| Using stochastic distance and directional distance functions I assess the efficiency of firms in the U.S. Electric Utility Industry. Further, I compare utilities' marginal control costs for SO2 emissions to prices observed in the SO2 allowance marketplace to evaluate the accuracy of the model for predicting permit prices.;I estimate stochastic distance functions and use them to calculate Malmquist productivity change (PC) indices which I decompose into efficiency change and technical change. Data Envelopment Analysis (DEA), which uses linear programming techniques, has been widely used to compute Malmquist indices as ratios of fitted distances from a convex hull frontier. As an alternative, I develop and estimate a flexible, stochastic distance frontier, using a generalized method of moments strategy. This allows for statistical inference and imposes no restrictions on returns to scale, Finally, comparisons are drawn between the stochastic approach and the less flexible non-parametric and non-stochastic DEA method using a panel of 43 electric utilities over a time period from 1961 to 1992. Applying the same methodology I analyze a panel of 78 firms over the time period from 1988 to 1997. Based on my results I identify the most and least efficient firms in the industry.;Further, I develop a methodology to explicitly take into account undesirable outputs. These outputs are produced in conjunction with desirable outputs and are not costlessly disposable. I specify and estimate input and output based stochastic distance and directional distance functions and compare the alternative approaches using panel data on 41 U.S. electric utilities observed in 1980, 1985, 1990, and 1995. I apply the preferred approach, the output directional distance function, to a sample of 56 primarily coal-burning utilities observed in 1995, 1996, and 1997. Of particular interest is the shadow price for the undesirable output. I compare the shadow values to prices from the SO2 allowance market and find that the estimated marginal industry control cost is reasonably close to prices observed in the marketplace. |
