchased power costs reflect the production costs of the seller, not the purchaser. The purchased power is simply a transfer from the producer to the consumer; it does not reflect any productive activity by the purchasing utility in and of itself. The productive activity of the purchasing utility involves establishing the purchase agreement; that is, making arrangements for the transfer of power from producer to consumer The purchaser undertakes this brokerage function in order to serve its ultimate consumers. Thus, the purchase agreement is one aspect of the utility's distribution function. The costs of obtaining the agreement are reflected in the utility's operating expenses and general plant. (8.) See Stevenson (1980). Stevenson uses a 20 year interval to calculate a weighted average price index. The denominator, 210, is the sum of the years digits. (9.) EEI graciously supplied fuel cost data published in the Uniform Statistical Reports for PacifiCorp, and Puget Sound Power and Light. (10.) The average ratios of non-nuclear steam generation to total net generation and total steam generation are 91 percent and 93 percent, respectively, for the sample utilities. (11.) See Christensen and Green (1976), Stevenson (1980), and Sing (1987) for examples. This view is not universal, however. See Atkinson and Halvorsen (1984). (12.) The estimated standard errors for each cost complementarity condition is based on the approximation formula given in Kmenta (1971, p. 444). (13.) Specifically, there is a 94.4 percent, 63 percent, and 86.4 percent probability that the Generation-Distribution, Generation-Transmission, and Transmission-Distribution estimates, respectively, are greater than zero. (14.) Calculation of the standard errors follows the formula as noted in footnote 12. (15.) The value of the output variables at the point of approximation are as follows: Generation-15,106,086 MWE; Transmission456,187,645 CVM; and Distribution-660 ultimate customers. (16.) Combining transmission and distribution, capital costs may account for as much as 35 percent and 75 percent of the total cost associated with generation and transmission-distribution, respectively. See Gilsdorf (1990, p.57). (17.) Shin and [Y.sub.i]ng (1992) recently applied the E&H subadditivity test on local exchange carriers and addressed the concerns raised by Roller (1990a).

REFERENCES

[1] Atkinson, Scott E. and Robert Halvorsen. 1984. "Parametric Efficiency Tests, Economies of Scale, and Input Demand in U.S. Electric Power Generation." Review of Economics and Statistics 25: 647-662. [2] Barten, A. R 1969. "Maximum Likelihood Estimation of a Complete System of Demand Equations." Eurapean Economic Review 1: 7-73. [3] Baumol, William J., John C. Panzar, and Robert D. Willig. 1982. Contestabe Markets and the Theory of Industry Structure San Diego: Harcourt Brace Jovanovich. [4] Brown, Randall S., Douglas W Caves, and Laurits R. Christensen. 1979. "Modelling the Structure of Cost and Production for Multiproduct Firms." Southern Economic Journal 46: 256-273. [5] Burns, Robert E. 1987. "Legal Impediments to Power Transfers." Pp. 51-99 in Non-Technical Impedimmts topower Transfers, edited by Kevin Kelly. Columbus: National Regulatory Research Institute. [6] Caves, Douglas W., Laurits R- Christensen, and Michael W. Tretheway. 1980. "Flexible Cost Functions for Multiproduct Firms." Review of Economics and Statistics 62: 477-481. [7] Chambers, Robert G. 1W. Golub. 1984. "Estimating Effective Concentration in Deregulated Wholesale Electricity Markets." Rand Journal of Economics 15: 12-26. [34] Shin, Richard T and John S. [Y.sub.i]ng. 1992. "Unnatural Monopolies in Local Telephone." Randjoumal ofeconomics 23:171-183. [35] Sing, Merrile. 1987. "Are Combination Gas and Electric Utilities Mtiltiproduct Natural Monopolies?" Review of Economics and Statistics 69. 392-398. [36] Standard and Poor's Compustat Services, Inc. 1987. Utility Compustat II. [37] Stevenson, Rodney. 1980. "Measuring Technological Bias." American Economic Review 70: 162-173. [38] U.S. Department of Energy. 1985. Financial Statistics of Seleted Eletric Utilities. Washington, DC: Government Printing Office. [39] Wales, Terence J. 1977. "On the Flexibility of Flexible Functional Forms." Journal of Econonletrics 5. 183-193. [40] Weiss, Leonard W. 1975. "Antitrust in the Electric Power Industry." In Promoting Competition in Regulated Markets, edited by Almarin Phillips. Washington DC: Brookings Institution. [41] Zellner, A. 1962. "An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias." Journal of American Statistical

Printer friendly Cite/link Email Feedback | |

Author: | Gilsdorf, Keith |
---|---|

Publication: | Quarterly Review of Economics and Finance |

Date: | Sep 22, 1994 |

Words: | 7548 |

Previous Article: | Common stochastic trends in Pacific rim stock markets. |

Next Article: | A methodological investigation of risk exposure of bank off-balance sheet loan commitment activities. |

Topics: |