Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium Member Libraries http://www.archive.org/details/doescompetitionrOOmark 1 Technology Department of Economics Working Paper Series Massachusetts Institute ot DOES COMPETITION REDUCE COSTS? ASSESSING THE IMPACT OF REGULATORY RESTRUCTURING ON U.S. ELECTRIC GENERATION EFFICIENCY Kira Markiewicz Nancy Rose Catherine Wolfram Working Paper 04-37 November 2004 RoomE52-251 50 Memorial Drive Cambridge, 02142 MA This paper can be downloaded without charge from the Social Science Research Network Paper Collection http://ssrn.com/abstract=6 1 828 at MASSACHUSETTS INSTITUTE OF TECHNOLOGY DEC 7 200^ LIBRARIES Does Competition Reduce Costs? Assessing the Impact of Regulatory Restructuring on U.S. Electric Generation Efficiency Kira Markiewicz UC Berkeley, Haas Nancy School of Business L. Rose MIT and NBER Catherine Wolfram UC Berkeley and NBER* November 2004 * markiewi(a),haas. berkeley.edu nrosefajmit.edu wolfram@haas.berkelev.edu . from the MIT , . Rose acknowledges support We thank NBER 10 Summer Institute Meeting, the Center for Energy and Environmental Policy Research and the Hoover Institution. participants at the NBER Productivity Program Meeting, the POWER conference, and the MIT Center for Energy and Environmental Policy Research conference, as well as seminar participants at Harvard, MIT, UC Berkeley, UC Davis and Yale for their suggestions. We are particularly grateful for the detailed comments on earlier University of California Energy Institute drafts provided also thank Tom by Al Klevorick, Mark Roberts, Charles Rossman and Johannes Van Biesebroeck. Wilkening for assistance in coding restructuring policy characteristics across states. We Does Competition Reduce Costs? Assessing the Impact of Regulatory Restructuring on U.S. Electric Generation Efficiency Kira Markiewicz Nancy L. Rose Catherine Wolfram Abstract Although the allocative efficiency benefits of competition are a tenet of microeconomic theory, the relation between competition and technical efficiency models of profit-maximization subsume static is less well understood. Neoclassical cost-minimizing behavior regardless of market competitiveness, but agency models of managerial behavior suggest possible scope for competition to influence cost-reducing effort choices. This paper explores the empirical effects of competition on technical efficiency in the context of electricity industry restructuring. Restructuring programs adopted by many U.S. states made utilities residual savings and increased their exposure to competitive markets. claimants to cost We estimate the impact of these changes on annual generating plant-level input demand for non-fuel operating expenses, the number of employees and for the most fuel use. part unaffected We find that municipally-owned plants, whose owners were by restructuring, experienced the smallest efficiency gains over the past decade. Investor-owned utility plants in states that restructured their wholesale electricity markets had the largest reductions in nonfuel operating expenses and employment, while investor- owned plants in nonrestructuring states fell between these extremes. The analysis also highlights the substantive importance of treating the simultaneity of input and output decisions, which we do through an instrumental variables approach. JEL Codes: L Keywords: Efficiency, Production, Competition, Electricity restructuring, Electric 1 1 , L43, L5 1 , L94, D24 Generation, Regulation Economists have long argued that competition generates important efficiency benefits for an economy. These generally focus on allocative efficiency; the implications of competition technical efficiency are less clear. Neoclassical models of profit-maximization cost-minimizing behavior by however, in all firms, regardless of market competitiveness. 1 subsume for static Agency models, recognizing the interplay of asymmetric information with the separation of management and control, suggest possible deviations may imply managers. These models from cost-minimization by effort-averse a role for competition in constraining managerial behavior, by rewarding efficiency gains and confronting less-efficient firms with the choice of cost reduction to the level of their lower-cost counterparts or exit; see Nickell (1996) for a brief discussion of some of these theoretical arguments. Their actual relevance is ultimately an empirical question. This paper assesses the effect of competition on technical efficiency using data on the U.S. The electric generation sector. past decade has witnessed a dramatic transformation of this industry. Until the mid-1990s, over ninety percent of the electricity in the vertically-integrated investor-owned utilities (IOUs), IOUs in many states own sold by most operating as regulated monopolists own within their service areas. Today, non-utility generators capacity nationwide, and US was roughly a quarter of generation only a small fraction of total generating capacity and operate in a partially deregulated structure that relies heavily on market-based incentives or competition. may have been motivated While studies of state-level in large part by rent-seeking many proponents of restructuring argued outcomes would yield efficiency gains prices. (e.g., electricity restructuring suggest politicians (e.g., White, 1996, and Joskow, 1997), that exposing utilities to competitive, market-based that could ultimately reduce electricity costs and retail Research on other industries suggests productivity gains associated with deregulation Olley and Pakes, 1996, on telecommunications and Ng and Seabright, 2001, on airlines) and with increased competitive pressure caused by factors other than regulatory change Galdon-Sanchez and Schmitz, 2002, on iron ore mines). The considerable body of academic work on (e.g., 2 electricity restructuring within the U.S. and abroad has thus far focused on assessing the performance of competitive wholesale markets, with 1 The implication of competition dynamic efficiency through innovation is the subject of an extensive from Schumpeter's 1943 classic Capitalism, Socialism, and Democracy. for theoretical and empirical literature in economics, dating at least example Borenstein, Bushnell and particular attention to the exercise of market power Wolak, 2002 and Joskow and Kahn, 2002). While many of the costs of electricity restructuring have been intensively studied, relatively little (see for effort has been devoted to quantifying any ex post operating efficiency gains of restructuring, although a few studies (e.g., Knittel, analyzed efficiency effects of various incentive regulations in this sector. the first such, is it 3 2002) have This study provides substantial analysis of early generation efficiency gains of electricity restructuring. contributes to the broad economic debate on the role of competition in the of direct policy relevance to states As economy and contemplating the future of their electricity restructuring programs. The results of this work indicate that plant operators most affected by restructuring reduced labor 5% or more relative to and nonfuel expenses, holding output constant, by roughly and by 15-20% owned utility plants, which were largely unaffected by restructuring the (IOU) plants, medium-run efficiency gains that relative to Joskow (1997, government- and cooperatively-owned p. incentives. treating the simultaneity productivity may interpreted as 214) posits "may be associated with Our work of input and output choice. may be These improving the operating performance of the existing stock of generating the productivity of labor operating these facilities." other investor- facilities and increasing also highlights the importance of Failing to recognize that shocks to input induce firms to adjust targeted output leads to overstatement of estimated efficiency effects, in some cases by a factor of two or more. While endogeneity concerns have been long recognized in the productivity generation to compensate for this. literature, Finally, we ours is one of the first studies of electric explore the sensitivity of the estimated efficiency impact to the choice of control group to which restructured plants are compared, and discuss the issues involved in determining the appropriate counterfactual. 2 Some hint municipally situations 3 of this possibility owned in electricity is provided by Primeaux (1977), and found a significant decrease in costs per One exception is who compared firms facing competition to a matched sample of municipally kWh for firms owned a sample of monopoly firms in facing competition. who uses stochastic frontier production functions to estimate generation One set of independent variables he includes is indicators for regulatory Hiebert (2002), plant efficiency over 1988-1997. orders or legislative enactment of restructuring reforms in 1996 and in 1997. While he finds significant reductions in mean inefficiency associated with restructuring laws in 1996 for coal plants, he finds effects for gas plants, nor for either fuel type in 1997. Our work uses a longer time no period, richer characterization of the restructuring environment and dating of reforms consistent with the U.S. Energy Information Administration, and an alternative technology specification that allows for more complex Joskow (1997) describes the significant moved from state-owned competitive generation market, although these mix restructuring and productivity shocks and treats possible input endogeneity biases. labor force reductions that accompanied restructuring in the monopoly to a privatized, privatization effects. UK, as the industry The remainder of the paper is organized as follows: Section competitive effects of efficiency, and discusses efficiency. Section 2 details our empirical how 1 describes existing evidence on the restructuring might alter electric generation methodology for testing these predictions, describes our strategy for identifying restructuring effects. The data and are described in Section 3. Section 4 reports the results of the empirical analysis, and Section 5 concludes. Why Might Restructuring Affect Generator Efficiency? 1. Exit by less-efficient firms is a well-understood efficiency benefit of competition: as output shifts from (innately) higher-cost firms to lower-cost competitors the production cost for a given total output level decline. Olley and Pakes (1996) provide empirical evidence of this their plant-level analysis of the magnitude and source of productivity gains phenomenon in in the U.S. telecommunications equipment industry over 1974-1987. They find substantial increases in productivity associated with the increased competition that followed the 1984 divestiture and deregulation in this sector, and identify the primary source of these gains as the re-allocation of output from less productive to more productive plants across firms. In a similar vein, Syverson (2004) finds that more competitive local markets in the concrete industry are associated with higher mean, less dispersion, and higher lower-bounds in plant productivity, effects he attributes to the exit The of less-efficient plants existing evidence efficiency gains in more competitive environments. on whether competition also leads to cost reductions through technical by continuing producers and plants is relatively sparse. Nickell ( 1 996) uses a panel of 670 U.K. manufacturing firms to estimate production functions that include controls for the competitive environments in which firms operate. He finds some evidence of reduced productivity levels associated with market growth rates in power and strong support for higher productivity more competitive environments. Concerns about the ability analysis to control adequately for unobservable heterogeneity across sectors specific evidence tighter and more convincing. 4 Schmitz (2002) study of labor productivity gains A notable example at iron is the of cross-industry may make sector- Galdon-Sanchez and ore mines that faced increased competitive pressure following the collapse of world steel production in the early 1980s. They find unprecedented rates of labor productivity gains associated with this increase in competitive 4 A number of studies have analyzed see, for cases efficiency gains following regulatory reform in various industries; (1998) on airlines. Unfortunately, in many on efficiency (e.g., operating restrictions imposed those sectors) from the indirect effects of reduced example, Bailey's (1986) overview and Park it is et al. difficult to disentangle direct regulatory effects on trucking firms or competition. airlines by regulators in pressure, "driven by continuing mines, producing the same products and using the same technology as they had before the 1980s" (Galdon-Sanchez and Schmitz, 2002, Several features of the electric generation sector competitive effects on technical efficiency. make 6 First, it p. 1233). 5 an attractive subject for testing potential generation technology is reasonably stable and well-understood and data on production inputs and outputs at the plant-level are readily available to researchers. This has made electric generation a common application for new production and cost function estimation techniques, dating at least to Nerlove (1963). Second, policy shifts over a relatively short period have resulted in a dramatic transformation of the market for electric power. Through the early 1990s, the U.S. electricity industry was dominated by vertically Most operated integrated investor-owned utilities (IOUs). as regulated monopolists over generation, transmission, and distribution of electricity within their localized geographic market, though there was some wholesale power traded among utilities or purchased from a small but growing number of non-utility generators. Prices generally were determined by based on accounting costs of service at the firm level. By state regulators 1998, every jurisdiction (50 states and the District of Columbia) had initiated formal hearings to consider restructuring their electricity sector, and by 2000, almost half had approved legislation introducing some form of competition including retail access. 7 This provides both time series and geographic variation environments. Third, static and dynamic efficiency claims bolstered measuring these benefits is a vital prerequisite to assessing the much of the wisdom of these in competitive policy reform; policies. has long been argued that traditional cost-of-service regulation does relatively well in limiting It rents but less well in providing incentives for cost-minimizing production; see Laffont (1993). Under pure cost-of-service regulation, regulator-approved costs of the utilities are passed and reductions directly through to customers, profits until rates are revised to reflect the in the cost raises operations costs Ng and of service yield new lower costs asymmetric information between regulators and firms, 5 above minimum cost at the at next rate case. inefficient behavior levels generally most short-term would be 8 Given by managers that reflected in increased rates Seabright (2001) estimate cost functions for a panel of U.S. and European airlines over 1982- 1995, and conclude that potential gains from further privatization and increased competition European and Tirole carriers are substantial, among though they point out that the best-measured component of these gains ownership rather than market structure differences. Understanding possible reallocation of output across plants relates to 6 available databases 7 when they is hampered by the exit of plants from most are sold to non-utility owners. of California's electricity crisis in 2000-2001, restructuring has become less popular and have delayed or suspended restructuring activity, including six that had previously approved In the aftermath many states retail access legislation. See US Energy Information Administration (EI A), 2003. and passed through to customers. Joskow (1974) and Hendricks (1975) demonstrate that in cost-of-service regulation, particularly those arising resetting hearings), may impact generally limited, is provide some incentives frictions from regulatory lag (time between pricemargin for cost-reducing at the effort. Their however, apart from periods of rapid nominal cost inflation (see Joskow, 1974). This system led economists to argue that replacing cost-of-service regulation with higher- powered regulatory incentive schemes or increased competition could enhance the 1 980s and early 1 incentive regulation. 990s, The many state utility monopoly structure, suggests studies a variety of incentive regulations in use through 1996, plant performance or fuel cost 9 Over commissions accordingly adopted some form of limited empirical evidence available on these reforms, price setting within the regulated More efficiency. were associated with gains mixed which modify results. Knittel (2002) and finds that those targeted at in plant-level generation efficiency. 10 general reforms, such as price caps, rate freezes, and revenue-decoupling programs, typically were associated with insignificant or negative efficiency estimates, all Restructuring, in contrast to incentive regulations, fundamentally changed the earn revenue. At the wholesale level, plants sell either else equal. way plant owners through newly created spot markets or through long-term contracts that are presumably based on expected spot prices. In the spot markets, plant owners submit bids indicating the prices at which they are willing to supply power from their plants. plant is paid to all Dispatch order is set by the bids, and, in most markets, the bid of the marginal plants that are dispatched. High-cost plants will be forced order, reducing likely revenue. Plant operators that reduce costs order, increasing dispatch probability, and increase the profit move 8 way in which retail customers are allocated." the dispatch way own costs and the retail rates are Retail access programs in Rates are constant between rate cases, apart from certain specific automatic adjustments (such as fuel adjustment clauses), so changes in cost would not be reflected 9 in higher in the dispatch margin between expected market price. Most restructuring programs also changed the determined and the down in rates until the See, for example, Laffont and Tirole, 1993, for a theoretical justification, or next rate case. Joskow and Schmalensee, 1987, for an applied argument. 10 Knittel uses OLS and stochastic production frontier techniques to estimate Cobb-Douglas generating IOU plants over 1981-1996. His plant production functions in capital, labor, and fuel for a panel of large results from first-differenced models, which implicitly allow for plant-level fixed efficiency effects, 1-2% associated with these reforms. Equations that do not allow for plant suggest gains on the order of fixed effects suggest States much larger magnitudes. have used a variety of approaches to link retail rates market. Over the short term, most states decoupled freezes, often at levels discounted utility from pre-restructuring under restructuring to wholesale prices revenue from costs by mandating prices. Some aggressively trying to encourage entry by competitive energy suppliers, retail customers. states, in the retail rate such as Pennsylvania, are who may contract directly with combination with the creation of the new wholesale spot markets may increase the intensity of cost-cutting incentives, leading to even greater effort to improve efficiency. While the most significant savings long-run investments in new from restructuring are capacity, there may be likely to be associated with efficient opportunities for modest reductions in operating costs of existing plants (see Joskow, 1997). This paper attempts to measure the extent of that possible improvement for the existing stock of electricity generating plants The implicit null hypothesis null, there should be no change in We discuss below our plant-level efficiency measures associated with restructuring activity. for estimating plant efficiency and identifying deviations from this hypothesis. the effects of restructuring requires specification of how generating plants operated absent the policy change. Constructing this counterfactual 2. U.S. before restructuring, operators were minimizing their costs, is that, given the capital stock available in the industry. Under the method in the is Assessing would have been crucial, but difficult. Empirical Model For a single-output production process, productive efficiency can be assessed by estimating whether a plant is maximizing output given its inputs and whether it is using the best mix of inputs given their relative prices. Production functions describe the technological process of transforming inputs to outputs and ignore the costs of the inputs; a plant production frontier. Cost minimization assumes that, is efficient if it is on the given the input costs, firms choose the mix of inputs that minimizes the costs of producing a given level of output. A plant could be producing the most output possible from a given input combination, but not minimizing costs for instance, labor labor. Even materials, it if was cheap relative to materials, yet the plant used a lot of materials relative to the firm were producing the would not be if, efficient if by substituting labor for materials. it maximum output possible from its workers and could produce the same level of output less expensively We explore the impact of restructuring on specifying a production function and then deriving the relevant input efficiency by demand equations implied by cost minimization. We adopt the convention of representing generating plant output (Q) by the net energy the generating units produce over some period (measured by annual megawatt-hours, data), a choice that is discussed in further detail in the data section below. studies of electric plant productivity model this output as MWh, in our While a multitude of a function of current inputs, often using a Cobb-Douglas production or cost function, the characteristics of electricity production argue strongly for an alternative specification. We derive a model of production and cost minimization that is sensitive to important institutional characteristics of electricity production that have been largely ignored in the earlier literature. First, observed output given its in general will be the lesser of the output the plant (Q ) capable of producing, available inputs, and the output called for by the system dispatcher. Because the system demand dispatcher must balance total production with p is and actual (Q A ) output for a given plant other plants available for dispatch, and plant i will at each moment, the gap between probable be a function of demand realizations, the set of position in the dispatch order. i's 12 Second, while fuel inputs are varied in response to real-time dispatching and operational changes, other inputs to a plant's production are determined in advance of output realizations. Capital typically changed chosen is at the time of a unit's construction (or retirement), and at the plant level From relatively infrequently. input. Utilities hire labor and the manager's perspective, set operating it may be and materials expenditures in is considered a fixed advance, based on expected demand. While these can be adjusted over the medium-run, staffing decisions as well as most maintenance expenditures are not tied Q We therefore advance of actual production, and determining a target level of probable treat these as set in output, to short-run fluctuations in output. p . Finally, while labor, materials, and capital may be to some extent substitutable to produce probable output, the generation process generally does not allow these inputs to substitute for fuel in the short-run. for plant i Q Given in year A lt A t = min[ g(E Q input, K for capital, unobserved of the technology, we posit a Leontief production process of the following form: where is this description it , TE , actual output and L e it E Q for labor, ), p is and (to the econometrician) Q p it (K L i; it , M it , Tp p , £ it )-exp(£ it A ))] probable output; inputs are denoted by E for energy (fuel) M for materials; T denotes parameter vectors, and z denotes mean zero shocks. See Van Biesebroeck (2003) for the derivation of a similar production function he uses to model automobile assembly plant production. 12 Random shocks to a plant's operations, such as unexpected equipment failures or equipment that longer than expected, will cause inputs. it to produce less or more than its lasts probable output from a set of available ) As noted above, made fuel input decisions are in real time, after the shocks associated with the plant's probable output productivity, plant, A £it output, , Q and the plant's energy-specific productivity p , is in P Ej t , manager has observed any the actual operation of the in the current period, £ it contrast determined by input decisions made in productivity of the plant's capital is in . Probable advance of actual production. We assume that capital, measured by the nameplate generating capacity Labor and materials decisions are made E advance of production, but of the plant, after the level it is dispatched, based on the targeted output productivity shocks that we assume known to are Q p . The the plant error term £ it manager in p from probable output by a multiplicative shock exp(£ made time fuel input choices are known but not shock would be, for example, negative if at the failure, or positive if the plant the case if a number of plants ahead of it realizations were unexpectedly high. (PF 1 P L Q < Qo(Ki)-(L y -(M it lt In preliminary analysis, we in the usual dispatch )™- exp( £l to be is and material prices on each measured S; t , at the determined. This down due anticipated, as to a might be demand order were unavailable or yields the specification: p ) Those results suggested The work reported here imposes an on cost-minimization, to estimate input demand functions, and restructuring effects observed estimated the parameters of the production function, including terms productivity gains associated with restructuring. 13 assumed function of labor and materials and (Q (K)) term. This it available We allow actual output time probable output that allowed for differential productivity under restructuring. constraint, based ), were run more intensively than We model probable output (Q p ) with a Cobb-Douglas capital effects in a constant A it is incorporates a generating unit were unexpectedly shut mechanical embedding and advance of scheduling labor and materials inputs, but are not observable to the econometrician. to differ fixed. observed. This reflects the quasi-fixity of these inputs over time: staffing decisions and maintenance plans are designed to ensure that the plant when 14 is would solve input. additional isolate possible A cost-minimizing plant manager, facing wages for the optimal inputs to produce probable output Qj P t W by: In fact, over a short time period, maintenance and repair expenditures will be inversely related to output since the boiler needs to be cool and the plant offline for most major work. We deal with this potential simultaneity bias below. 14 The empirical analysis defines a new plant-epoch, so that within each plant-epoch, capacity is /', whenever there approximately constant. are significant changes in capacity, lt W , W, ,-Lj + min Lj„ Si , M t -Mi demand (LI) Lit = (X,YL (Ml) Mi^^Q/yS;, = oo + ln(Li t) = oto L ln(X.y A it = A P Qi e t it restructuring, ln(L L is we expand irt ) ) t - A £i t - ln( measurement error L it ) A irt )-ln(W irt ) measures the remaining error L ; . Note that L <p r L cp r in labor used many this at the plant, this i equation will hold demand associated with year and regulatory in /, as: + aL + i 5t L +cp r L - in the labor input equation. picks up mean in the of the availability constraint (X) or that Making . A Eirt + E irt L 5, L captures year-specific captures restructuring -specific shifts in labor demand, and could reflect systematic changes Note p it I5 a is now subsumed L Ei rt in the plant-specific residual changes in labor input for a plant in a restructured regime relative to that plant overall and to 15 Q shadow value of the probable output or in the to irt to include plant and re-write (L2) r, = ln(Q the subscript differences in labor demand, It ) over time, or across regulatory regimes measures a plant-specific component of labor demand, demand, a p t ) ; , are particularly interested in changes in input restructuring regime a; O^-expfe becomes: of the production function (y As we with error. A ln(Qi constraint (X), or if there where -(Mj rather than probable output, , If there are differences across plants, ). in the coefficients (L2') ) equations: substitution and taking logs of both sides, (LI) where YL t P Qit )/Wit We observe actual output, Q (L2) < Qo(K,)-(Li the Lagrangian on the production constraint. is X, p t it yielding the following factor where & s.t. all other plants at the marginal productivity of labor (y in optimization errors. plant-level differences, such as capital stock, L ), same point in the in time. shadow value I6 and many time-varying shocks, such as technology-neutral productivity shocks, drop out of this equation through the conditioning on output choice. 16 If there may were systematic differences also reflect the change mean £ i rt . seems plausible that these be mean zero the restructuring sample we observe. setting planned output, could be nonzero in in of probable and actual output across restructuring, y Since e ^ reflects shocks unobservable by the firm when in the relation it in expectation, L r but their realizations 10 (Ml) becomes: Similarly, equation ln(M (M2) which is with equality, Q Assuming in ) - ln(S rt ; t q> r component of the Leontief production A M - £ A irt + £ „ M t function, which will in general hold = g(E ,YE,£ E ) irt that g(«) irt is monotonically increasing in E, we can simply invert it to get an expression terms of Q. Note that the price of fuel does not enter into the demand for fuel except through the level of output the plant input specifications, (El) where + aM + 5M + ) as: (PF2) E A rt directly analogous to (L2'). We model the energy for = ln(Q irt ) we InCEirt) dispatched to produce. For consistency with the other is specify a log-log relationship: = yQ E -ln(Q irt ) + E cp r as before, the plant-specific error, Oj specific term, E cp r , energy to electricity E + E a, + 5 E + E £irt the year-specific error, 5 t E , and the restructuring- capture systematic changes in the efficiency with which plants convert — that is, changes in plant heat rates — across plants, over time, or correlated with restructuring activity, respectively. We confront two (L2'), (M2) and equations may important endogeneity concerns in estimating the basic input (El). The first is the possibility that shocks (£ L irt £ in M E , demand £in ) in equations, the input demand be correlated with output. If output decisions are made after a plant's manager observes the plant's efficiency, managers may increase planned output in response to positive shocks to an input's productivity, or reduce planned output in response to negative shocks. This behavior would induce a correlation between the error in the input demand equation and observed output. Though one can control directly for plant-specific efficiency differences and for secular productivity shocks in a given year, idiosyncratic shocks remain a source of possible bias. Second, the estimates may be subject to selection bias if exit decisions are driven by unobserved productivity shocks. In this case, negative shocks could lead to plant shutdown, implying that the 11 errors for observations problems is Consider first we unique to our observe will be drawn from a truncated distribution. Neither of these setting, and they have been raised in many earlier papers. We face a potential simultaneity problem the simultaneity issue. for instance, a if, malfunctioning piece of equipment reduces the plant's fuel efficiency, leading the reduce its operation of that plant and consequently to use less fuel. There may be 17 utility to deviations from predetermined employment and materials budgets caused by unanticipated breakdowns that require increased use of labor and repair expenditures and result in lower output. efficiency shock to an input may lead managers to run that plant concern. electricity the We choose to use an instrumental variables approach, using a measure of state-level demand as an instrument for plant output. This amount of output a plant efficiently intensively over the year, A variety of methods have been used to address this increasing output as well as input use. 18 more will be called is likely to be highly correlated with to provide, but uncorrelated, for instance, an individual plant's feedwater pumps are working. This approach particularly effective for the energy equation, given the responsiveness demand A positive fluctuations in real time, and for identifying is likely to powerful in identifying variation in The of our exogenous output fluctuations at state demand. 19 stable than those studied in is more many difficult to address. non- It may be in part We therefore explore the The plants in our sample seem more other contexts (especially see Olley and Pakes, 1996), suggesting that the selection problem may be somewhat less severe for electric generation. exit increases in restructuring regimes, typically not because the plant but because divestitures remove the plants from the reporting database. divestitures are to results to alternative instruments. potential selection issue However, plant be ex ante labor and maintenance choices, depending on the extent to which plant managers anticipate sensitivity how of energy input choices baseload plants, which are more strongly influenced by marginal swings in demand. less with mandated by the restructuring legislation, this To is retired the extent that the should not create selection problems. But without better information on what determines discretionary divestitures, we have 17 Nerlove (1963) provides an early discussion of simultaneity bias in production functions. Olley and Pakes (1996) propose a structural approach to addressing simultaneity, which is compared to alternatives Griliches and Mairesse (1998). Ackerberg and Caves (2003) discuss this issue and compare treatments in proposed by Olley and Pakes (1996) and Levinsohn and Petrin (2003). While many papers have estimated production or cost functions for electric generating plants, from the classic analyses in Nerlove (1963) and Christensen and Greene (1976) to very recent work such as Kleit and Terrell (2001) and Knittel (2002), electricity industry studies typically 18 19 See the references cited A in have not treated either simultaneity or selection problems. note 17, supra. drawback to this instrument is that we measure demand only at the state, rather than plant which means that we do not use the cross-plant variation within a state to identify output coefficients. further level, 12 no direct way to assess their potential selection effects work for to those for a panel which is impact on the to compare results. results for the of plants that continue may indirect way to assess the significance of unbalanced panel to operate potential selection effects are likely to be those results One we use in most of our through the end of our sample period, most severe. Substantial differences across suggest the need to more carefully treat potential selection biases. Identification strategy There is substantial heterogeneity across plants, utilities and states, operate has changed considerably over time. In addition, environment in which restructuring is not randomly assigned across political jurisdictions is utilities and the economic — earlier work suggests strongly correlated with higher than average electricity prices in the cross-section. Fortunately, we have plants operated in this sector own and legal it 20 a database rich in variation. There are thousands of generating by hundreds of utilities subject each setting their that to regulation institutional environment. by dozens of political jurisdictions Panel data on the costs and operations of these plants are available, with some recent exceptions, from well before any restructuring until the present. 21 This allows us to construct benchmarks that we believe control for most of the potentially confounding variation. The plant-specific effects, plants in the sample. N {oii These }, measure the mean use of input effects may be and vintage, ownership (government year-specific shock, {5 N t }, secular technology trends, effects v. private utilities), macroeconomic was initiated in 1993 and the outcome of those hearings, and time-invariant N {(p r }. The state effects. shifts The over time, such as The heterogeneity activity allow the data to distinguish all states latest in in the timing and between temporal held hearings on possible restructuring, the 1998. There is considerable variation in the as well, with just under half the jurisdictions (23 states and the District of Columbia) enacting restructuring legislation between 1996 and 2000. 20 relative to other fluctuations or energy price shocks. Restructuring on plant productivity correspond to a non-zero shocks and restructuring effects. While ; associated with differences in plant technology type measures the efficiency impact of sector-level outcomes of state-level restructuring earliest N at plant significant role of sunk capital costs in regulatory ratemaking means 22 The remainder that high prices necessarily imply high operating costs for generation facilities within a state, however. See do not Joskow (1997) for a discussion of the contributors to price variation across states. Cost data are not publicly available for plants owned by exempt wholesale generators, including those acquired from regulated 22 We collected utilities. information on state restructuring legislation from various Energy Information Administration and National Association of Regulatory Utility Commissioners publications and state 13 : considered and rejected, or considered and simply did not act on, such legislation. This variation allows us to use changes in efficiency at plants in states that did not pass restructuring legislation to identify restructuring separately from secular changes in efficiency of generation plants over time. It is possible that plants in this control group also altered their behavior over the post- 1 992 period. This could be due perhaps to the introduction or intensification of incentive regulation within states that did not enact restructuring, to the expectation of potential restructuring that did not occur, or spillovers from restructuring their information movements in other states (e.g. if regulators updated about the costs necessary to run plants of a certain type, or multi-state utilities operating under differing regimes improved efficiency of all their plants, not just those in restructuring states). To the extent this occurs, our comparison will understate the magnitude of any efficiency effect of restructuring. We therefore consider a second control group, consisting of cooperatively-owned or publiclyowned municipal and federal plants, although the group broader than is which strictly for convenience we will refer to as An implied by this label. "MUNI" plants, extensive literature has debated the relative efficiencies of private and public ownership in this sector under traditional regulation, with somewhat mixed effects that absorb results. any levels differences We abstract from that by allowing for plant-specific in input use across ownership type. Restructuring generally altered the competitive environment only for private investor-owned state, that leaving those for publicly- and cooperatively-owned utilities unchanged. MUNIs may We adopt a parameterization that measures publicly-owned plants during the period that investor-owned 1 993 forward. Using This suggests equation 1 1 irt - } relative to of restructuring, N to denote input (labor, nonfuel expenses, or fuel), and to denote the relevant input price (none for the fuel equation), A ln(N irt ) = ln(Q ) N {cp r utilities are at risk PRICEN (1 1 ) within a provide a second benchmark against which to measure changes in efficiency associated with restructuring. defined as utilities 23 N \n(PRICE in ) + yMUNI + it N a, + 5t N + we have N <p ir - e A irt input use + e N , r t utility commission websites. Since 2000, no additional states have enacted restructuring legislation, and several have delayed or suspended restructuring activity in response to the California crisis. public With the exception of Arizona and Arkansas, which included government-owned utilities in restructuring programs. 14 This specification implicitly provides two "non-treatment groups" to which investor-owned plants in restructuring regimes may be compared: investor-owned plants in non-restructuring regimes (with the restructuring effect measured by N cp plants over 1993-1999 (with the restructuring effect Data 3. The r ), and public- and cooperatively-owned measured by N - y). cp r & Summary Statistics analysis in this paper is based on annual generating plant-level data for U.S. electric Plants are comprised of at least one, but typically several, generating units, to or retired from service over the several-decade would allow ideal data set life utilities. which may be added of a typical generating plant. While an us to explore efficiency at the generating unit level, inputs other than fuel are not available at the generating unit level, and some, such as employees, are not even assigned to the unit level as they are shared across units at the plant. 24 We therefore use a plant- year as an observation. The Federal Energy Regulatory Commission (FERC) plants annually in the Utilities Service FERC Form (RUS) 1 , collects data for investor-owned utility and the Energy Information Administration (EI A) and Rural collect similar data for municipally-owned plants and rural electric cooperatives, respectively. These data include operating statistics such as size of the plant, fuel usage, percentage ownership held by the operator and other owners, capacity factor, operating expense, year built, and set includes all large data were reported to other plant-level statistics. (CCGT) steam and combined cycle gas turbine FERC or EIA plants, defined as those for our sample years. many number of employees, generating plants for which over the 1981 through 1999 period. which gross capacity exceeded We also excluded approximately 25 undue influence. Further We follow the year, 1,500 observations where data were missing, on the data are provided literature in characterizing output Some labor may be is, in reality, One utility total screen for Appendix. energy output of the plant over the NET MWhs. This is shared across multiple plants, though assigned to one particular plant in our data. unfortunate consequence of restructuring is in the plant employment that available data variable. on plants sold by utilities to non- generators are extremely limited after the sale, due to changed reporting requirements. This that plants will an multidimensional, although most dimensions are not This will lead induce measurement error, particularly 25 by the tests to in the measured by annual net megawatt-hours of electricity generation, imperfect choice. Output 24 details We excluded smaller 00 megawatts for fewer than three of 1 and dropped several hundred observations based on regression diagnostic outliers or Our base data be excluded from the dataset after such means sales. 15 recorded in the plant data. For example, generating plants (such as spinning reserves, when may also provide reliability services the plant stands ready to increase output at short notice), voltage support and frequency control. While the production process varies considerably across these different outputs, only net generation is not a the homogenous product. Because Friday in July first Sunday down is in is well measured in the data. 26 Moreover, electricity is non-storable, electricity how produced 5PM on with taking their plant to balance the costs associated do maintenance against the probability that a poorly maintained plant will to at 5AM on the second a separate output from electricity produced at March. Firms must decide electricity output during fail peak demand hours, and the availability of the plant may be an important modifier of output Changes quality. in incentives associated with restructuring may have of these tradeoffs, although the expected direction of the effects is altered firms' assessments theoretically ambiguous. Hourly output prices and output from individual plants might allow us to better assess Lacking such data, We have we rely on a single output dimension, but acknowledge information on three variable inputs. The employees The second, at the plant. first, EMPLOYEES, NONFUEL EXPENSE, includes is all its 27 this. limitations. a count of full-time non-fuel operations and maintenance expenses, such as expenses for coolants, maintenance supervision and engineering expenses. This variable less is than ideal as a measure of materials, both because expenditures rather than quantities, and because counted in EMPLOYEES, although that expense NONFUEL EXPENSES includes EMPLOYEES will coal, barrels reflect changes in Btu content of each Input prices pose a challenge. includes the wage bill for the reflects employees not separately delineated in our data. is As payroll costs (not separately identified), both this and 28 staffing. of oil, and mcf of natural gas). plant-specific it it The We convert fuel into BTUs fuel to obtain total Wages perceived regulatory environment. third input is fuel use in particular by type of fuel (tons of using the reported annual BTU input at the plant for each year. may be endogenous to the firm and Hendricks (1975) suggests that utilities may its bargain less 26 The inputs required to produce a given level of energy (MWh) from a specific plant also will depend on whether the plant runs continuously or intermittently and on its average capacity utilization. Starting a plant frequently and running per 27 mwh generated than does it at low capacity For instance, under traditional regulation, utilities blackouts or brownouts, leading to investment maintenance resources to increase plant restructured wholesale markets because 28 The this is when elasticity may in more inputs (particularly fuel) at its rated capacity. may have faced strong political incentives to avoid greater capacity to increase reserve margins and in greater reliability. On the other hand, competitive firms producing in face even stronger incentives to be available when demand peaks prices are highest. of NONFUEL consistent with our back total utilization rates typically use running a plant continuously EXPENSES with respect to EMPLOYEES is of the envelope calculations suggesting about .5 in our data, broadly that labor costs are roughly half of the nonfuel operating budget. 16 aggressively over input prices such as wages during periods in which higher costs could be readily passed firm was on customers through higher regulated prices, and more aggressively when the to likely to be the residual claimant to cost savings. In other industries, regulatory reform has sometimes been associated with substantial reductions in wages, suggesting rent-sharing under regulation (see Rose, may 1 not be exogenous to the firm, and workers of similar may not reflect the opportunity cost to managers of the We address this by using state-level marginal unit of labor. skills These suggest that observed wages 987, on the trucking industry). average wages from industries with and training to power plant operators, including natural gas petroleum refining and hazardous waste treatment facilities, denoted as WAGE. distribution, This reflects opportunity wages, and avoids confounding the employee price measurement with any changes in recorded wages due to changes in labor force composition or changes in wage bargaining. materials prices that comprise EXPENSES therefore We do not have plant- or even firm-specific indices NONFUEL EXPENSES. final input is the capital stock Our data record the plant capacity in megawatts. Our empirical model of NONFUEL capacity of the plant we retirement, is create a changes we measure by of the plant, which retirements to define plant-epochs. Each plant is identifier for in input use. two other The is and associated new plant-specific plant may vary within plant-epoch and lead may become The second is FGD affects the environmental output, unmeasured by state, we have less efficient or the addition of a flue-gas in some \n{NET MWhs). plant data with information on state-level restructuring activity. identified the date at which formal hearings on restructuring began, the enactment date for legislation restructuring the state's utility sector, if any, the date for retail access under that legislation, and some rate freezes This allows FGD (also called scrubber), to reduce sulfur-dioxide emissions We supplement the operational For each effect. AGE in years, dated from the installation date of the require additional inputs for a given level of output. coal plants. time the of the plant. oldest operating generating unit at the plant: as plants age they desulfurization system, or Any an identifiable unit addition or unit plant characteristics that first is information on unit assigned a unique identifier. significantly changed, or there new plant capacity and vintage. We combine this with to alter the underlying input efficiency We include controls changes for the corresponds to an input demand equation with constant prices and a price The to associated with restructuring of one. coefficient capital at utilities implementation associated aspects of restructuring such as and mandatory divestiture of generation. Testing for restructuring-specific shocks 17 requires a determination of how to match this when were information with firm decisions: plant operators in a given state likely to have begun responding to a policy change? Consultations with industry participants and readings of these events suggest that utilities often acted in advance of final outcomes. The legislative and regulatory process leading up to state restructuring typically lasted a number of years, allowing utilities to anticipate the coming change, and alter their behavior in advance. For example, Boston Edison's 10-K filed in March 1994 discussed Massachusetts' consideration of restructuring, stating "The Company is responding to the current and anticipated competitive pressure with a commitment to cost control and increased operating efficiency without sacrificing quality of service or profitability" (p. 6). begun to phase in input 29 Utilities changes, especially those involving labor and particularly unionized workers. Moreover, as policy changes were discussed, rates were frozen in explicitly enabling by policy makers or utilities to In this work, restructuring. turns we 31 in effect by implicit PUC decisions not to hear capture the savings from incremental cost reductions. The primary start RESTRUCTURED, variable of interest, of formal proceedings is A second variable, RETAIL ACCESS, number of plants PECO's CEO Joseph Paquette discussed have been focusing on our all PECO operations. an indicator variable that indicates the start of retail access for plants in 32 Table 1 reports the cost saving accomplishments and strategies for the future, restructuring of the utility industry and strategic plans to becoming more competitive." effectiveness of rate cases, 30 each year that were in states that had RESTRUCTURED and the in our database In a 1993 article outlining and new states, either a state that eventually passed restructuring in the four states that implemented retail competition during the sample. 29 many allow restructuring effects to begin with the opening of formal hearings on on with the legislation. may have One enhance our abilities to satisfy initiatives cited in the article included particular was quoted Chairman as stating, "we our customer needs by improving the cost accomplishment noted was the reduction in total employment from 18,700 to 12,900. 30 As noted earlier, some of these changes may have also affected utilities in non-restructuring states. For example, the number of utility rate cases dropped dramatically in the 1990s, implying that many or most utilities may have been short- or medium-run residual claimants to cost reductions. Knittel (2002) identifies a number of incentive regulations adopted in various jurisdictions during the 1990s. Many of the fuel-related regulations (modified pass-through clauses, heat rate and equivalent availability factor incentive programs) were strongly correlated with ultimate restructuring. (e.g., price 31 The RESTRUCTURED variable is some delays or suspension of planned While regulations based on whether a state had passed legislation as of mid-2001, although in the aftermath of the California electricity 32 Some of the broader caps and revenue decoupling programs) were almost orthogonal to eventual restructuring. RESTRUCTURED indicates approval of retail access legislation, the specified phase-in of retail access was often slow. Only five states implemented in 1997, California, Massachusetts, there has been no additional restructuring, and crisis, restructuring activity. and New York in retail access during our sample period: Rhode Island 1998, and Pennsylvania in 1999 (U.S. EIA, 2003). Because we have no valid observations on Rhode Island plants in 1997 or beyond, retail access effects will be determined by the 4 states implementing in 1998 or 1999. Divestiture requirements in California and Massachusetts further reduces the post-retail access sample of plants, as investor-owned plants in those states were largely divested by 1999. 18 number of plants in states that had started restructuring legislation or regulation retail access by 1999. 33 was enacted and the policy uncertainty RESTRUCTURED will underestimate the true effect by averaging evaluate this possibility beginning in the we If utilities did not introduce a third variable, in 34 Similarly if actual will be reflected in an incremental effect of RETAIL municipally-owned plants over the restructuring time period, 1992, equal to one for all we 1999. The final column is important to ACCESS. To examine define the variable municipally-owned plants from 1993, the RESTRUCTURED is one, through resolved, non-response years. To implementation of retail access and the associated wholesale market reforms it until LAW PASSED, an indicator equal to one year the state passes restructuring legislation. efficiency gains, respond first reports the MUNI*POST year for which number of plants in this category. Details on the data sources and 2b report summary summary statistics are statistics for plant-level provided in the appendix. Tables 2a and data in 1985 across three categories: investor-owned plants in states that later restructure, investor-owned plants in states that do not restructure, non-IOU plants. changes across initiatives. We choose this date to ensure that comparisons are made prior to any significant states in the competitive or regulatory From and these tables, it from the same population. The environment, even prior to restructuring appears that the plants from these groups are not random draws first three variables measure employees and non-fuel operating expenses, scaled by the plant's capacity, and fuel use in millions of British thermal units (mmBtus), scaled by the plant's output. In 1985, before state-level restructuring initiatives were considered, plants in states that eventually restructured had higher intensities of employees and non-fuel operating expenses, although the difference The first two rows in is significant only for non-fuel expenses. Table 2b show that municipally-owned plants had significantly higher employment and non-fuel input use than plants in non-restructuring states. heat rates and capacity factors are not significant in either of the tables. The The last differences in four variables in both tables describe the stock of plants in the two types of states. Although plants are very similar in size across IOUs, 33 MUNIs plants are considerably smaller. IOU plants in restructuring For the tables and the regression analysis that follows, plants are assigned to the state in which they are A plant located in one state may be owned by a company with exclusive service territory in a regulated. and that second state is the state by which the regulatory policy is measured. Some plants company with service territory in more than one state and some plants are owned by several companies that are regulated by different states. In the regression analysis, we found that separately different state, are owned by a characterizing "mixed" regulation and "shared" plants had very little impact on our results. There is on average about a 2.6-year lag between the initiation of hearings and the passage of the law. We have experimented with a number of alternative measures of restructuring activity, including variables 34 that begin with hearings regardless of restructuring outcomes, those that measure years since hearings were 19 states tended to be older, more likely to use gas, and less likely to use coal, than non-restructuring states. likely to use gas, than MUNI plants tended to be younger, IOU plants in non-restructuring states. less likely to IOU plants in use coal, and more The regression analysis will control for these differences directly or with the use of plant-epoch effects. If investor-owned utilities achieved efficiency improvements restructuring of the generation sector, one generation for affected companies, and distribution costs would expect little when facing impending to see a relative decrease in the cost of difference in the change in transmission and between the affected and not affected states since restructuring programs leave transmission and distribution comparatively untouched. If restructuring did not affect operating efficiency in the generation sector, we might would not be between restructuring and non-restructuring companies, or (2) we would statistically different same see the expect either (1) the change in generation expenses pattern of change in costs for the transmission and distribution sectors as for the generation sector. 35 Table 3a and 3b display the difference in mean and non-restructuring states for a change tests for in costs investor-owned utilities in restructuring between 1990 and 1996. Table 3a reports the percentage change in total costs for each category of cost, and Table 3b reports the percentage change in costs per significant at the level. The MWh. The larger decrease in generation costs 1% level, and the result for generation costs per at restructuring MWh is companies significant at the difference in costs for companies in restructuring and non-restructuring states significant for either the transmission or distribution costs. These aggregate preliminary support for the expectation that the portion of the utility statistics is is 6% not provide company faced with competition (the generating sector) responded with a decrease in costs, while other sectors and companies not faced with competition did not share 4. The Effects of Restructuring on Input this response. Use and the presence of restructuring-associated rate freezes. change materially to the conclusions we draw below. 35 For the analysis comparing costs of generation, transmission, and distribution services, we rely on data reported annually by utility companies to the Federal Energy Regulatory Commission (FERC) in the FERC Form 1, page 320, 321, and 322 respectively. We use a balanced sample composed of all companies with data reported for all three sectors in both 1990 and 1996. This amounts to 49 companies in states that did not deregulate and 74 in states that did deregulate for the comparison of costs, and 48 and 72 respectively for the comparison of costs per MWh. Using costs per necessitates the exclusion of a few companies for which data was not available in one of the two years. initiated for states that eventually restructured, None of these seem to MWh MWh 20 Following equation (II), we (EMPLOYEES, NONFUEL EXPENSE, and BTUs) ln(N irt ) = (Rl ) with the following basic regression model: n ^ \n(NETMWh M) + tfhiQPRICE"* ) + &A GE N IOU*RESTRUCTURED + yMUNI1993 m9 + where we allow for non-unity coefficients on the output term input price term (p 2 Oj . r N on WAGE) in the EMPLOYEES important plant characteristics that vary over time: invariant fixed effect for input N at plant-epoch i, N ) (Pi equation, 36 k + ilt N (p N estimate the influence of restructuring on the use of input FGD fa +5 N + t ilt + N p\ s A lrt + e N irt and on the for all equations and include controls for two AGE and FGD (scrubber). which may contain a N is otj a time- state-specific and ownership-specific error that will not be separately identified. These plant-specific effects control for much of the expected variation in input use across plants arising from heterogeneous They technologies, state or regional fixed factors, and basic efficiency differences. mix between for differences in the plant each plant to rejects the is itself restructuring and non-restructuring states over time, removing any time-invariant plant effects. exogeneity of plant effects, an industry-level effect in year t, all As a also control by comparing Hausman test reported results include plant-epoch fixed-effects. which controls for systematic changes in input demand 37 5 N t across plants over time. all N £irt time is t. assumed to be a time varying This shock is mean zero shock for input N at plant-epoch i in unlikely to be independent over time for a given plant. There regime is r at likely to be persistence in input shocks, particularly for labor, from year to year. Indeed, estimated rhos based on assumed first-order range for non-fuel expenses. serial correlation are in the .65 It is autogressive process, however. range for labor inputs and unlikely that the correlation The is in the .33 as simple as a first-order physical operation of power plants is correlation at longer differences. For example, routine maintenance cycles likely to induce may involve scheduled shutdowns but increased labor and nonfuel expenses every three or four years. have explored GLS specifications based some on an assumption of first-order autoregressive We errors, and the results are quite similar to those reported in the tables below. Rather than impose this assumption, however, 36 Recall that we do we choose to estimate the model without a GLS and simply not have a price associated with nonfuel expenses, and that according to equation (El), fuel prices should not enter into the fuel input function. We experimented with the price of a given plant's fuel relative to the prices of other fuels in the output but the variable had no 37 correction, This use of fixed effects is power using a variable measuring same region as an instrument for in the first stage. work of Joskow and Schmalensee (1987), who used generating between operating performance and unit characteristics. Our similar to the unit-level data to explore the relationship 21 report standard errors that are corrected for a general pattern of correlation over time within a given plant. 38 Input use over time will vary with the level of plant operation, which specifications as the net generation by the plant in megawatt-hours is measured in these (NET MWh). We treat the endogeneity and measurement problems described earlier by instrumenting for output with state demand (the log of total state electricity sales, a consumption rather than production measure). This instrument affects the likelihood that a given plant in the state will be dispatched more over the year, but not influenced by the characteristics of the plant or the choices of individual plant is operators. We consider specifications that include interactions of IOU ownership with the three primary restructuring indicator variables described in section 3: RETAIL ACCESS. and In the input regressions, a negative coefficient on the restructuring variables would imply increased input efficiency associated with the regulatory reform. for the input analysis are presented in Tables 4 for and 6 for BTU. RESTRUCTURED, LA W PASSED, We first discuss the results for EMPLOYEES, 5 for The core results NONFUEL EXPENSES employment and nonfuel expenses, and then discuss the results for fuel use. Column 1 of tables 4 and 5 reports a simple output. In this column, OLS formulation that excludes any control for IOU*RESTRUCTURED captures the mean differential in input use for investor-owned plants in states that eventually pass restructuring legislation, measured over the period following the first This corresponds to the restructuring hearings, relative to mean IOU plants in non-restructuring states. within-plant shift in input use, independent of output. The results suggest statistically and economically significant declines in inputs during restructuring. Employment relative to declines by almost IOU 6% (2%) and nonfuel expenses decline by almost 13% (2%), plants in regimes that have not restructured. Controlling for plant output reduces the estimated impact of restructuring by and statistically distinguishable employment and -10% (2%) more than one-quarter, though from zero (see column 2 of each for nonfuel expenses. fixed effects are actually finer than plant-level, as we 40 We use [exp(<p r N the effects remain large table), at -4% (2%) Measuring restructuring permit a, to and other significant changes in rated plant capacity. 38 Reported standard errors are calculated using the cluster option 39 39 effects for from the change with unit additions or retirements in Stata. )-l]*100 to approximate the implied percentage effect of IOU* RESTRUCTURED on input use. 40 Note that the Cobb-Douglas functional form assumption suggests that the coefficient on output should be one, substantially larger than the coefficients estimated in these regressions. If we impose this constraint, the effect of restructuring is estimated to be positive and significant. We have estimated production 22 enactment date of legislation does each little to change the qualitative conclusions (see column 3 in table.). This finding hints at an interesting pattern in the data: restructuring is associated with same-plant output reductions relative to same-plant output in non-restructured regimes is why at the same date. This estimated input use reductions associated with restructuring are smaller in the presence of output controls. Some of this might be restructuring, as the less in a raw input and expected: if these plants were less efficient prior to more competitive environment. exempt generators, this output reduction also consumption the case, they may be used encourages entry by non-utility or If restructuring might further reduce output may may have been price data suggest at less efficient plants. 41 Some plant-level be generated by the apparent lower growth in state-level electricity in restructuring states, relative to electricity consumption growth in non- 42 restructuring states. This result is not only of interest economically, but also estimation of the input demand relationship. The is important for the econometric inverse correlation of output and restructuring policy attaches great importance to obtaining a consistent estimate of the relationship between output and input demand. Understating the response of input use to output changes will load more of observed input reductions onto the restructuring variable, overstating the responsiveness will lead to underestimates of the restructuring The second notable is While group. pattern IOU the dependence of the implied restructuring effect on the control plants in restructuring states exhibit nonfuel expenses relative to than twice as large IOUs when compared MUNI*POST 1992 coefficients 21-23 percent (columns 2 and in modest reductions in employment and in non-restructuring states, the implied reductions are to public Employment drops by 13-15 percent functions in effect. more and cooperative plants over the restructuring period. (see the difference in the columns 2 and 3 in table 4) 3 in table 5), relative to IOU*RESTRUCTURED and and nonfuel expenses decline by municipal plants over the 1993-1999 EMPLOYEES and NON-FUEL EXPENSES using more flexible functional forms than Douglas, and the results also suggest efficiency gains associated with restructuring. We Cobb- have also estimated instrumental variables versions of equations (L2) and (M2) that include the other input instead of output and obtained very similar results to those reported here. 41 Total state-level electricity consumption, while increasing on average during the 1993-1999 period in restructuring states, appears to 42 This is grow less fast than consumption in non-restructuring states. unlikely to be causally related to restructuring, as restructuring effects state electricity electricity on retail rates were neutral however, a negative correlation between consumption and restructuring, conditioning on state means and national growth rates in or negative (due to rate freezes) during this period. There is, consumption. 23 period. This suggests that even IOU plants in non-restructuring regimes were reducing their input use to a significant extent during the 1993-1999 period, perhaps in response to latent threats of increased competition and restructuring. Given the importance of controlling We return to this issue in greater detail below. for output in assessing restructuring effects, we next turn to instrumental variables estimates of the input equations that treat potential measurement error and simultaneity bias with respect to output. Column 4 of these tables instrument for the log of current plant output with the log of total electricity sales in the state. on the estimated output coefficients for the two input demands. For 43 This has opposite effects EMPLOYEES, instrumental variables (IV) estimates of the output coefficient are below the although the imprecision of the IV estimates make consistent with there being a positive relationship although as we Both respect to output are quite small (at OLS estimates, difficult to reject equivalence. between shocks discuss below, the direction of this relationship specifications of the instrument. to efficiency 8.1% (1% standard is error) largely is is not robust to alternative is elasticity with and 6.5% (7.5%), respectively for Consistent with relatively fixed. estimated coefficient on IOU*RESTRUCTURED This and input use, and IV estimates of the labor demand the basic specification), suggesting that labor at it OLS the unchanged this, the in the basic specification, -4.5% (2.3%). In the case of nonfuel expenses, instrumenting for output substantially increases the estimated output elasticity, from 15% (1%) to 40% (9%) in the basic specification. negative correlations of input shocks and output, as for example, if This is consistent with large maintenance expenditures are associated with outages at the plant. Given the strong link between output and nonfuel expenses implied by these results, and the fact that both declined with restructuring, instrumenting also has a substantial effect on the estimated effect of restructuring. The estimated coefficient on IOU*RESTRUCTURED coefficient drops by more than range of the estimated labor input effect Columns 5 This is SALES) Table 44 it into the about -4.8% (2.8%). ^ Measuring restructuring by an important determinant of plant-level output. is bringing and 6 of the tables explore robustness to alternative measures of restructuring, maintaining the use of IV estimates. 43 at half, zero in the first stage for Column (4) is The F-statistic LAW PASSED from the test that in column 5 \n(STATE F(l,9674)=18.13 for Table 4 and F(l,9686)=16.68 for 5). Results are broadly similar in unreported regressions that measure restructuring from the date of legislation, although the effect number of years since hearings of RETAIL ACCESS is further tempered, and in regressions that add the were begun, which suggest that employment and nonfuel expense reductions begin relatively small and increase over time. 24 yields slightly larger coefficients on restructuring in the employment regressions and slightly smaller (and statistically indistinguishable from zero) coefficients in the nonfuel expense Column regressions. identified 1999, and RETAIL ACCESS variable. We by no more than two years of data it is note that in the four states that -5% (5% labor demand is coefficient its implement not particularly stable across alternative instrument sets. RETAIL ACCESS in additional 6 adds the The is access as of retail coefficient on imprecisely estimated, though the point estimate suggests an when standard error) change in employment The estimated impact of retail access on nonfuel expenses is states implement retail access. -15% substantially larger, at (7.8%). In the instrumental variables results, as in the statistically IOU plant larger, restructuring and economically in a restructured on the order of 1 results, the upon the chosen benchmark or restructuring effect depends demand between OLS and non-restructuring regime relative to in control group. states, conditional The gap on output, in is IOU input generally though relatively modest. The performance gain of an significant, 5% reductions implied magnitude of the MUNI plants over the same period is employees and 20% substantially reductions in nonfuel expenses. We therefore experimented with a number of more flexible specifications of the MUNI control group. Column 7 of tables MUNI plants over the 4 and 5 reports one variant, in which 1993-1999 period and (MUNI*RESTRUCTURED). MUNI In neither case can independent of restructuring (point estimates for we we include separate controls for plants in restructuring regimes reject the null that MUNI plant input use is MUNPRESTRUCTURED are positive for employees and negative for nonfuel expenses). In unreported specifications we have allowed for more flexible specifications coefficients. The of the MUNI controls over time and for differential MUNI output pattern of results is consistent with input performance gains for restructured regimes relative to nonrestructured IOU plants, IOU plants in and even greater gains when measured against MUNI performance. To provide further insight into the question of column (4) without the of benchmark group, we re-estimate the basic IOU*RESTRUCTURED and MUNI*POST 1992 variables, but allowing for separate year effects for each of three categories of plants: eventually restructure, model IOU plants in states that do not restructure, IOU plants in states that and MUNI plants. Figures 1 (employees) and 2 (nonfuel expenses) plot the estimated year effects for each plant group. Figure 1 suggests that period. The MUNIs use more labor than do IOU plants, figures suggest greater divergence measures as the 1990s progress. As this is a all else equal, throughout the between MUNI and IOU plants in sample both input period of increasing competitive pressures and 25 substantial movement toward information in the restructuring, these patterns suggest to us that there is considerable MUNI benchmark comparisons. Table 6 reports results from variants of our basic specification for fuel inputs. In column output coefficient is constrained to unity (effectively estimating IOU the right-hand side variables). In this specification, IOU plants use slightly more fuel than effect is ln(BTU/MWh) the 1, as a function of plants in restructured regimes appear to in nonrestructured regimes, all else equal, Column 2 imprecisely estimated (1.2%, standard error 0.8%). reports though the OLS results relaxing the output coefficient constraint, which generates a lower estimated output elasticity standard error regimes (-1 .1), and an implied reduction in fuel use associated with IOU plants in restructuring .2%, again relatively imprecise with a standard error 0.7%). The remaining columns report results instrumenting for output with state electricity consumption. in all four (.9, — columns are very similar fuel use is The qualitative results very highly correlated with the megawatt-hours a plant produces, with an estimated output elasticity that is slightly below one though statistically indistinguishable from unity. This could reflect slight increasing returns to scale with respect to fuel, consistent with the well-understood engineering relation between more efficient fuel burn and higher levels of output at a given plant, though the data are not precise enough to establish this. In all and on IV specifications, the coefficient estimates for statistically indistinguishable IOU fuel efficiency relative to RETAIL ACCESS in column operating under retail access fuel use in restructured IOU*RESTRUCTURED are small, positive, from zero. These provide no evidence of restructuring effects plants in non-restructuring states. Indeed, the coefficient 5 implies 5% (3%) higher fuel —an enormous use, all on else equal, for plants increase in economic terms. While somewhat higher wholesale markets could be consistent with more startup and ramping costs as plants adjust their output to follow hourly changes in marginal revenue, the estimated magnitude of this effect is The IOU difference between the other inputs. The too large to be explained even by this behavior. and MUNI plants is more difficult to pin MUNI coefficient point estimates are slightly distinguished from zero, are not generally distinguishable from the estimates. We find little evidence of gains though a caveat is in order. in fuel efficiency down larger, and though they can be IOU*RESTRUCTURED point over our horizon from restructuring, While variations on the order of even 0.5%- 1% are economically significant, it may be difficult to aggregated data. Because fuel efficiency at a plant for fuel use than for in fuel productivity measure these sufficiently precisely with our is heavily influenced by factors such as the 26 allocation of output across units at a plant, the and for number of times units are stopped and started, long the units were running below their capacity, our inability to measure or control how for possible changes in these operational characteristics capture fuel efficiency changes. Sample its may make it particularly difficult to We continue to explore this aspect of utility input choice. selection As noted earlier, our results could be affected by sample ownership transfers that eliminated data reporting. attrition Plant sales are particularly an issue in retail access states, as exempt wholesale generators (EWGs) purchased required to divest during restructuring. Since EWGs companies are not required sense, these due to plant retirements and plants that incumbents are not regulated to report operating data by FERC were in the traditional on their plants, and hence drop out of the database. If the plants dropping out of the sample had been systematically higher or lower in efficiency relative to their plant-epoch effect contaminate the restructuring estimates. for purchase were very efficient One could and to surviving construct plausible stories that plants selected and thus good acquisition the database due to retirements or sales that offered the substantial gain if efficiency could be improved. The targets, or that inefficient plants left new owners the opportunity for correlation coefficient of the last year of data on the plant with the efficiency measures used in the regression MW, -0.05 employees per plants, selection bias could MW), for non-fuel expenses per is negative (-0.17 for suggesting that less efficient plants tend to drop out of the database, either because they are sold or shutdown. of attrition, we 1 999. The in each case restricting the sample to plants that were in the dataset results for this The one exception is statistically is the coefficient on IOU*RETAIL, which indistinguishable from zero. This suggesting that utilities may in the last which could be picked up with the coefficient simply due to the difficulty of estimating the is not significant in our declines in magnitude and be consistent with anecdotal evidence deferred maintenance on plants they Testing robustness of the RETAIL sample (available from the authors) are for the most part very similar to the results for the full sample, suggesting that selection bias results. assess the impact estimated both basic IV specifications and specifications that included the ACCESS variable, year, To knew they were going on IOU*RETAIL until we RETAIL ACCESS effect to divest, limit the sample, or with so 45 little data. RESTRUCTURED effect 45 Imposing this selection rule eliminates, for example, almost due to required divestitures before 1999. all Massachusetts and California IOU plants 27 We next consider the robustness of our results input demand to a variety of alternative specifications of the equations. Given the null results in our basic fuel use regressions, we focus on labor and nonfuel expense input choices in this analysis. In Table 7, we indicator that turns on in 1993 if the plant utility generation as of 1993 to higher intensity is (HIGHNUGS). above median penetration of non- in a state that has This measure should capture any to the policy variable in columns 2 and in addition to 4, which include larger than the estimated effect of larger in column 2, RESTRUCTURED 3% IOU 3, responses We and as a complement of RESTRUCTURED and estimated impact of plants (columns 1) -6% (3%). for restructuring, at standard error). are close substitutes, however, as the presence of RESTRUCTURED (with a positive though and BOTH. The in table 4, at which includes controls estimated direct effect of restructuring (-6%, mechanisms at 1 direct effects an interaction of these variables, high levels of non-utility generation on employment somewhat utility of actual generation competition from unregulated market participants. consider this as a replacement for the policy variables in columns HIGHNUGS on an report results that use an alternative measure of competitive pressure, based It is negative and The effect -8% (3%), as is is the appears that these BOTH largely offsets the effect of insignificant effect estimated at 6%, standard error NUG penetration appears to have no detectable direct effect on IOU plant input use, and the results in column 4 suggest that controlling for NUG penetration 3%). For nonfuel expenses, high eliminates the direct measured restructured effect: Only IOU plants in regimes with both high NUG penetration and restructuring reduce nonfuel expenses relative to other IOU plants (see the coefficient on BOTH MUNIS and IOUS is in column 4). In all four specifications, the estimated difference around 15%, very similar to this difference in tables In tables 8 (employees) and 9 (nonfuel expenses), large v. small plants (columns 1 and 2), and old v. we IOU new plants (columns 3 and 4). Recall that IOU plants. plants in restructured regimes exhibit lower input use than nonrestructured regimes (see the coefficients on 5. explore results for two sampling strata: these are dimensions on which municipal plants appear to differ from specifications, 4 and between IOU*RESTRUCTURED), For all do IOU plants in though the magnitude of the estimated effect varies with the subsample. Employee reductions appear to be largest at small and old plants; nonfuel expense reductions appear to be largest at large plants, and similar across old and coefficient IOUs new plants. More interesting, perhaps, is the comparison to MUNI plants. The on MUNPPOST 1992 measures the difference between publicly-owned plants and in nonrestructuring positive, consistent with regimes over the 1993-1999 period. MUNI plants using more inputs than While the point estimates remain IOUs during this time, the relative performance plants (see at ofMUNIs column old plants. 3 of both Changes They appear differs across subsamples. tables), and are indistinguishable from IOUs employment in in general less inefficient for old at in employment changes MUNI plants are more in line with IOUs and for nonfuel expenses, the difference does not appear to vary with the size of the these differences are of some interest, with respect to the In table 1 0, we it small plants, at While plant. does not appear that the conclusions of the earlier tables MUNI benchmark are substantially affected by these sample plant differences. consider whether the results are explained by a regression to the phenomenon among IOU plants: is the gain in efficiency among mean plants in restructuring regimes because they had on average low productivity draws prior to restructuring, and simply return to mean investigate the extent to low cost losses at plants. mean this, we which efficiency gains To input use from a regression the To examine efficiency over time? at the higher cost plants are offset by efficiency on data for the pre-restructuring period, 1981 "expensive" and those below zero as "cheap." indicators with the restructuring variables, more IOU detailed restructuring variables. plants, at and residuals 1992. We calculate above zero as We then interact these expensive and cheap IOU roughly 46 The results in columns 1 and 2 plants in non-restructured regimes are 1 0% reductions in both labor and nonfuel expenses for expensive plants. The coefficients on the interactions are economically mean - predict which are post- 1992 and re-run the basic regression suggest most of the input declines relative to associated with expensive we separate plants into "cheap" and "expensive" categories, residual for each plant and classify plants with specification using these and low cost plants and identify high IOU*RESTRUCTURED* CHEAP statistically indistinguishable from zero, contrary to mean reversion predictions, and perhaps suggesting that the form of efficiency improvement was to bring less efficient plants into line with we have more efficient plants. This received from several industry participants, who is consistent with comments claimed that restructuring led their firms to identify high-cost plants as those disadvantaged in the dispatch order, and to focus attention on bringing the costs of those plants closer MUNI benchmark implies to an efficient larger efficiency gains for both cheap and expensive both categories of MUNI plants experienced higher input use than states, We though the increase was largest at the cheap benchmark IOU plant. IOU plants, as plants in nonrestructured MUNI plants. have implemented a number of additional robustness checks, including alternative instruments and instrument strategies and more flexible dynamics in input choice. complete discussion and example results are available 46 Using the The direct effects of cheap and expensive are absorbed in a technical appendix. A more Of particular note in the plant fixed effects. 29 were specifications of fixed costs of input adjustments. that allow for the possibility We find that the restructuring estimates are robust to allowing for differential effects of output on input use depending on whether current output to is above or below plant mean output, and to allowing inputs respond to future as well as current output Labor use estimates were levels. using lagged values of output as instruments, following Blundell and although the estimated difference between remained economically important. IOU Bond less robust to (1998, 2000), plants in restructuring regimes and MUNIs 47 Conclusion 5. This research provides some of the first estimates of the impact of electricity generation sector The restructuring in the United States on plant-level efficiency. yield substantive medium-run efficiency restructuring regimes reduced their labor gains. The results suggest restructuring estimates suggests that to a The estimated benchmark based on municipal, reductions in labor use and plants over the same time 20% plants in and nonfuel operating expenses by about anticipation of increased competition in electricity generation relative to did not restructure their markets. IOU federal, IOU plants, 5% when compared on the order of 15% reductions in nonfuel operating expenses relative to period. There evidence of increases is little to plants in non-restructuring regimes, although the power of these in plants in states that efficiency gains are even larger and cooperative may non-IOU in fuel efficiency relative tests is limited given the plausible range of possible fuel use improvements. These same-plant reductions in input use suggest an important role for competition in promoting technical efficiency, buttressing the findings of Nickell (1996), Galdon-Sanchez and Schmitz (2002), among Ng others. This finding the industry context. Generating plant technology is 48 is particularly interesting given reasonably well understood by engineers, and the pre-restructuring industry was remarkably open operations and input use across plants and firms. and Seabright (2001), and in sharing detailed information on plant Presumably, external benchmarks were more accessible in this setting than in most industries. This could suggest that competition induced greater effort effort, rather on cost reduction by increasing the sensitivity of returns to managerial and worker than by reducing informational asymmetries over managerial effort (see Nickell, 1996). 47 48 Whether lagged output is a valid instrument for this application is, moreover, a debatable proposition. Evidenced, for example, by our use of publicly available, plant-identifiable, data in our analysis. Additional work remains to be done electricity industry efficiency. 49 to fill out the picture of the overall effects of restructuring on We have started by existing utility plants both because this is looking at operating efficiency within one of the few places where gains are likely to before restructured wholesale markets open up and because rich data are available on owned plants. As our results show up utility- suggest, even these data are inadequate for the fine-level analysis required to estimate within and across-plant changes in fuel efficiency. This analysis will require datasets with both cleaner measures of fuel efficiency and richer information factors that affect fuel use. Finally, assessing efficiently after restructuring requires power plants are so long-lived, very more whether investment decisions are made more time, and access to better non-utility data. Since few new additions are made each no more than a handful of anecdotes about investment It is on independent year, and currently we have after restructuring. important to recognize that these efficiency estimates are, however, only one input to judging the ultimate benefit of restructuring policies. realized magnitude of potential dynamic The overall assessment depends as well efficiencies, on the and possible offsetting effects from higher investment expenditures, restructuring costs, the loss of coordination and network economies within vertically integrated systems, and the exercise of market markets. Dynamic costs could be higher productivity growth over time. It is striking as firms respond to the new human and if restructuring power in unregulated generation reduces knowledge sharing that affects possible, however, that longer run effects will be incentives created more by restructuring with investments in both physical capital that further enhance efficiency. If California's crisis does not induce reversals of the restructuring movement, and regulators do not shut down data reporting and researcher access to detailed plant-level data, time may enable us to distinguish among these possibilities. See Wolfram (2004) for a discussion of the general issues involved efficiency changes accompany in assessing different types of electricity restructuring. 31 Table Note: 1: Number of Plants Affected by Restructuring in Each Year Total IOU* IOU*RETAIL MUNI* Year Plants RESTRUCTURED ACCESS POST 1992 1981 465 1982 490 1983 511 1984 527 1985 546 1986 547 1987 582 1988 588 1989 592 1990 597 1991 600 1992 591 1993 602 27 119 1994 598 103 129 1995 608 166 129 1996 589 187 119 1997 569 239 101 1998 557 236 47 101 1999 486 183 30 91 TOTAL 10645 1,141 77 789 RESTRUCTURED indicated those that have started formal hearings that eventually go on to pass legislation, and states that have started retail RETAIL ACCESS indicates plants in access. 32 Table 2a: Summary of Data for Plants Larger Than 100 Non-Restructured lOUs Variable NON-RESTRUCTURED (N = 244) (N = 196) AGE %COAL %GAS 0.29 0.21 19780 14354 0.26 0.14 .03 16490 8496 3291 11 2.0 11 3.1 0.21 0.41 0.20 -.01 806 28 52 36 662 645 .8 50 806 24 80 48 16 11 Mean (TJ MW %GAS -.54 -.54 13 3.4 40 37 -28 -6.48** 21 5.11** 1985: Munis versus = 196) Difference in Means T-statistic for Difference Std. Dev. AGE %COAL MW, -.1 1.79 2.99** 0.01 2.95** NON-RESTRUCTURED MUNIS T-statistic for Std. Dev. 0.39 (N=106) EMPLOYEES /MW NONFUEL EXPENSE/ MW HEAT RATE CAPACITY FACTOR Means Difference Mean Table 2b: Summary of Data for Plants Larger Than 100 Non-Restructured lOUs Variable Difference in Std. Dev. MW 1985: Restructured versus RESTRUCTURED Mean EMPLOYEES /MW NONFUEL EXPENSE/ MW HEAT RATE CAPACITY FACTOR MW, Mean Std. Dev. 0.27 0.13 0.26 0.14 .01 .45 15220 9292 16490 8496 -1269 -1.17 12 5.9 11 3.1 .4 .68 0.40 0.23 0.41 0.20 -.01 -.23 703 608 -102 -1.36 11 -5 -4.8 -3.53** 68 27 47 45 806 24 80 645 19 40 37 -12 -2.16** 12 2.27** 16 Significant at .05 level or better 33 Table 3a: Percentage Change Difference of lOUs in in Costs From 1990 Means Tests to 1996 Non-Restructuring Versus Restructuring States N Distribution Transmission Generation 74 12.3% 22.7% -17.0% 49 17.3% 33.8% 4.2% Difference -5.0% -11.1% -21.3% tstat -0.99 -1.34 -3.65** Restructuring Mean NonRestructuring Mean *significant at All 5% measures are Table 3b: Percentage Change in in in 1 % level nominal dollars. Costs Per Difference of lOUs "significant at level MWh* From 1990 to 1996 Means Tests Non-Restructuring Versus Restructuring States N Distribution Transmission Generation 72 1.5% 13.1% -13.5% 48 -1.6% 12.6% -5.1% Difference 3.1% 0.4% -8.3% tstat 0.70 0.06 Restructuring ; Mean NonRestructuring Mean *significant at All 5% level measures are "significant at in -1.87 1 % level nominal dollars. Transmission and Distribution costs per MWh are costs per MWh sales to ultimate customers, while Generation costs per MWh are costs per MWhs generated at company * plants. Note: Three observations dropped due to lack of MWh data. 34 — — * * .-' — —^ — — — — » in ro •a CU i_ 3 C O 3 3 * *— i^ > s is ^^ ^ * * o O o O oo o o o o o o o d d d d dd d — d o— o ^_ * ?- ^r -3- (N * * , s-^ * , CD LO co sf LO en CM CM •* T LO r-r^ LO r^ CM co ^r CO T— d. CD oo od -*— C CD o * co cu 1 'c az. 'co *—* co _ •= > J3 "cu « * w o cu LO " O *S * i ^ 'S- 1 CD C CD u O CD cu * * * re m > F^ T— co CM co csT ^r co o O r^ LO T— o o q o O o o o d^ d d d d d d CM oo o odod d O dO +• --*» * ^ ^~ T— LO LO *CO CM CD IN CM f * , q o £ re co co CM en S" r^ LO tf r^ * o O ^ o o o o O o o d d d o — — d d d <p^ T q^ 1 _i Nf LO~ oo — Od o T3 + , LO * co CM tf CO CM CO Q , r~- * 3 co t -fu CO o cu it o ^> * * re CD — * o o d d cu l_ CD cu — * , LO LO CM LO co" co "3" CM , LO Lo~ CO CM n T^- CO r^ co LO •^r o o o o o o o o oo o d d d d d d cd oq od CXI >* — CO "if c . . => "5 CD 1 T3 i»S C "to h~ ca cu * +j co co O * re * Q * o~ * , ^ LO co CM o o d q^ 5 , ^^ * , o" , j: * Uj o o o o o o co o o o o o T— d^ o d d d^ d d o d £2.8 c + * CD CM LO re * oo od CD CO + T co CM* LO CM co LO i 1 CD CO OJ 8 O to _1 So o '35 re * * rsi * co oo oo o co oo o o co d d d d d o d^ o d 1 <*~^ O * CD lo * ST ^_ * ^ T CD CM * * CD ~ T o * q o^ o O o o o dq d d d— " 1 Q Q 0) m CO- o: 0) to re ? o s CO co CO ** fl) J2 re 1- c: CD v t— o CO CD T3 u_ CO CM LO CO* z. CD "a CD LO T <J> * o q d d o o c^ o zo _^ * 1 cu CD CO CD II CD j^ 'o 0. CD LO CD T CO ^~ _i CO "™' +J 00 ST a> co * oo dd q^ 4c _c C o LO LO CM CD * * * O 1 B, <D~ 1 a! * * oo mo * _5 o 9? o o CN -J 1CO 2 CO "co Ul CO CO UJ p 1— o -I— CO JD o o U4 o: «* -J UJ o; 0. «- b o b o b I § a: CO UJ cc *» 1 UJ" O s 1 ef § LU Ul o Q | s *' 73 a> f .— > c ° co o 'E CD (0 _ w ID -*— CO CO *3 *-» CO o "O 0) "2 !?Uj .2 W -J CO Q m £ceo — .£ * CO to ro r -S £,re^ OtS c S •*- ac .g>§ w re * to LU CO J= §.</>£ 5 j^-g Jo£ C I CD in 3 O C>4 u. 8 CD '* ...£ (D ii — . to -Z. o T3 - CD CD «*- a a a. CO 2 a> to to w o re CD in _a> n re re co o 'o -a C c o CD £ K = g E$ CD O O t 1- CO O re r~ \M < o CO , E b o o g o to o to UJ cc co _2 o o 2 •t— a to to UJ E 8 U) -Q O 1 cc UJ UJ b o b o I I 2 Q O U. — — — — — ' — —— — rro TJ 0> 3 oo ._ +» co > c U 3 3 i_ s , ^^ * t LO CM CM oo * T— T— 00 'd- a> h~ ^t^ o o o o d^ T CM CM f^ T T o o o o O o o o o O Oo d d— o d d d d d d d 'd- CT> > ^ ' ' CO a> CO <D r CD CD Qi ro a. »™ -^^ lO > CO +-» CD q; 10 CO a> * ? o T— u o o o d d^ LO CD~ CO CD LO CM CO T < CO CM CM CO o o o o o o oV o o^ o o o o d oOoo d O^ d d^ * Co" CD r^ CD ^J- T o ac o^ t- ro ro ro a> co o co T— oo od oo T— ^> Q T 5 * co * r^- CM CM LO oo o o o O O o oo o CD d d d d d — d — t*- o o o d d— CO to -a cu «= —I s _l "53 o o •~^» 00 > '(/) CO m * ^ o o o d d^ T_ LO T s~ * csT CM T ^ -3- en co v> LU - cm" LO T o o o o oo oo d^ d d d d d oo dd co *- T t^ CD o ro -^- CD ^O _3 # ro CO CM _l "~~- t _c c o *3 o" * oo T— oo 00 d^ CO _l ^O i 1 s~ O o o d d CM P o o o oo o o d d^ d d CM cm" CM f^~ co o o d d_ CO w o * ro t- >> ro T3 ro 'c sz ,ro i Q LU ¥ Q LU £ ? 1o s CO CO LU o <N CO CO K CO LU q; s «s -J LU o: J2 o O o CO O 0. «- 1 "c ro ? 1— o CO LU 0£ «2 1 § CD LO in 1 CD o § LU 5 LU CD <* Q O U. ro "55 C CD m 0) co £- Q. CO CD _ o o OQ o o CO * re a \ ii^ M- 'o ro LU CM CM ?7 o" o o o o o o to d d dd d * T Oo d CM ,- 1- 21 £" r*_ o o d CD o s P o si o c * CO — co CD "c CD O CO ^ * * * co o * o CD o O o co o o o m co o CD d q^ q d d IV ln(NON-FUEL EXPENSES) * m * co co 1 1 * CM LO * CD in LO o o o o co o o q LO eq d d CM d o d d CD CO co • CD o -^ c W It— OjUj »^ I 3 U. CD UJ CO in * * CO CD CM en CD * * ltT co co o en m o O d^ Co * * co LO <r CD LO d^ d i * CM~ CM O o O o o d d CM d d * o CO O o s CO 6 3s-oMJ I— Ul ttf ,C UJ g re 52 ** CO c — o ca -^ cz V) 0) +" CD re co S> * s E (0 en (A * * * UJ s >3 (A LO CD * * ST * * CD lo CN CD CM co" co CM co CD CM LO CO CO T i 1 3 5 0) CD CO" Ul 3 s T" S2 >3 o re i— * * * * * in *LO CN CO * * ST CM en co CO o co O o o o o o o o CD q o d^ d o o o o q d O d d IN o o o cd 1 <=CO co o o O O CM o in o CO o CM O o q q q o o o q O d^ O d CD d d d d d iq o <q O d d CM q d i +-> * * * CO co lo cp co i CD CO CD 0) CT) <U T3 CD XI c T3 c © i— "co 3 a c o z CO cd =) a: cd CN O) o» o E3 o CQ re l- b o CO o lco b o b O O uT 0. s. CD |c 1 Ul UJ CD Q CD U. Z -— £ CO o c O Q) °- £ ® 5 c S — _CD a. £ CD S CD 1 I— Uj "= CD £ 5 w -c ^ _ ' — — ^ ^ •' CO j£ "— O) 'to * j * o o— *o 5Oo o o q oo O zo o o q oo o ooo o _^-^ fc ^— _^_, ^— ,, * ^— CM T- CO 'S- ** CM CM * 1^ CD CO CO CO T— CO a> LO "tf" in CO CM i^ T— T— in TUJ ld <— T— in CD CD 1 •^5 in 1 re -§— c CO o > U— 'c D) ^-^ * * T— o CO o 00 ^ ^ ^_ a CO s~ CO CM ^r * h- in CO o o CD r^ CO _i o o o qqoo q q o o o ooo oq o o o in £n* co Ul LU > o _ _ 0) * * _i * CO co~ * _i CD CO * -3t^ CO CD CO <y> CD CO CO CO LO o o <Oo o o qooo oo o o CO O CO CD CD in CD^ o o o o o — O o (0 CD 5! 1c o cx> * i > .J2 CD re OQ * o CD CD o P CO CO CM CM CO CO in CM o q CD o o oo O ooO o ooo CO o oo oq CO V .c * * T co~ in CO CM CO qm do in O (0 a> 0) re 1- J; to u. * CO <D £ CO CD OT Q) j= £ O a. CD ito ca-e T3 CO Q UJ § K o X3 c is "to CM O) T3 to CO 00 C CD re 3 j| o c Q. >N .Q to "D 12 o * CT> '& re CD 1? 1 i 'to ir- T i^~ to Ul DC h o O ul a. CD c o 1 § CO CD u. CD 4— to 3 o •*-* to 1 § |c c LU <3 «* Q CD U. to J2 o J3 O a: — " ! ' o * * CO LO * 00 CD CM * ^— ^ LO CO CM 5 CO in o oo o oo O o *" LU LO o LO T— o oo LO CO LO z O d d d— d d d d o d— lo * ro 05 1 '^r s 1 ' >3- * ' ' " * s " > ^ i CO # * * * * * 00 CD" * C\T * * CO Q o CO CD CD -a _l o o <- 5o O o Oo o CO O d o^ d d d— d— o LO a d 3 d d d "o CO LU CO "3"3- CD s - § ' ' ^c CO <D O re * * * * * * ^ 00 * o LfT CO LO oo O CD CO o < o co o o o O LO o CO 00 o dV CO CN d^ LO S d d dd d o *•> *• 'vf ST ** S" ^F > 73 X M L_ CO CD "a * * O CQ * co~ * co CD * * * oo" * CO CM CD O oo o d^ d d d * * CD O, * * * c co ST LO CM o co LO o CO o o CM d d^ d d_ LO O sz o o Q. CD CO K c Q >» UJ rCO re o^ *— CO E ro « CO O >*= c c "a t3 -S a D) * CD * 1— CD ^p . CO o "o *-^ CO CO 13 -^ c CO o o -Q fc B o (0 CD 1- CO "(0 -Q o ,. clJ a: iE re Q. en .S> CO CD re 3 .c H= cd i (A ro o c*_ u. ^ ° -1—1 _l _l o m co CD s 9>< UJ 5 CO *J CO UJ h o o a. (A UJ § c o 1 § ^ *3 re LU CD <* Q CO CO u. o Table 10: Test for Mean Reversion Estimates Effect, IV In(EMPLOYEES) ln(NON-FUEL EXPENSES) IOU*RESTRUCTURED_cheap -0.011 IOU*RESTRUCTURED_expen (0.028) -0.106*** (0.022) -0.102*** MUNI*POST 1992 cheap (0.031) 0.143*** (0.022) 0.193*** (0.041) MUNI*POST 1992 expen 0.056** (0.022) 0.095*** -0.004 (0.021) (0.027) -0.115*** ln(WAGE) (0.045) ln(NET MWH) 0.409*** 0.037 (0.084) AGE (0.061) -0.003 0.003 (0.002) FGD 0.028 j (0.123) (0.056) 10222 842 plant-epoch and 19 year fixed Robust clustered standard errors significant at 1 0% ** ; significant at Instruments for 5% (0.003) 0.235* 10258 effects included. in parentheses, *** ; significant at 1 % MWh: \n{STATESALES) 41 Figure 1: Year-Effects by Group from Basic IV Specification: Employment 0.2 0.1 ^--^•«^ "•^Munis • * - lOUs: Non-Restructuring States lOUs: Restructuring States -0.5 42 Figure 2: Year-Effects by Group from Basic IV Specification: Non-fuel Expenses — •— HI -a a x - -* ra °P N$> Nc£> #> N £p £}> $> ,$> Nc£> ^> ^ ,d? ,d? c$s s ^ ^ ^ ^> ,dp Munis - lOUs: Non-Restructruing States lOUs: Restructuring States Sources Ackerberg, Ando, Dan and Kevin Caves. 2003. Amy W. "Structural Identification of Production Functions," and Karen L. Palmer. 1998. "Getting on the Map: The Electricity Restructuring," Political UCLA mimeo. Economy of State-Level Resources for the Future Working Paper 98-19-REV. Averch, Harvey and Leland L. Johnson. 1962. "Behavior of the Firm Under Regulatory Constraint," The American Economic Review, 52 (5), 1052 1069. - Change Following Deregulation: The US Experience," Bailey, Elizabeth E.. 1986. "Price and Productivity The Economic Journal, 96 (381), Baron, David P. and Raymond R. De 1 - 17. Bondt. 1979. "Fuel Adjustment Mechanisms and Economic Efficiency," Journal of Industrial Economics, 27 (3), 243-261. Blundell, Richard and Stephen Bond. 2000. to Production Functions," . "GMM Estimation with Persistent Panel Data: An Application Econometric Reviews, 19 1998. "Initial Conditions and Moment (3), 321-340. Restrictions in Dynamic Panel Data Models," Journal of Econometrics, 87, 115-143. Borenstein, Severin, James B. Bushnell and Frank A. Wolak. 2002. "Measuring Market Inefficiencies in California's Restructured Wholesale Electricity Market," American Economic Review, 92 (5), 1376- 1405. Christensen, Laurits R. and William H. Greene. 1976. "Economies of Scale in U.S. Electric Generation," Journal of Political Economy, 84 Courville, Leon. Griliches, Zvi and Jacques Mairesse. Press: Cambridge, 5 (1), 53-74. 1998. "Production Functions: Econometrics and Economic Theory in the Iron Ore Markets in the 1980s," M and Stephen H. Karlson. Economic Good, David Efficiency," The 20th Century, Steinar Strom Jr.. American Economic Review, 92 Cambridge University 1222-35. (4), (4), Mechanism on 574-584. and Robin C. Sickles. 1993. "US Airline Deregulation: Economic Journal, 103 (July), 1028-1041. "The Determinants of the Cost Efficiency of Electric Generating Stochastic Frontier Production Approach." Southern L.. ed., 1978. "The Impact of the Fuel Adjustment Implications for European Air Transport." The Joskow, Paul for Identification," in 2002. "Competitive Pressure and Labor Productivity: Review of Economics and Statistics, 60 H., Lars-Hendrick Roller, Hiebert, L. Dean. 2002. The Search UK. Galdon-Sanchez, Jose E. and James A. Schmitz, Gollop, Frank 655-676. 1974. "Regulation and Efficiency in the Electric Utility Industry," Bell Journal of Economics and Management Science, World (4), Power Economic Journal 68 1997. "Restructuring, Competition and Regulatory The Journal of Economic Perspectives, Reform in the (4), Plants: A 935-946. U.S. Electricity Sector," 11 (3), 119-138. 44 Joskow, Paul L. and Edward Kahn. 2002. "A Quantitative Analysis of Pricing Behavior in California's Wholesale Electricity Market During Summer 2000," Energy Journal, 23 (4), 1-35. Joskow, Paul L. and Richard Schmalensee. 1987. "The Performance of Coal-Burning Electric Generating Units in the United States: 1960-1 980,' 'Journal of Applied Economics, 2 (2), 85-109. Hendricks, Wallace. 1975. "The Effect of Regulation on Collective Bargaining in Electric Utilities," Bell Journal of Economics and Management Science, 6 Knittel, Christopher R.. 2002. "Alternative Regulatory Evidence from the US Electricity Industry," 45 1 -465. (2), Methods and Firm Efficiency: Stochastic Frontier The Review of Economics and Statistics, 84 (3), 530- 540. Kleit, Andrew and Dek Terrell. 2001. Electricity Generation: "Measuring Potential Efficiency Gains From Deregulation of A Bayesian Approach," The Review of Economics and Statistics, 83(3), 523- 530. Kumbhakar, Subal C. and C.A. Knox Lovell. 2000. Stochastic Frontier Analysis, Cambridge University Press: Cambridge, UK. Laffont, Jean- Jacques and Jean Tirole. Press: Cambridge, 1993. A Theory of Incentives in Procurement and Regulation. MIT MA. Levinsohn, James and Amil Petrin. 2003. "Estimating Production Functions Using Inputs to Control for Unobservables," Review of Economic Studies, 70 Moore, Thomas Gale. 1986. "U.S. Airline Deregulation: Journal of Law Nerlove, Marc . & Economics, 29 (1), 1 - (2): Its 317-41. Effects on Passengers, Capital, and Labor," The 28. 1963. "Returns to Scale in Electricity Supply," in Measurement in Economics, Christ eds., Stanford University Press: Stanford, et al. CA. Ng, Charles K. and Paul Seabright. 2001. "Competition, Privatization, and Productive Efficiency: Evidence from the Airline Industry." The Economic Journal, Nickell, Stephen (4), 1 1 1 (July), 591-619. 1996. "Competition and Corporate Performance." Journal of Political Economy, 104 J. 724-746. Olley, Steven and Ariel Pakes. 1996. "The Dynamics of Productivity in the Telecommunications Equipment Industry," Econometrica, 64 (6), 1263-1297. Primeaux, Walter J. 1 977. "An Assessment of X-Efficiency Gained through Competition." Review of Economics and Statistics 59 Rose, Nancy L.. (1): 1987. "Labor Rent Sharing and Regulation: Evidence from the Trucking Industry," The Journal of Political Economy, 95 Scully, Gerald 105-08. W. (6), 1 146 - 1 178. 1998. "Reform and Efficiency Gains in the New Zealand Electrical Supply Industry." Journal of Productivity Analysis, 11, 133-147. 45 Rota, Paola. 2004. "Estimating Labor Demand with Fixed Costs," International Economic Review, 45 (1), 25-48. Spann, Robert M.. 1974. "Rate of Return Regulation and Efficiency in Production: An Empirical Test of the Averch-Johnson Thesis," Bell Journal of Economics and Management Science, Syverson, Chad. 2004. "Market Structure and Productivity: A Concrete Example." 5, 38-52. National Bureau of Economic Research Working Paper 10501. Cambridge, MA. U.S. Energy Information Administration . 2003. "Status of State Restructuring Activity Timeline as of February 2003." http://vvfww.eia.doe.gov/cneaf/chg retail access timeline.html [accessed Van Biesebroeck, Johannes. 2003. November "Productivity str/ 23, 2003]. Dynamics with Technology Choice: An Application Automobile Assembly," Review of Economic Studies, 70 (1), to 167-198. White, Matthew. 1996. "Power Struggles: Explaining Deregulatory Reforms in Electricity Markets," in Brookings Papers on Economic Activity: Microeconomics 1996, 201 Wolfram, Catherine. 2004. "The Efficiency of Electricity Generation forthcoming in James Griffin and Steve Puller, - 250. in the U.S. After Restructuring." eds., Electricity Deregulation: Where to From Here . University of Chicago Press. 46 Data Appendix - Table Al: Summary of Variables Variable Source Definition Output and Input Variables LN_NONFUEL EXPENSE Log of the annual non LNEMPLOYEES production expenses ($) Log of the average number of LN_BTU employees at the plant. Log of the total btus burned by the + UDI UDI of fuel plant. Calculated as (tons of coal * btu/lb) UDI fuel (barrels 2000 lbs/ton* of oil*42 + (Mcf gas* 1000 cf/mcf btu/cf). The gal/barrell*btu/gal) values of the btu content of each fuel are provided in the UDI data for each plant each year. LN_NETMWH Log of the of the net MWh generation UDI plant. Restructuring Variables IOU Dummy equal to classified as 1 for plants IOU, holding, or 1997 UDI Utility Datapak Book private companies. MUNI Dummy equal to owned by 1 for plants utilities classified as 1997 UDI Utility Datapak Book government or cooperative utilities. IOU*RESTRUCTURED Dummy equal to one for IOU plants in states that restructured in years MUNI*RESTR UCTURED following the advent of EIA "Status of State Electric Industry Dummy equal to one for MUNI Restructuring plants in states that restructured Activity" Timeline as following the advent of formal hearings Dummy equal to one for IOU of July 2002, EIA "The Changing Structure of the Power plants in states that restructured, Electric beginning with year that Industry: legislation IOU*RETAIL ACCESS information compiled from: formal hearings. in years IOU*LAW PASSED Restructuring was enacted Dummy equal to one for 12/96," IOU plants in states that restructured in years following the advent of retail access. An EEI Competition Update, "Electric in the States" February, EIA "The Changing Structure of 2001, 47 MUNI*POST 1992 Dummy equal to one for MUNI the Electric plants in years 1993-1999. Industry: 2000 Update," NARUC Power An "Utility Regulatory Policy in the United States and Canada, Compilation" 19941995 and 1995 -96, "Restructuring the Electricity Industry," The Council of State Governments, 1999, and State PUC websites, relevant legislation IOU*HIGHNUGS Dummy equal to one beginning in 1993 for IOU plants in states EIA and Electric reports. Power Annual. with above median penetration ofNUG plants in 1993. BOTH Interaction of CHEAP (EXPENSIVE) RESTRUCTURED and HIGH NUGS dummy variables. Dummy indicating whether plant average residual from input use regression for 1981-1992 period is below (above) INDEPENDENT7 OTHER VARIABLES PLANT AGE Number of years unit at the plant FGD zero. since the oldest came on UDI line. Dummy equal to indicating the operation of a FGD scrubber at 1 UDI the plant. CAPACITY FACTOR calculated as (NET MWH) (GROSS MW * LNJVAGE UDI Capacity factor of the plant, hours 1 in year) wage bill employment BLS State-level annual divided by total at plants in the following industries: SIC 4923-4925 4953 (natural gas distribution), (hazardous waste treatment), 2911 (petroleum refining). STATE SALES State-level utility sales in Sales: Sales to gigawatthours, not including Ultimate Customers non-utility sales. from EIA's "Electric Sales and Revenue" Table 6 and EIA's 48 "Electric Power Annual" Tables 117 and 90. PLANT CHARACTERISTIC VARIABLES Based on the load type reported LOAD TYPE in FERC Form through 1992. UDI 1 We assign plants to the last load type reported if they change load type no more than twice, otherwise we do not assign load types to missing data years. We use this cutoff to include plants had a "stray" change—one peak, cycle or shutdown type in the middle of a string of "base" types, for example. Based on the primary fuel burned FUEL TYPE UDI at the plant Dummy variable equal to one if BIG the plant capacity (Gross greater than OLD MW) is 550MW. Dummy equal to one if the youngest unit in service ECONOMIC AND WEATHER VARIABLES ANNUAL UNEMPLOYMENT Annual ANNUAL HDDAYS at the plant UDI came before 1960. average unemployment Bureau of Labor in the state. Statistics Population weighted heating U.S. Dept. of degree days for each state-year. Commerce National (use ANNUAL CDDAYS UDI MD for DC) Oceanic and Atmospheric Population weighted cooling degree days for each state-year. (use MD for DC) Administration Historical Climatology "Monthly State, Regional, and National Heating Degree Days Weighted By Series, Population" Note: UDI stands for the Utility Dal a Institute O&M Production Cost D atabase. All data in the U production cost data bases are extracted from the EIA Form 412 electric FERC Form by municipal and other government cooperatives). These are annual reports. (filed 1 by investor-owned utilities), and RUS Form 7 & 12 (filed by (filed utilities), 49 Data Appendix - Table A2: Variable Summary Statistics Mean Obs LN NONFUELEXP LN EMPLOYEES LN BTU LN NETMWH PLANTAGE FGD LN WAGE IOU MUNI IOU*RESTRUCTURED MUNI*RESTRUCTURED IOU*LAW PASSED IOU*RETAIL ACCESS MUNI* POST 1992 IOU*HI NUGS BIG OLD LN STATE SALES ANNUAL UNEMP ANNUAL CDDAYS ANNUAL HDDAYS 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10645 10608 10608 15.880 4.562 30.304 14.075 27.925 0.124 10.543 0.805 0.195 0.107 0.013 Std.Dev. 1.137 1.108 1.631 1.760 12.940 0.330 0.255 0.396 0.396 Max Min 9.674 18.694 19.839 5.476 7.099 33.067 16.905 76.000 1 9.815 1 1 .488 0.007 0.309 0.113 0.272 0.085 0.074 0.103 0.500 0.262 0.304 0.500 0.476 0.499 0.850 8.285 12.627 0.081 11.171 0.064 1439.436 4382.568 0.021 0.022 0.180 931.661 81 2144.882 414 3879 10758 50 Technical Appendix: Sensitivity Analysis Robustness of results The conclusion to alternative instruments that same-plant input use declined after restructuring for IOU plants appears robust to a variety of alternative specifications, although the precise magnitude of the estimated efficiency gain depends crucially on the estimated relationship between inputs and output. In tables Tl and T2, we analyze the sensitivity of the results to alternative output specifications and instrument specification of the regression In the first column of each model we table, (similar to low output low output levels based on the 60 th although specifications using cutoffs ranging from the results. 1 relative to th its IV using the basic 5). on output with variables report results that interact the coefficient indicating whether the plant achieved a high or cut-off between high and column 4 of tables 4 and sets, We use a historical performance. percentile of the plant's historical output, to the 80 th percentile yielded qualitatively similar Introducing this flexibility could be important if fixed labor and materials costs cannot be scaled back when the plant operates relationship at historically low output between output and employment (table levels. The results in column (1) suggest that the Tl) or output and nonfuel expense (table T2) is not materially different across output levels, however. Columns 2 through 4 of tables Tl and T2 introduce an column state (2), we add the square of \n{STATESALES) as array of nonlinearities in the instrument In an instrument. This allows extreme departures from average sales to have an additional predictive effect on plant output. 50 Column (3) term between \n(STATE SALES) and RETAIL ACCESS, allowing the relation between output to vary after implementation of retail access. set. 51 Column (4) interacts adds an interaction state demand and \n(STATESALES) with plant fuel type indicator variables, allowing the plant output to respond differentially to state demand depending upon fuel type. For employment, reported in table Tl, these instrument sets almost magnitude of the coefficient on \n(NET MWhs) all increase the estimated relative to the specifications reported in Table 4. The restructuring effect remains negative and statistically significant, although smaller than reported in Table 4. For non-fuel expenses, reported in table T2, these changes in the instrument set have varying effects on the estimated magnitude of the output elasticity as well as the impact of restructuring, relative to results in Table 5. 50 Efficiency effects remain large relative to municipal plants in In the first stage, the coefficient on state sales \n(STATESALES) and \n(STATESALES) uses \n(STATESALES). 2 is all specifications, significant, although the F-statistic are equal to zero is and roughly the on the hypothesis that both smaller than the F-statistic from a specification that just Technical Appendix - page 1 magnitude of those found in table 5 with respect to exception of results in column The final IOU plants in non-restructuring regimes, with the (3). two columns of tables Tl and T2 use lagged values of output discussions in Blundell and applications to input (1998, 2000) on the use of lagged values of endogenous variables in demand and production exogenous instruments. specifications, Bond 52 as instruments. This follows function specifications, particularly in the case of weak In our case, lagged values of output are very highly correlated with output in even those estimated using first differences, although first differences may all not be appropriate given that most plants have scheduled major outages for equipment maintenance and overhauls on a two to We explore two variations. four year cycle. instrument. We report results from this as The simplest, in a base, though this column (5), uses the prior year output as an valid only if there is is no first-order serial correlation in the time-varying efficiency shock, unlikely given the estimated rho in our data. variation uses lags from years two through four years before the set substantially reduces, or eliminates (Table relative to nonrestructuring column IOU plants, 1 1 , column current. For employment, instrument 6) the estimated efficiency gain of restructuring though the performance gain relative to (6) results also suggest a coefficient this The second on output that is much MUNIs remains large. The higher than in any other specifications. In the case of non-fuel expenses, these instruments suggest results for restructuring effects that are broadly consistent with our other results. Robustness of results to including forward-looking output We have explored specifications that allow employment. 53 for a fixed cost would appear especially unproductive employment prior to an in a discrete change. If this were the case, the employment following the reduction until the lower output levels are realized. considerations on our results, forward looking changes and 51 52 levels, particularly This could cause a plant that foresaw declining output over a several year period, for instance accompanying restructuring, to ratchet plant of changing input we estimated input in output. statistically significant in the In the first stage, the coefficient 54 OLS The To demand equations coefficient reduction, and especially productive investigate the effect of dynamic that included variables on the forward looking changes specifications similar to those in on the interaction term is columns (3) in measuring the output were positive of tables 4 and 5, but highly significant. and Bond describe instrumenting with lagged values in a common factor model that allows for a first-order auto-regressive term and is first-differenced to eliminate firm-specific effects. We have estimated this form of the model and the results seem to yield an AR(1) term that is over 1 .6. For the reasons discussed in the text, we don't think our data have a simple first-order autoregressive error structure. 53 There is an extensive labor literature on the presence of fixed costs and other non-convexities in employment In fact, Blundell decisions. Technical Appendix - page 2 negative and statistically indistinguishable from zero once changes 54 in output. This input In both cases, the coefficients demand equation is we instrumented for both output and forward on our restructuring variables were based on one derived by Rota (2004), face a fixed cost to changing their employment who specifies a virtually unchanged. dynamic model where firms levels. Technical Appendix - page 3 ) uTIo"^ CCS ||| to l~ K- I— U| lij UJ CM CM • * * * + * in CM CO K CD CM CO in CO 00 CO CD CO o oO O oqoo q d ddd o d in CD CX I -3 ££— c o D. CX < * 1UJ * * "3 o CM o o o CO O O q d d od d d q dd d Ji lo CNI io si ST v- ¥ CM 00 o 00 CD o a) c ro o i^ to LU a <* <o 5? 5t ° 1tO u. * * un c\T cnT 00 CM in CM co 'c D) 'w * 00 CO CD in o o O o q d d^ o d o d d d d d CNI O i ID "5 if o c CO o M— c to to LU LU * ^^=J JO * o * CO CO o SJo^ft CM m CM CO LU LU £ Yi id o 5 E C m o o d o qd o qd do do ? ? fc ^ d d d D) * LU i CO CD i :3 c5^ C) O (/) CO o to to E 3 i_ LU LU -J -J «* >-: to to S CO CD * * * * * T3 CO CO CO CD CM CO en CO o a> o CM q o q oo o d d_ d q^ d d d d O <L) X o c M— Dl CO CC a) to to u c c n. (U 03 d) CO CD -C .c c o c u CD 9* LU -J 1 < c E UJo c tO Ki-i D. * * * * fl) * * in c\T * o CM CM o m oo o o o d d d q^ d d CD oo CM 3- CM CM CM i CO CD O O dd dd O CXI i to t_ co Q. c c IU 0. CO o l_ LCO T3 i_ CO (0 ac 3 © CO S CO -a (A (A C -Q (0 I- E 3 i_ u c 1 to O a> a. 2 1 0; 1^ 1 C o re CD | CD L_ CD CO 3 o 4—* LU 5? 5^ 3s CO >< (0 a> * a) -CM 0) > CO O o a * c — W +-> c (0 H i CO 3 o CD to .a o a: — — to to ' <o c -c £: I 1 CO to — u ao ca a. c c = I a w Hi < to 11 o o u H CO HI CO § t ® •4—* ICO LL -4— CO c ca o c to UJ uj o CO ° a^ CD 00 = c j= q: I w *-• c .E ro O f= a3 "> CO E 3 i_ cn +> X U) **" c CO > >»JZ c a (0 E. CO CO < c w CO UJ -J = « c 2> CD CO r- i_ CO CD CO o a. o c Q.— to c UJ I- Q. to 1-1 CO i— CO 73 C (0 CO •4—* (O (0 CO H _co JQ re CN OJ Ci *~ T3 B K. o CO O 0. E CO UJ I W o g DC a) to UJ £. CO -^ Iff c CO CO m S * i— to °: .Q O a: Date Due Lib-26-67 MIT LIBRARIES 3 9080 026 :i!!il 1111