In a recent article ("The Efficient Utility: Labor, Capital, and Profit," Sept. 1, 1995), Taylor and Thompson attempt to measure the
economic efficiencies of 19 investor-owned utilities.
The authors use a method of efficiency measurement proposed by M.J. Farrell in a pioneering paper published nearly 40 years ago. Farrell's approach decomposes overall profit efficiency into two components: technical efficiency (the ability to produce a given physical output with a minimum quantity of inputs) and allocative efficiency (the choice of the optimal combination of inputs given the prices of the inputs).
Farrell's model belongs to a class known as "deterministic frontier models," which suffer from several critical deficiencies. First, any deviation from the frontier is attributed to the firm's inefficiency. In reality, some departures from the frontier are due to random exogenous factors (such as the weather). Second, these models are sensitive to extreme observations. If these extreme observations arise from measurement errors, estimates of the frontier will obviously be inaccurate.
We believe that a "stochastic" approach ameliorates these shortcomings by reflecting exogenous shocks that shift firms away from the efficient frontier.
In spite of its weaknesses, the deterministic frontier approach can yield important insights. Unfortunately, Taylor and Thompson's approach is flawed. First, the authors specify profit rather than physical production as the measure of the firm's "output" produced by labor and capital services. This specifi-cation confounds managerial
inefficiencies with the effects of differences in input prices, factors over which managers have limited control. Optimized profit is not directly a function of labor and capital. Rather, profit is indirectly achieved by (i) maximizing the output from a given combination of labor and capital inputs, and (ii) minimizing total costs, given input prices. Thus, deviations from a "profit frontier" include both technical inefficiency (waste) and allocative inefficiency (failure to minimize cost).
Second, we believe all variables in their analysis are inappropriately measured. The authors measure capital by the book value of total assets, which includes cash and investments in associated or subsidiary companies. We believe the appropriate proxy for the physical capital of the utility is the value of total utility plant. The labor variable appears to be measured by the total number of utility employees, with no adjustment for part-time workers, which could provide a distorted measure of labor hours used in production.
Third, the authors omit fuel as an input in the production process and make no adjustment for purchased power. Fuel efficiency is an important component of overall utility efficiency. Further, purchased power serves as an input in the production process of delivered power and its omission potentially distorts the evaluation of utility efficiency.
Finally, the authors make no attempt to distinguish short-run efficiency (capital stock is fixed) from long-run efficiency (capital stock can be varied). Observed technical inefficiency may be caused by poor management, but may also be due to the use of capital of older vintage. This distinction is important in the electric power industry, which is extremely capital-intensive and subject to lengthy adjustment lags.
Figure 1 compares the two frontiers: Taylor and Thompson's, with gross profits as the output measure; and an alternative that uses net generation as the output measure. Each ratio variable has been standardized to the unit interval for easier comparison. Perhaps the most striking feature of the plot that uses net generation as output is the relative position of San Diego Gas & Electric (SDG&E), which now appears a significant distance from the frontier.
The Center for Regulatory Studies is currently engaged in an extensive analysis of electric utility efficiency issues, applying stochastic frontier methods. Our results should shed considerable light on the types of questions raised but left unanswered by Taylor and Thompson.
Matthew J. Morey, Director
L. Dean Hiebert, Associate Director
Center for Regulatory Studies
The authors respond:
We welcome the analysis by Morey & Hiebert (M&H). But their arguments led us to imagine the mood in Detroit's boardrooms in the 1970s: "We don't need to worry about those Japanese cars." Today we know how that turned out.
First, M&H claim better results from stochastic measurements (em e.g., statistical regression (em than from DEA Best Practice. Experts have long debated the merits and demerits of these alternative techniques. The recent monographs edited by Fried et al. (The Measurement of Productive Efficiency, Techniques and Applications, New York: Oxford Univ. Press, 1993) and Charnes et al.(Data Envelopment Analysis, Theory, Methodology and Applications, Boston: Kluwer Acad. Pub., 1994) include bibliographies of well over 500 publications. We doubt that utility managers have time to address this debate, but several key points warrant examination.
First, decisions based on the "average" can lead to mediocrity. Figure 2 depicts four hypothetical firms (A, B, C, & D) using two factors (em labor and capital. A, B, & C determine the Best Practice frontier; D lags behind with excessive labor and capital use. If A, unaware of its position, used the average as the norm (em reducing labor costs, but failing to increase capital outlays (em it could lose its Best Practice status and see its costs rise. If D pursued the average, its situation would improve, but remain mediocre.
Stochastic frontier methods provide an average, as do virtually all statistical methods, particularly regression techniques. DEA Best Practice, in contrast, reveals the best-in-class performers. No known statistical technique provides such revelations.
Second, M&H claim better input-output identification. Measurement experts have long argued about the best variable representations. We sought to use definitions in accordance with both FERC reports and financial statement accounting practices. FERC Form 1 accounts solely for utility activities. This point is important because some utilities today engage in nonutility business, such as cable TV. And quibbling about cash as capital ignores that it forms a basic requirement of continuity. Similarly, some firms are reportedly outsourcing much of their labor under a part-time classification.
Third, M&H measure output by electricity units, but that focus ignores the real world. As the data shows, some utilities produce more than one output (em namely, gas and electricity. Not accounting for the gas part of the business may significantly underestimate total output and bias, and could flaw the efficiency estimates. Gas revenues at SDG&E made up nearly 20 percent of revenues in 1992-93. Moreover, the inputs of labor and capital cannot be separated in the data into electricity and gas functions. DEA Best Practice provides a method to analyze multiple output problems; stochastic frontier methods are limited to one output.
We believe that utility managers need simple tools, based on real-world data. DEA's Best Practice approach provides strategic guidance for the uncharted future.
As an analysis of utilities, the Taylor and Thompson article (Sept. 1, 1995), offers a starting point, but appears too academic to provide valuable insight. The authors provide data, but they treat it as information. They fail to give the reader additional knowledge that explains the data (em knowledge that would turn the data into information. For example, what are the underlying reasons for the movements depicted by the article's Figure 3? Let me suggest other areas for more exploration: [Editor's note: All references to figures cite the original article.]
s It is well known that the net investment per kilowatt of nuclear generating stations greatly exceeds that for fossil-fired stations. Yet the authors' group of 19 utilities includes five that have no nuclear investment. Why do only two of the nonnuclear utilities show up at the extreme left of Figures 1 and 2? And why do the nuclear utilities range from the extreme left to the extreme right?
s Texas Utilities (TU) placed two nuclear units in service and went through a well-publicized downsizing during the authors' analysis period. How did these events influence past asset and profit amounts and future profit expectations?
s For many years Texas-New Mexico Power outsourced its power supply, then embarked upon the construction of a large generating unit that caused sufficient difficulties to trigger a management change during the authors' analysis period. How did this situation influence the amounts disclosed in the authors' Appendix?
s Some time ago, San Diego Gas & Electric (SDG&E) decided to quit building generating units. How do Figures 1 and 2 reflect outsourcing risk and the SDG&E decision? Also, the authors suggest that six of the utilities in the lower left corner of Figure 2 show "Good Cost and Revenue Management" (Figure 4). However, varying uses of outsourcing could be distorting the relationships.
In addition, I question the validity of the authors' suggestion that the relationship of TU to Tampa Electric Co. (TEC) and SDG&E on Figure 2 verifies the Averch-Johnson hypothesis. It's apples and oranges: TU is a nuclear utility, TEC is not, and neither has outsourced any significant portion of their power-supply function.
I also question the validity of the statement: "A utility can trade off labor and capital to move back and forth along the efficiency frontier." I doubt that electric utilities have the suggested substitution capability.
Greater elaboration of these and other aspects of the data would have truly informed the reader.
John S. Ferguson
The authors respond:
We wanted to show that financial statement data can be analyzed using the DEA Best Practice technique. We skirted many additional complexities, such as:
s The economic dynamics of adjustment (as in Figure 3)
s Rates of return on possible stranded investments (e.g., TU)
s The price of reliability, including who pays for it.
Ferguson does raise some key issues that beg for further analysis, but we believe his letter distorts our example of the Averch-Johnson (AJ) hypothesis. Historically, TU could have relied more heavily on outsourced, wheeled interutility electricity within Texas (under ERCOT). Its current fixed-asset problems result from its previous strategic choices. Ignoring that, TEC and SDG&E might be dropped from the peer group in Figure 2. Then, Florida Power & Light would move up to best-in-class utility. Again, TU is out of economic position because it employs too much capital and too little labor. That is the essence of the AJ hypothesis.
As economists have long taught, "Everything has its price." That is true in the management of all competitive firms. Utilities must now exploit their substitution possibilities to meet the competitive challenge. t
Articles found on this page are available to Internet subscribers only. For more information about obtaining a username and password, please call our Customer Service Department at 1-800-368-5001.