- Email: [email protected]
Power plant investment in restructured markets
ARTICLE IN PRESS
Energy 32 (2007) 1403–1413 www.elsevier.com/locate/energy
Power plant investment in restructured markets$ W.D. Wallsa,, Frank W. Ruscob, Jon Ludwigsonc a
Department of Economics, University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada T2N 1N4 b U.S. General Accounting Office, Washington, DC, USA c U.S. General Accounting Office, Denver, CO, USA Received 9 February 2006
Abstract Investment opportunities in electric power generation have changed dramatically since electricity industry restructuring first began. In contrast to regulated utilities adding capacity in line with central planning and regulation, power plant investment is now more often made by independent companies who market power across multiple utility jurisdictions. Because the regulatory approval process is long and outcomes uncertain, developers often plan multiple options for a given development budget. As more information is revealed about the future prospects at different sites, options are abandoned sequentially until only projects that will be completed remain. In this paper we evaluate the decision to build new generation facilities in a changing and uncertain regulatory environment. We estimate hazard rates for new power plant projects using a database of generation projects in North America and examine in greater detail the development patterns in California and Texas, two states with very different regulatory regimes. We find that regulatory uncertainty significantly affects the pattern of development in the electric power generation industry. r 2006 Elsevier Ltd. All rights reserved. JEL classification: L94; Q4; L2 Keywords: Power plant investment; Electricity market restructuring; Regulatory uncertainty
1. Introduction The electricity industry in the United States is in the midst of fundamental change as a result of federal and state efforts to restructure the industry, thereby introducing and increasing the intensity of competition in wholesale and retail markets. Unlike a number of other countries, the U.S. restructuring effort has been hampered by divided regulatory jurisdictions.1 The federal government has $ An earlier version of this paper was presented at the American Economic Association meeting, Washington DC, January 2003, and at the International Association for Energy Economics North American meeting, Mexico City, October 2003. We would like to thank Chris Knittel, other conference participants, and three anonymous referees for helpful comments on earlier versions of this paper. The views expressed in this paper are solely those of the authors and are not to be attributed to their employers. Corresponding author. E-mail address: [email protected] (W.D. Walls). 1 See, for example, the discussion of Brennan [1] for legal and economic perspectives on the roles of different levels of government in a federation.
0360-5442/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.energy.2006.11.002
jurisdiction over wholesale electricity sales and movement because electricity at the wholesale level crosses state borders and therefore qualifies as interstate commerce.2 Retail markets, on the other hand, are under individual states’ jurisdiction. This historical fact has led to a patchwork of different rules and regulations governing electricity markets. For example, while wholesale markets have largely been restructured to allow the emergence of a large and active private sector of electricity generating companies, retail restructuring has been spotty and incomplete. Currently, according to a recent GAO [3] report, 24 states and the District of Columbia have enacted legislation or issued regulatory orders to open their retail markets to competition. However, of these, seven states have either delayed or suspended implementation of 2 In addition to interstate trade, there is substantial Canada–U.S. trade in electricity that adds another level of institutional complexity. See Feldberg and Jenkins [2] for a brief legal and institutional analysis of this issue.
ARTICLE IN PRESS 1404
W.D. Walls et al. / Energy 32 (2007) 1403–1413
restructuring and the remaining 26 states have not yet taken any steps to introduce competition at the retail level.3 The result of this divided jurisdiction and diverse approaches to restructuring has been to introduce a great deal of regulatory uncertainty into the market. This uncertainty is having an impact on the development of new generation facilities. One key feature of restructuring has been a move away from centralized planning, wherein utilities, in conjunction with state public utilities commissions, planned for development of new generating capacity and transmission upgrades in order to meet expected increases in future demand. In its place, a decentralized process of development and investment decisions—largely by non-utility companies—is evolving. Unlike the rate-regulated regime of the past, the development and investment plans of these myriad companies are not subject to approval of public utilities commissions, nor are they coordinated in any way by a central body. This is particularly true in states that have aggressively pursued retail restructuring—sometimes requiring or encouraging their utilities to divest generating resources—but it is also the case in other states to the degree that non-utilities find it attractive to develop new generating resources in those states. Although under restructuring states will no longer oversee the entire process of development and investment in new generating capacity, state entities still wield significant power to influence investments through licensing and permitting processes, through the terms of interconnection agreements, and more generally, through state decisions regarding whether and how far to pursue restructuring of their retail markets. Specifically, state and local agencies responsible for air and water quality and land use decisions must grant approval for companies to begin construction or operation of new power plants. The role of these agencies is to ensure that any new development is in compliance with relevant laws, ordinances, and regulations. There is considerable variation across states in the administration of the development process and thereby in the costs developers must incur to gain approval from state and local entities. 4 Firms balance expected return against the effect of new investment on the risk of their portfolios of projects. In order to mitigate the risk across regulatory jurisdictions 3 See Joskow [4] and Brennan et al. [5] for a discussion of these and other issues surrounding restructuring. See Borenstein [6] and Joskow and Kahn [7] for non-technical examinations of the California market. 4 Federal environmental laws and regulations, as well as laws protecting endangered species, also play a role in determining where and how new power plants are built. For example, proposed new power plants in any area that is not in compliance with EPA air quality regulations are subject to ‘‘new source review,’’ requiring plant owners to purchase or otherwise acquire air emission credits equal to or in excess of their planned emissions. Often the new source review permits are issued by state agencies that have gained approval from the EPA to grant such permits. In the event that a proposed new power plant might impinge on the habitat of an endangered species, developers must also get approval from other federal and state agencies.
and over time, firms may diversify their investments across regions and states, and across power plant types and fuel sources. In addition, because the regulatory approval process is long and outcomes uncertain, firms may plan multiple options for a given development budget. As more information is revealed about the future prospects at different sites, options are abandoned sequentially until eventually only projects that will be completed remain. The costs of early development—the so-called soft development costs, incurred prior to breaking ground for construction—are a small fraction of total costs to build, but they are significant in magnitude running between several hundred thousand and many millions of dollars depending on the characteristics of the site and the state and local requirements. These soft development costs reflect the cost to developers of acquiring an option to build a power plant. In addition to the development costs associated with acquiring regulatory approval, new power plants must be interconnected with the transmission grid, frequently requiring costly upgrades to the system to maintain reliability. The terms under which these new power plants are allowed to interconnect and the distribution of the costs of upgrades is another critical factor that determines where and when power plants are built. Again, there is considerable variation across states in the interconnection costs, and a developer’s share of these interconnection costs can run from a few hundred thousand to tens of millions of dollars, depending on the characteristics of the existing transmission system and on how the costs are assessed. Many hazards lurk in the regulatory arena. Because the development process can be long—running to many years in some cases—regulatory and market conditions may change considerably, causing developers to reassess the relative merits of each of their projects. In addition to regulatory uncertainty, energy prices have been volatile both across time and across regions, adding to the risk of committing development resources. This underlying market risk is increased by long state and federal approval processes and by regulatory uncertainty and unstable market rules because unforeseen events may occur during the course of development that change price or load forecasts. We present in this paper an empirical examination of the overall investment patterns in North America, with a closer look at two states—California and Texas—in the context of those states’ very different regulatory regimes. We evaluate the permitting process for individual firms in the context of the decision to build power plants in an uncertain environment. In the next section we explore the decision-making process of power plant developers and develop a series of conjectures about the expected pattern of investment. In the remainder of the paper we examine— using a micro dataset of power plant development and investment—the recent experiences in North America overall, and of California and Texas, specifically, in attracting new generation development. In the latter case,
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
we will examine the types of generators being built, the actions of these states vis-a´-vis restructuring, and the impacts of state actions on variations in the rate at which plants transition between different phases of the development project and the rate at which they are completed (or fail). 2. Uncertainty affects power plant development We consider firms contemplating irreversible investments to build new electricity generating units under a regime of regulatory and other uncertainty. Specifically, there are three principal areas of uncertainty that influence new generation development. First, regulatory uncertainty with respect to both the timing and pace of restructuring varying across states creates a complicated regulatory environment for investors. Second, uncertainty about future demand—and therefore electricity prices—is a critical feature for merchant power producers who plan to sell a significant proportion of their power in spot or other relatively short-term contracts. Third, the permitting process for siting new generating plants varies a great deal across states and is fraught with uncertainty about the length of time, cost, and ultimate success of applications for approval to build. Because permitting takes a significant amount of time, the firm will generally receive updated information about forward prices and the regulatory regime during the time it takes to negotiate the permitting process. Also, because multiple firms often compete to build generating units in a given market, conditions may change considerably over the period of time in which a firm is acquiring the necessary permits to build. In this context, each stage of the permitting process represents an option to pursue the project and can be evaluated in light of contemporaneous information about forward prices and other relevant factors before deciding whether to invest in the next stage of permitting, and ultimately whether or not to build the plant. A full model of a firm’s investment decision in which the firm has multiple options, each with potentially different costs of development, degree of regulatory uncertainty, and forward price expectations has proven to be intractable. However, we believe the intuition flowing from the standard real options literature as well as the theory of the firm is sufficiently clear to allow us to make a number of conjectures about a firm’s optimal choices in response to various changes in its exogenous environment. A series of such conjectures are made below and then compared against observed investment patterns in the following sections. The value of waiting to invest increases as the relative value of information that may be revealed by waiting increases. The firm is therefore more likely to wait to invest the more uncertainty exists, ceteris paribus. In the standard two-period model in the real options literature, firm’s only have the options of building in period 1 or waiting to get the information in period 2. In the choice to invest in new
1405
generating facilities, firms can respond to greater regulatory uncertainty in several ways. First, a firm can begin the permitting process in more than one jurisdiction, knowing that it will likely not build in each of these sites—the final choice of which units to build will depend on the realization of information during the permitting phase. This will naturally lead to less observed investment, holding forward price expectations constant, in regions in which uncertainty is greater because delays will be longer as firms wait longer to learn more. It also implies that firms will sometimes explore several options for each increment of investment, even investing time and financial resources to start the permitting process while they await information that will enable them to choose between competing options. Alternatively, firms that face greater regulatory uncertainty can vary the expected payoff period of their investment. This can be done contractually if the firm can find buyers willing to bear some of the risk by entering into long-term contracts to buy electricity, or it can vary the type of investment. With regard to the latter, a firm facing greater regulatory uncertainty over the life of an investment may choose investments that have a lower capital cost relative to the capacity of the generating unit. Firms compete with one another to build power plants in the right locations and at the right time to meet expected demand. Suitable locations generally require a nexus of access to fuel sources, transmission lines, and water for cooling. For example, developers of natural gas fired power plants—the predominant technology being built in recent years—look for sites with access to high-volume gas pipelines with excess capacity. Similarly, coal fired plants need access to rail lines, or direct access to coal at the source. Access to transmission suitable for interconnection is also critical for developers. The terms under which interconnection is approved vary a great deal across states and control areas as do the distribution of costs of upgrades. For example, in Texas, the upgrades required to interconnect new power plants are paid for through a surcharge on electricity sold to consumers, while in California the developers have been required to pay for any upgrades deemed necessary by the local transmission owner—generally a local utility company. The implications of different approaches to assessing interconnection costs on power plant location can be profound. When the costs are borne by consumers, developers can focus more on finding locations with lower development costs, easier access to fuel sources, and water for cooling. It is possible under these conditions that there will be overbuilding of transmission upgrades, because developers do not bear the costs of any negative externalities they impose on other grid users.5 In addition, this approach may lead to concentration of generating 5
These externalities will exist in as much as adding generating capacity at a point increases transmission congestion, thereby limiting incumbent generating plants’ outputs.
ARTICLE IN PRESS 1406
W.D. Walls et al. / Energy 32 (2007) 1403–1413
units at some distance from the load it serves, because land costs are lower and environmental issues, such as air quality, less prevalent on such sites. On the other hand, when developers bear the full cost of upgrades, they look for sites with lower interconnection costs, which—given the nature of the flow of electricity and of congestion in the existing transmission grid—may encourage development closer to the load it will serve. However, it may cause underbuilding of upgrades because the developers are not compensated for any positive externalities accruing to electricity consumers.6 Cooling water is essential for many of the most commonly built power plants. For the most part, this requires locating near a source of fresh water, although some designs allow the use of waste water.7 The volume of water drawn by power plants in the United States is quite large, ranking second only to agriculture. The watercooling processes used in most newer power plants lose less water to steam in the atmosphere than do older technologies—most of the water is recaptured and returned to its source. However, returning warmer water to a fresh source can have negative environmental implications, and these issues have led to controversy and delays or denials of permits in some cases. Finally, developers and investment bankers also prefer, other things equal, stable regulatory jurisdictions and clear market rules for trading electricity. Very few states have established and maintained clear paths to retail restructuring and this creates regulatory uncertainty. Specifically, only 17 of the 50 states and the District of Columbia have enacted and implemented legislation allowing consumers to choose their retail electricity provider. Even among the states with retail choice programs, the states have simultaneously reduced and frozen retail rates at levels that have discouraged retail competition. A lack of retail competition also feeds back into the development of new capacity by limiting the ability of developers to enter into long-term supplier contracts with large consumers or multiple retail sellers. The absence of multiple buyers in a given utility’s control area puts non-utility sellers at a strategic disadvantage in negotiating terms of electricity trades. Based on the discussion above, we make the following conjectures:
We expect to find multiple proposed projects for every project ultimately built and that these projects fall out of
6 From the perspective of efficiency, the ideal is to assess upgrade costs on developers in the amount equal to the negative externalities imposed on other transmission users, and assess costs on consumers in the amount that they benefit from the new capacity. In practice, there is a great deal of uncertainty about the value of either externality, but it is fairly clear that neither extreme—assessing all costs to consumers or all costs to developers—is optimal except under extreme conditions. 7 It is also possible for some units to use a largely air-cooled process, although this is much more costly than water cooling and has the disadvantage that it works least well during hot periods when, in many parts of the country, electricity demand is frequently at its peak.
consideration at different phases of the development process as information becomes available and firms narrow their options. We expect to find a distribution of power plant types that is consistent with access to fuel sources. This means that the fuel mix that has evolved historically will be expected to continue to vary by region according to the relative availability and cost of acquiring different fuels. With regard to interconnection costs, states that assess greater proportions of interconnection costs on developers will attract, ceteris paribus, less investment overall. The investment pattern will be toward more generators close to the load it will serve. States that ‘‘socialize’’ the cost of interconnections can expect greater investment and more geographic concentration of generators in areas with lower land or licensing costs—typically areas further from the load it will serve. Further, we expect to find proportionally more development by non-utility companies in states that have implemented retail restructuring, thereby signaling a commitment to developing competitive markets for electricity. In addition to differences in the amounts of development by utilities and non-utilities, we expect to find differences in the types of projects undertaken by these groups in various jurisdictions. For example, in states in which regulatory uncertainty is deemed to be more extreme, firms will tend to substitute away from capital intensive investments, such as combined cycle baseload generating plants, toward less capital intensive investments, such as combustion turbine peaker units. Finally, we expect that development will be more prevalent, again controlling for price expectations, in jurisdictions with more stable market rules and regulatory regimes, than in jurisdictions with greater regulatory uncertainty.
3. Empirical analysis 3.1. Data collection The data used in this paper are compiled primarily from monthly reports of the NewGen database published by RDI, a division of Platts.8 RDI gathers data on new generation projects from trade publications and state and federal data sources and reports the status of each of the projects they identify as of the reporting month. The status reports identify projects as being in one of six categories— proposed/early development, advanced development, under construction, operating, tabled, or canceled.9 For the 8 A more detailed description of the data collection exercise and of the assembled data set is contained in Ludwigson et al. [8]. 9 A seventh category—retirement—applies to projects that are being retired. We are only dealing with gross additions to generation in this paper because we are focusing on how projects transition from one status to the next, and retirements do not transition through the status categories in the same way as new generation projects.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
purposes of this paper, we define a project as a unique power generating plant that could be completed independently of any other units. In making the individual power generating plant the unit followed through time, we diverge from the definition of project adopted by RDI. RDI often groups multiple generating plants under the same project identifier, as long as the development of these plants is planned to be completed simultaneously. However, we found instances in which generating plants that were grouped together subsequently diverged in the timing of their development. Therefore, we separated multiple-unit projects into their individual generating units.
1407
Table 1 Taxonomy of project phases Project phase
Description
Early development Advanced development
Initially proposed project Formal requirements of construction have been met including receipt of permits, securing of turbines, and/or secured financing or output sales. These soft development costs range from $1 to $3 million Ground broken, construction in progress Generator is online and producing power Project is delayed for indefinite period Project is canceled
Under construction Operating Tabled Canceled
3.2. Description of new investment Projects progress through the stages of early development, advanced development, construction, and finally operation.10 The various stages of development are set out and defined in Table 1. Projects may also be tabled or canceled at any point in the process. Because we are following projects through their lives, we do not include in our analysis projects that in the initial month—January 2000—were already operating or permanently canceled. This amounted to deletion of 14 canceled projects and 16 operating projects carried over from previous months. The adjustments and deletions described above resulted in 1907 unique projects in North America that we follow over the 30-month interval. Of all new projects, 77% are owned by non-utility companies and 23% by utilities, with considerable variation across jurisdictions ranging from 90% of new projects being non-utility in California to 79% of new projects being utility-affiliated in Quebec.11 About 71% of all new projects are of the combustion turbine or combined cycle types and this accounts for 78% of the entire generating capacity of all new projects. Approximately 80% of these combined cycle and combustion turbine projects are owned by non-utility companies. Non-utility development is 10
In the NewGen database, the advanced development status is assigned to projects that meet two or more of the following criteria:
A power purchase agreement (PPA) for a large portion of the output has been signed with a marketer that is not an affiliate of the developer.
Financing has closed or notification of an expected closing in three months has been received.
Turbines for the project have been secured. The siting permit (often called the Certificate of Public Need and
Necessity) and the air permit have been obtained, or the acquisition of these licenses is imminent. Strong local support is indicated or there is no visible local opposition. The project involves re-powering with no emissions increase (such as nuclear re-power projects or air inlet cooling) or is a project with no emissions byproducts (such as wind or solar power). Projects in advanced development status may still not materialize due to the developer’s decision to abandon the project, or other unexpected developments.
11 Detailed descriptive statistics and cross-tabulations of the data are contained in Ludwigson et al. [8].
responsible for the bulk of renewable fuel generation (portfolio standards). Non-utility companies account for 86% of the projects involving geothermal, solar, waste, or wind, and 52% of hydroelectric projects. Non-utilities account for about two thirds of the coal plants under development in the sample. Natural gas is the predominant fuel source in new power plant development. The combined cycle and combustion turbine categories, accounting for 78% of generating capacity under development, use natural gas as fuel source almost exclusively. Nonetheless, coal plants make up about 10% of all plants under development and (although not shown in the tables in this paper) these are predominantly in regions that already generate large proportions of their power using coal. Hence a regional fuel mix that follows traditional patterns, albeit in different proportions, is still projected for the future. In addition, most of the coal plants under development are in states that have not restructured retail markets. Overall, as shown in Table 2, the majority of development projects—including investments by both utilities and non-utilities—have been in states that restructured. This includes California, which has recently suspended retail choice, but still has a centralized wholesale market run now by the California Independent System Operator. When we include states that delayed restructuring—states that passed some sort of restructuring legislation, but then delayed its implementation—61% of all projects under development have been in states that took some actions that signaled restructuring plans, compared to states that have been inactive entirely. In part this may be explained by the fact that the states taking restructuring actions generally had higher retail rates to begin with. For this reason, the value of additional units was greater in these states than in the inactive states. However, this is not the whole story. The ability of private generators to make money depends on the restructuring status, because a state that allows retail competition will have more buyers of power than a state that still relies on a monopoly utility structure at the retail level. In addition, state actions to restructure signal intent on the part of state legislators and regulators to develop competitive electricity markets,
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1408
Table 2 Projects by plant type, restructuring status and owner entity
Table 3 Status of new projects at end of sample period
Plant type
Status
Non-utility Active
Delayed
Not active
Owner entity type and EIA restructuring status CC/cogen 18 5 18 CT/cogen 21 3 16 Coal 16 16 36 Coal cogen 3 1 1 Comb cycle 222 50 136 Combust turb 225 34 194 Geothermal 1 1 Hydro 3 3 7 Intern combust 2 3 4 Nuclear 7 Other boiler 4 1 6 Solar 4 1 Waste 12 4 Wind 46 13 24
Suspended
3 4 1 36 115 5 1 6 5 8 4 5
Utility Active CC/cogen CT/cogen Coal Coal cogen Comb cycle Combust turb Geothermal Hydro Intern combust Nuclear Other boiler Solar Waste Wind
Delayed
2 1 6
1
16 61
5 10
1
1
11 5 6
1 1
2
1
Not active
Suspended
3 4 23 1 47 120
6 11
2 10 9 5
1 1 1 2
1 11
making these states more desirable for non-utility investors. Table 2 illustrates this point in that the bulk of development undertaken by utilities as opposed to nonutilities is in states that took no restructuring steps. Specifically, utilities accounted for 35% of total projects under development in states that did not pursue restructuring, but only accounted for 14% of projects in states that were either actively pursuing restructuring or had delayed their restructuring implementation. 3.3. Project survival Table 3 shows a tabulation of the status of the projects at the end of the sample period. Of the 1907 projects proposed, 413 were canceled and 219 tabled, while 525 of the projects were already being operated and another 267 were under construction. We cannot determine to what extent these relative magnitudes reflect a steady state, but we have reason to believe that subsequent to this time period, there were a greater proportion of projects that transitioned to delayed or canceled status. This was caused in part by a general contraction in the electricity industry
Owner entity type Non-utility
Early develop Advan develop Under constr Operating Tabled Canceled Total
Total Utility
274 268 211 338 196 180
68 73 56 187 23 33
342 341 267 525 219 213
1467
440
1907
following Enron’s demise and coincident with a general downturn in the economy. It has also been attributed in part to a natural boom-bust cycle. In particular, we know there were a large number of cancellations in California subsequent to the sample period and we will discuss this in greater detail below. We have calculated life tables for all new generation projects. These tables show the empirical hazard function—the probability that a project will die in the following time interval—where the project ‘‘died’’ if it were canceled and it continues to ‘‘live’’ otherwise. Table 4 shows the empirical hazard functions and cumulative failure rates for all projects disaggregated by utility and non-utility generation. The cumulative hazard rates corresponding to Table 4 are plotted in Fig. 1.12 It is clear from inspection of the figure that the cumulative failure rate for non-utility projects rises more rapidly than for utility projects. During most intervals the hazard rate is higher for the non-utility projects as well. The results of likelihood ratio and logrank tests lead us to conclude that the survivor functions are statistically different for these two groups of projects. This observation is consistent with our conjectures and with the theory of real options in general. Utilities, particularly in states that have not restructured their retail markets face far less uncertainty with respect to the return on their investment than do non-utility plants in general. As a result, we would expect utilities to purchase fewer options to build in the form of multiple applications for a single plant investment budget than their non-utility counterparts. The observed lower and flatter hazard rate for utilities is consistent with this explanation. We also calculated life tables for new generation projects in the major regional transmission organizations.13 These 12 See Cox [9] and Kalbfleisch and Prentice [10] for the details of calculating the life tables and hazard rates. An alternative to the simple life tables is the Kaplan–Meier [11] estimator of the survivor function; in the present analysis, the results are not at all sensitive to the estimator employed and the Kaplan–Meier estimates mirror those presented in the more easily understood life tables. 13 Ideally the analysis would be performed at the state level, but there are not enough observations in most individual states to perform the statistical analysis. Grouping the projects into the major transmission organizations that form their market areas provides enough data to
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1409
Table 4 Life tables for new projects: utility and non-utility generation Interval
Non-utility 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Beg. total
Cum. failure
Std. error
Hazard
Std. error
generation 2 1467 4 1466 6 1464 8 1463 10 1462 12 1457 14 1451 16 1448 18 1438 20 1427 22 1417 24 1396 26 1364 28 1338 30 1315
0.0007 0.0020 0.0027 0.0034 0.0068 0.0109 0.0130 0.0198 0.0273 0.0341 0.0484 0.0702 0.0879 0.1036 0.1227
0.0007 0.0012 0.0014 0.0015 0.0021 0.0027 0.0030 0.0036 0.0043 0.0047 0.0056 0.0067 0.0074 0.0080 0.0086
0.0003 0.0007 0.0003 0.0003 0.0017 0.0021 0.0010 0.0035 0.0038 0.0035 0.0075 0.0116 0.0096 0.0087 0.0108
0.0003 0.0005 0.0003 0.0003 0.0008 0.0008 0.0006 0.0011 0.0012 0.0011 0.0016 0.0020 0.0019 0.0018 0.0020
0.0023 0.0045 0.0114 0.0159 0.0182 0.0273 0.0455 0.0545 0.0705 0.0750
0.0023 0.0032 0.0051 0.0060 0.0064 0.0078 0.0099 0.0108 0.0122 0.0126
0.0011 0.0011 0.0034 0.0023 0.0012 0.0047 0.0094 0.0048 0.0085 0.0025
0.0011 0.0011 0.0020 0.0016 0.0012 0.0023 0.0033 0.0024 0.0032 0.0017
Utility generation 4 6 10 12 14 16 16 18 18 20 20 22 22 24 24 26 26 28 28 30
440 439 438 435 433 432 428 420 416 409
Likelihood-ratio test statistic of homogeneity: w2 ð1Þ ¼ 8:0153809, P ¼ 0:00463817. Logrank test of homogeneity: w2 ð1Þ ¼ 7:65, P ¼ 0:0057.
results are presented in Table 5 and graphically in Fig. 2. It is interesting to note the marked difference between California and the other major regional transmission organizations. The sample period encompasses the period of California’s electricity crisis and subsequent state actions that increased the level of regulatory uncertainty dramatically. We discuss this in more detail below. 3.4. Modeling the hazard rate Failure data of the type analyzed in this paper can be analyzed using a variety of statistical techniques. The life tables and empirical hazard rates set out above are the result of simple arithmetic; they make no statistical assumptions, which is good, but they also do not allow us to model directly the effect of covariates on the survival process because the hazard rate is estimated separately for each group of projects. In this section we apply a semi-parametric method to model the hazard rate. This method is based on a survival (footnote continued) perform the analysis; states in each group are similar in their state-level restructuring programs.
Fig. 1. Cumulative hazard rates for new projects by owner entity.
R1 process with a survival function SðtÞ ¼ t f ðxÞ dx where f ðxÞ is the density function of survival time. The instantaneous risk of failure given that a project has survived to time t, called the hazard function, is hðgÞ ¼ f ðtÞ=SðtÞ, and this can be modeled as a function of covariates using Cox’s [9] proportional hazards model.14 The hazard function is hðt; x; bÞ ¼ h0 ðtÞ expðb0 xÞ where h0 ðtÞ represents the unspecified baseline hazard function corresponding to the case where x is the zero vector. The coefficients b represent how the independent variables x multiplicatively affect the hazard. The parameters in this model are estimated by maximum likelihood and hypotheses are tested using standard Wald tests. The regressors in the hazard rate regression included dummy variables for restructuring status, non-utility projects, and the proportion of the project’s output that would be ‘‘merchant power’’, meaning that it would have to be sold in the spot power market and was not pre-sold under long-term contract.15 Table 6 displays the parameter estimates of a Cox hazard regression. Convergence of the model’s likelihood was achieved after six iterations and overall the model is statistically significant. Before discussing the individual regression coefficients it is useful to recall that they represent the multiplicative effect of each independent variable on the hazard rate, so coefficients
14 The advantage of using the proportional hazards model is that the baseline distribution of survival times need not be specified parametrically and the hazard can be modeled directly. As a robustness check, we also estimated parametric survival models that quantify life length—specifically the Weibull and lognormal—and found no difference in the qualitative results. 15 Projects often pre-sell a portion of their future output through longterm contracts, with the remaining generation capacity to be sold in the spot market. The data on the proportion of merchant power and whether or not the project is a utility or a non-utility project is part of the RDI power plant investment database. The data on restructuring status was obtained directly from the Energy Information Administration [12]; the remaining variables are we obtained from the NewGen database as discussed in the data description section above.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1410 Table 5 Life tables for new projects by RTO group Interval
rtogroup CA ISO 11 12 14 15 15 16 17 18 18 19 19 20 20 21 21 22 22 23 23 24 25 26 26 27 27 28 28 29 29 30
Beg. total
Cum. failure
Std. error
Hazard
Std. error
215 210 208 207 205 204 203 202 193 188 186 172 168 167 161
0.0233 0.0326 0.0372 0.0465 0.0512 0.0558 0.0605 0.1023 0.1256 0.1349 0.2000 0.2186 0.2233 0.2512 0.2791
0.0103 0.0121 0.0129 0.0144 0.0150 0.0157 0.0163 0.0207 0.0226 0.0233 0.0273 0.0282 0.0284 0.0296 0.0306
0.0235 0.0096 0.0048 0.0097 0.0049 0.0049 0.0049 0.0456 0.0262 0.0107 0.0782 0.0235 0.0060 0.0366 0.0380
0.0105 0.0068 0.0048 0.0069 0.0049 0.0049 0.0049 0.0152 0.0117 0.0076 0.0209 0.0118 0.0060 0.0149 0.0155
rtogroup ERCOT 8 9 129 22 23 128 23 24 126 25 26 121 28 29 118
0.0078 0.0233 0.0620 0.0853 0.0930
0.0077 0.0133 0.0212 0.0246 0.0256
0.0078 0.0157 0.0405 0.0251 0.0085
0.0078 0.0111 0.0181 0.0145 0.0085
rtogroup MISO 8 9 9 10 15 16 17 18 20 21 21 22 22 23 23 24 25 26 26 27 27 28 28 29 29 30
332 331 329 328 325 324 323 318 315 312 308 307 306
0.0030 0.0090 0.0120 0.0211 0.0241 0.0271 0.0422 0.0512 0.0602 0.0723 0.0753 0.0783 0.0813
0.0030 0.0052 0.0060 0.0079 0.0084 0.0089 0.0110 0.0121 0.0131 0.0142 0.0145 0.0147 0.0150
0.0030 0.0061 0.0030 0.0092 0.0031 0.0031 0.0156 0.0095 0.0096 0.0129 0.0033 0.0033 0.0033
0.0030 0.0043 0.0030 0.0053 0.0031 0.0031 0.0070 0.0055 0.0055 0.0065 0.0033 0.0033 0.0033
rtogroup PJM 2 3 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 25 26 26 27 27 28 28 29
339 338 337 336 332 331 330 329 328 327 323 321 318 312 310 306 305
0.0029 0.0059 0.0088 0.0206 0.0236 0.0265 0.0295 0.0324 0.0354 0.0472 0.0531 0.0619 0.0796 0.0855 0.0973 0.1003 0.1121
0.0029 0.0042 0.0051 0.0077 0.0082 0.0087 0.0092 0.0096 0.0100 0.0115 0.0122 0.0131 0.0147 0.0152 0.0161 0.0163 0.0171
0.0030 0.0030 0.0030 0.0120 0.0030 0.0030 0.0030 0.0030 0.0031 0.0123 0.0062 0.0094 0.0190 0.0064 0.0130 0.0033 0.0132
0.0030 0.0030 0.0030 0.0060 0.0030 0.0030 0.0030 0.0030 0.0031 0.0062 0.0044 0.0054 0.0078 0.0045 0.0065 0.0033 0.0066
Likelihood-ratio test statistic of homogeneity: w2 ð3Þ ¼ 39:555176, P ¼ 1:324e 08. Logrank test of homogeneity: w2 ð3Þ ¼ 50:81, Pr 4w2 ¼ 0:0000.
Fig. 2. Cumulative hazard rates for new projects by RTO grouping.
Table 6 Hazard regression Haz. ratio
Robust Std. err.
Haz. ratio
Robust Std. err.
Non-utility Restructuring Active Delayed Canceled Merchant (%) Combust turb Coal
1.811468
.3784981
1.959942
.4177448
1.083616 1.055826 3.371199 1.026478
.1919256 .3120933 .6099967 .0069154
1.058059 1.195648 2.957074 1.026404 1.988257 .5341334
.1883613 .3539154 .5283339 .0066883 .2936597 .2733636
Log likelihood Wald w2 ð5Þ Prob4w2
1464.5039 88.90 0.0000
1450.8163 103.71 0.0000
less than unity lower the hazard while those greater than unity increase the hazard. The results of the Cox regression show that projects in a jurisdiction that canceled restructuring—namely California—have a statistically higher hazard rate than the baseline hazard; projects in jurisdictions with active or delayed restructuring have hazards that are not statistically different from the baseline hazard of no restructuring. The coefficient on the percentage of merchant power was statistically larger than unity, indicating that projects with a lower percentage of locked-in sales had a significantly higher probability of failure. Non-utility projects also had a statistically higher hazard rate. Overall, the results of the hazard regression are consistent with the sequential abandonment of projects over time. In an auxiliary regression we included dummy variables for combustion turbine and coal power plant types and found that combustion turbine plants had a statistically larger hazard, holding other factors constant. Again, this is consistent with the real options nature of power plant development under regulatory restructuring.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
3.5. A closer look at California and Texas As discussed above, the regulatory environment in which permitting takes place as well as the stability of the regulations in a state can have a great influence on the pattern of investment in new generating facilities. To illustrate how different regulatory environments can lead to different investment patterns we focus on California and Texas—two states with very different regulatory environments. At the time of our review there were significant differences in regulation of new power plant development across these states. Overall, California’s regulatory process to approve the siting of a new power plant was more complex and often required more stringent environmental mitigation efforts. The overall process of receiving regulatory approvals is more complex in California than in Texas. The process in both states includes application and approval by state and local agencies to ensure compliance with environmental, land use, and other requirements. However, the process in California requires application and approval by the California Energy Commission, a state energy agency that conducts its own analysis of whether the benefits of additional electricity supplies outweigh any potential negative environmental or other effects. In addition, the regulatory process in California can also require new power plants to install more stringent air pollution mitigation equipment. Extensive areas in California have witnessed high levels of air pollution over many years and face more stringent federal oversight. As a result, federal air quality regulations generally require new power plants to undertake efforts to mitigate the pollution that they introduce into these already polluted areas. Exacerbating these federal requirements, the state of California has, in recent years, sought to require more stringent levels of air pollution mitigation than what has been required in Texas and many other states. Negotiating what types of air pollution mitigation efforts may be required can lengthen the time for some projects to be approved. Furthermore, in California the California Energy Commission can designate members of the public or other stakeholders as intervenors, allowing them to become formal participants in the regulatory process with the ability to request data, file motions and take other formal actions. Of the 72 siting cases filed from 1995 to 2001, 39 had one or more intervenors. The regulatory process in Texas also allows some members of the public to request contested evidentiary hearings, during which these members can present evidence. However, in only 15 of 84 of the siting cases filed since the law allowing these proceedings was enacted were these types of hearings held. The emergence of significant local opposition to a new plant is a significant factor in receiving the necessary approvals, delaying regulatory decisions in many cases. As a result, developers seek locations where their project will receive local support.16 16
For example, communities in Texas generally welcome new natural gas-fired power plants because they add to the communities’ tax base and pose few environmental concerns.
1411
It would be difficult to find two more disparate regulatory regimes than those of California and Texas with respect to siting new power plants. This difference is evidenced in the data: although the two states both have similar process for approving applications to build power plants, this process took 6 months longer on average in California (14 months) than in Texas (8 months) in the period from 1995 to 2001 [13]. Perhaps of more significance to developers is that California’s process was much more variable, with 5 of 21 medium to large-scale projects taking 18 months or longer. The uncertainty about the outcome of the permitting process also appears to be much greater in California than in Texas. In discussions with several large energy companies and several large investment banks, two of the authors repeatedly heard that California is a very difficult place to site a power plant because of the strength of opposition at the local level and the extent to which the permitting process in California can be held up by such opposition. In addition, Texas ‘‘socializes’’ the costs of interconnection, while California makes developers pay for all the interconnection costs associated with their new plant. If our conjectures are correct, then the data should show predictably different investment patterns in the two states. A look at the data indicate that this is the case. Table 7 shows the composition of power plant types in California and Texas. In California 58.6% of proposed projects are combustion turbines as compared to 17.05% in Texas. In contrast, 38.76% of projects in Texas are combined cycle plants as compared to 19.53% in California. This different pattern of investment is not inconsistent with our conjecture that greater uncertainty will cause investors to choose power plant types with lower capital to generating capacity ratios.17 Table 8 displays the composition of projects across development phases in California and Texas, and shows an 51 50 eventual observed success rate of 215 in California and 129 in Texas. This is a dramatic difference, but may not be reflective of a steady state. During the sample period, California underwent, along with the rest of the western states, an electricity crisis, largely caused by insufficient supply and a severe drought. During that crisis, prices were extremely high and the state responded by streamlining and expediting their permitting process for certain new generating plants. In response, many companies proposed a number of power plants in California and began the permitting process. Subsequent actions by the state to (1) suspend retail competition, (2) renegotiate long-term contracts signed during the crisis and (3) to demand
17 There are other likely factors in the explanation of this pattern, most notably, the greater prevalence of hydroelectric generation in California and nearby states. In high water years, there has typically been enough hydroelectricity generated to drive a number of base-load generators out of the market for significant amounts of time. This increases the number of years to expected payoff, which favors investments in combustion turbines (peaker units) to some extent.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1412
4. Conclusions
Table 7 Distribution of plant types in Texas and California Plant type
Texas
California
Freq CC/cogen CT/cogen Coal Comb cycle Combust turb Geothermal Hydro Intern combust Nuclear Other boiler Solar Waste Wind
9 10 1 50 22 0 0 0 2 3 0 12 20
Total
(%)
Freq
(%)
6.98 7.75 0.78 38.76 17.05 0 0 0 1.55 2.33 0 9.30 15.50
3 4 1 42 126 5 2 7 1 7 8 4 5
1.40 1.86 0.47 19.53 58.60 2.33 0.93 3.26 0.47 3.26 3.72 1.86 2.33
129
215
Table 8 Project phases in California and Texas Phase
Year/month 2000/01 2000/06 2001/01 2001/06 2002/01 2002/06
California Advan develop 2 Canceled 0 Early develop 19 Operating 0 Tabled 0 Under constr 3
8 0 13 1 1 3
4 5 26 2 3 7
16 10 55 10 7 23
25 29 47 44 32 16
26 60 22 51 40 16
Total
24
26
47
121
193
215
Texas Phase
Year/month 2000/01 2000/06 2001/01 2001/06 2002/01 2002/06
Advan develop 7 Canceled 0 Early develop 15 Operating 0 Tabled 1 Under constr 15
15 0 16 5 2 20
13 1 36 12 2 25
17 1 35 23 2 27
19 8 28 43 8 17
14 12 17 50 9 27
Total
58
89
105
123
129
38
refunds from generators who sold power in the state during the crisis caused many developers to flee.18 It is our understanding that almost all power plants proposed during this time were ultimately abandoned.
18 California’s suspension of retail competition was done in order to ensure that the state utilities could charge prices high enough to recover the costs of power purchased by the state at high prices during the height of the electricity crisis. These prices are considerably higher than current or expected future wholesale prices in the state.
Power plant investment is higher in states that have restructured electricity markets than in states that have taken no restructuring actions. Development is also more prevalent in areas of the country with a robust wholesale market infrastructure. Ownership of new power plants also differs across states, with non-utility companies accounting for the bulk of new power plants in states taking restructuring actions, while utilities still have a strong or dominant role in new development in states that have not restructured at all. States’ decisions to implement retail competition result in more investment in new power plants. These patterns indicate that state regulatory actions are an important determinant of how well restructuring at the national level will ultimately work. The bulk of the potential benefits of restructuring the industry will come from improvements in efficiency of wholesale generation and sale of electricity and this depends critically on the ability of new companies to enter and exit. In addition, reliability of the electricity system depends critically on the ability of companies to make timely investments in power plants and other infrastructure. However, non-utility companies are far less likely to make the investments necessary to achieve these benefits in uncertain regulatory environments. References [1] Brennan TJ. Provincial and federal roles in facilitating electricity competition: legal and economic perspectives. In: Walls WD, editor. Regional transmission organizations: restructuring electricity transmission in Canada. Calgary: The Van Horne Institute; 2003. p. 20–40 (Chapter 2). [2] Feldberg P, Jenkins M. Reciprocity, regional transmission organizations, and standard market design: some implications for Canadian participation in North American wholesale electricity trade. In: Walls WD, editor. Regional transmission organizations: restructuring electricity transmission in Canada. Calgary: The Van Horne Institute; 2003. p. 60–75 (Chapter 3). [3] U.S. General Accounting Office. Lessons learned from restructuring: transition to competitive markets underway, but full benefits will take time and effort to achieve. Technical Report GAO-03-271, GAO, Washington, DC; December 2002. [4] Joskow PL. The difficult transition to competitive electricity markets in the U.S. Working paper, Department of Applied Economics, University of Cambridge, UK; 2003. [5] Brennan TJ, Palmer KL, Salvador AM. Alternating currents: electricity markets and public policy resources for the future. Washington DC; 2002. [6] Borenstein S. The trouble with electricity markets: understanding California’s restructuring disaster. J Econ Perspect 2002;16(1): 191–211. [7] Joskow P, Kahn E. A quantitative analysis of pricing behaviour in California’s wholesale electricity market during summer 2000. Energy J 2002;23(4):1–35. [8] Ludwigson J, Rusco FW, Walls WD. Regulatory uncertainty and the development of new electric generation: competing risks with option value. In: Integrating the energy markets in North America: issues and problems, terms and conditions. Proceedings of the 23rd IAEE North American Conference on CD-ROM. International Association for Energy Economics; 2003. p. 1–10.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413 [9] Cox DR. Regression models and life tables. J R Stat Soc 1972;B34:187–220. [10] Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: Wiley; 1980. [11] Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457–81.
1413
[12] U.S. Energy Information Administration. Status of state electric industry restructuring activity. hhttp://www.eia.doe.gov/cneaf/electricity/ chg-str/restructure.pdfi; February 2003. [13] U.S. General Accounting Office. Restructured electricity markets: three states’ experiences in adding generating capacity. Technical Report GAO-02-427, GAO, Washington, DC, May 2002.
Energy 32 (2007) 1403–1413 www.elsevier.com/locate/energy
Power plant investment in restructured markets$ W.D. Wallsa,, Frank W. Ruscob, Jon Ludwigsonc a
Department of Economics, University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada T2N 1N4 b U.S. General Accounting Office, Washington, DC, USA c U.S. General Accounting Office, Denver, CO, USA Received 9 February 2006
Abstract Investment opportunities in electric power generation have changed dramatically since electricity industry restructuring first began. In contrast to regulated utilities adding capacity in line with central planning and regulation, power plant investment is now more often made by independent companies who market power across multiple utility jurisdictions. Because the regulatory approval process is long and outcomes uncertain, developers often plan multiple options for a given development budget. As more information is revealed about the future prospects at different sites, options are abandoned sequentially until only projects that will be completed remain. In this paper we evaluate the decision to build new generation facilities in a changing and uncertain regulatory environment. We estimate hazard rates for new power plant projects using a database of generation projects in North America and examine in greater detail the development patterns in California and Texas, two states with very different regulatory regimes. We find that regulatory uncertainty significantly affects the pattern of development in the electric power generation industry. r 2006 Elsevier Ltd. All rights reserved. JEL classification: L94; Q4; L2 Keywords: Power plant investment; Electricity market restructuring; Regulatory uncertainty
1. Introduction The electricity industry in the United States is in the midst of fundamental change as a result of federal and state efforts to restructure the industry, thereby introducing and increasing the intensity of competition in wholesale and retail markets. Unlike a number of other countries, the U.S. restructuring effort has been hampered by divided regulatory jurisdictions.1 The federal government has $ An earlier version of this paper was presented at the American Economic Association meeting, Washington DC, January 2003, and at the International Association for Energy Economics North American meeting, Mexico City, October 2003. We would like to thank Chris Knittel, other conference participants, and three anonymous referees for helpful comments on earlier versions of this paper. The views expressed in this paper are solely those of the authors and are not to be attributed to their employers. Corresponding author. E-mail address: [email protected] (W.D. Walls). 1 See, for example, the discussion of Brennan [1] for legal and economic perspectives on the roles of different levels of government in a federation.
0360-5442/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.energy.2006.11.002
jurisdiction over wholesale electricity sales and movement because electricity at the wholesale level crosses state borders and therefore qualifies as interstate commerce.2 Retail markets, on the other hand, are under individual states’ jurisdiction. This historical fact has led to a patchwork of different rules and regulations governing electricity markets. For example, while wholesale markets have largely been restructured to allow the emergence of a large and active private sector of electricity generating companies, retail restructuring has been spotty and incomplete. Currently, according to a recent GAO [3] report, 24 states and the District of Columbia have enacted legislation or issued regulatory orders to open their retail markets to competition. However, of these, seven states have either delayed or suspended implementation of 2 In addition to interstate trade, there is substantial Canada–U.S. trade in electricity that adds another level of institutional complexity. See Feldberg and Jenkins [2] for a brief legal and institutional analysis of this issue.
ARTICLE IN PRESS 1404
W.D. Walls et al. / Energy 32 (2007) 1403–1413
restructuring and the remaining 26 states have not yet taken any steps to introduce competition at the retail level.3 The result of this divided jurisdiction and diverse approaches to restructuring has been to introduce a great deal of regulatory uncertainty into the market. This uncertainty is having an impact on the development of new generation facilities. One key feature of restructuring has been a move away from centralized planning, wherein utilities, in conjunction with state public utilities commissions, planned for development of new generating capacity and transmission upgrades in order to meet expected increases in future demand. In its place, a decentralized process of development and investment decisions—largely by non-utility companies—is evolving. Unlike the rate-regulated regime of the past, the development and investment plans of these myriad companies are not subject to approval of public utilities commissions, nor are they coordinated in any way by a central body. This is particularly true in states that have aggressively pursued retail restructuring—sometimes requiring or encouraging their utilities to divest generating resources—but it is also the case in other states to the degree that non-utilities find it attractive to develop new generating resources in those states. Although under restructuring states will no longer oversee the entire process of development and investment in new generating capacity, state entities still wield significant power to influence investments through licensing and permitting processes, through the terms of interconnection agreements, and more generally, through state decisions regarding whether and how far to pursue restructuring of their retail markets. Specifically, state and local agencies responsible for air and water quality and land use decisions must grant approval for companies to begin construction or operation of new power plants. The role of these agencies is to ensure that any new development is in compliance with relevant laws, ordinances, and regulations. There is considerable variation across states in the administration of the development process and thereby in the costs developers must incur to gain approval from state and local entities. 4 Firms balance expected return against the effect of new investment on the risk of their portfolios of projects. In order to mitigate the risk across regulatory jurisdictions 3 See Joskow [4] and Brennan et al. [5] for a discussion of these and other issues surrounding restructuring. See Borenstein [6] and Joskow and Kahn [7] for non-technical examinations of the California market. 4 Federal environmental laws and regulations, as well as laws protecting endangered species, also play a role in determining where and how new power plants are built. For example, proposed new power plants in any area that is not in compliance with EPA air quality regulations are subject to ‘‘new source review,’’ requiring plant owners to purchase or otherwise acquire air emission credits equal to or in excess of their planned emissions. Often the new source review permits are issued by state agencies that have gained approval from the EPA to grant such permits. In the event that a proposed new power plant might impinge on the habitat of an endangered species, developers must also get approval from other federal and state agencies.
and over time, firms may diversify their investments across regions and states, and across power plant types and fuel sources. In addition, because the regulatory approval process is long and outcomes uncertain, firms may plan multiple options for a given development budget. As more information is revealed about the future prospects at different sites, options are abandoned sequentially until eventually only projects that will be completed remain. The costs of early development—the so-called soft development costs, incurred prior to breaking ground for construction—are a small fraction of total costs to build, but they are significant in magnitude running between several hundred thousand and many millions of dollars depending on the characteristics of the site and the state and local requirements. These soft development costs reflect the cost to developers of acquiring an option to build a power plant. In addition to the development costs associated with acquiring regulatory approval, new power plants must be interconnected with the transmission grid, frequently requiring costly upgrades to the system to maintain reliability. The terms under which these new power plants are allowed to interconnect and the distribution of the costs of upgrades is another critical factor that determines where and when power plants are built. Again, there is considerable variation across states in the interconnection costs, and a developer’s share of these interconnection costs can run from a few hundred thousand to tens of millions of dollars, depending on the characteristics of the existing transmission system and on how the costs are assessed. Many hazards lurk in the regulatory arena. Because the development process can be long—running to many years in some cases—regulatory and market conditions may change considerably, causing developers to reassess the relative merits of each of their projects. In addition to regulatory uncertainty, energy prices have been volatile both across time and across regions, adding to the risk of committing development resources. This underlying market risk is increased by long state and federal approval processes and by regulatory uncertainty and unstable market rules because unforeseen events may occur during the course of development that change price or load forecasts. We present in this paper an empirical examination of the overall investment patterns in North America, with a closer look at two states—California and Texas—in the context of those states’ very different regulatory regimes. We evaluate the permitting process for individual firms in the context of the decision to build power plants in an uncertain environment. In the next section we explore the decision-making process of power plant developers and develop a series of conjectures about the expected pattern of investment. In the remainder of the paper we examine— using a micro dataset of power plant development and investment—the recent experiences in North America overall, and of California and Texas, specifically, in attracting new generation development. In the latter case,
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
we will examine the types of generators being built, the actions of these states vis-a´-vis restructuring, and the impacts of state actions on variations in the rate at which plants transition between different phases of the development project and the rate at which they are completed (or fail). 2. Uncertainty affects power plant development We consider firms contemplating irreversible investments to build new electricity generating units under a regime of regulatory and other uncertainty. Specifically, there are three principal areas of uncertainty that influence new generation development. First, regulatory uncertainty with respect to both the timing and pace of restructuring varying across states creates a complicated regulatory environment for investors. Second, uncertainty about future demand—and therefore electricity prices—is a critical feature for merchant power producers who plan to sell a significant proportion of their power in spot or other relatively short-term contracts. Third, the permitting process for siting new generating plants varies a great deal across states and is fraught with uncertainty about the length of time, cost, and ultimate success of applications for approval to build. Because permitting takes a significant amount of time, the firm will generally receive updated information about forward prices and the regulatory regime during the time it takes to negotiate the permitting process. Also, because multiple firms often compete to build generating units in a given market, conditions may change considerably over the period of time in which a firm is acquiring the necessary permits to build. In this context, each stage of the permitting process represents an option to pursue the project and can be evaluated in light of contemporaneous information about forward prices and other relevant factors before deciding whether to invest in the next stage of permitting, and ultimately whether or not to build the plant. A full model of a firm’s investment decision in which the firm has multiple options, each with potentially different costs of development, degree of regulatory uncertainty, and forward price expectations has proven to be intractable. However, we believe the intuition flowing from the standard real options literature as well as the theory of the firm is sufficiently clear to allow us to make a number of conjectures about a firm’s optimal choices in response to various changes in its exogenous environment. A series of such conjectures are made below and then compared against observed investment patterns in the following sections. The value of waiting to invest increases as the relative value of information that may be revealed by waiting increases. The firm is therefore more likely to wait to invest the more uncertainty exists, ceteris paribus. In the standard two-period model in the real options literature, firm’s only have the options of building in period 1 or waiting to get the information in period 2. In the choice to invest in new
1405
generating facilities, firms can respond to greater regulatory uncertainty in several ways. First, a firm can begin the permitting process in more than one jurisdiction, knowing that it will likely not build in each of these sites—the final choice of which units to build will depend on the realization of information during the permitting phase. This will naturally lead to less observed investment, holding forward price expectations constant, in regions in which uncertainty is greater because delays will be longer as firms wait longer to learn more. It also implies that firms will sometimes explore several options for each increment of investment, even investing time and financial resources to start the permitting process while they await information that will enable them to choose between competing options. Alternatively, firms that face greater regulatory uncertainty can vary the expected payoff period of their investment. This can be done contractually if the firm can find buyers willing to bear some of the risk by entering into long-term contracts to buy electricity, or it can vary the type of investment. With regard to the latter, a firm facing greater regulatory uncertainty over the life of an investment may choose investments that have a lower capital cost relative to the capacity of the generating unit. Firms compete with one another to build power plants in the right locations and at the right time to meet expected demand. Suitable locations generally require a nexus of access to fuel sources, transmission lines, and water for cooling. For example, developers of natural gas fired power plants—the predominant technology being built in recent years—look for sites with access to high-volume gas pipelines with excess capacity. Similarly, coal fired plants need access to rail lines, or direct access to coal at the source. Access to transmission suitable for interconnection is also critical for developers. The terms under which interconnection is approved vary a great deal across states and control areas as do the distribution of costs of upgrades. For example, in Texas, the upgrades required to interconnect new power plants are paid for through a surcharge on electricity sold to consumers, while in California the developers have been required to pay for any upgrades deemed necessary by the local transmission owner—generally a local utility company. The implications of different approaches to assessing interconnection costs on power plant location can be profound. When the costs are borne by consumers, developers can focus more on finding locations with lower development costs, easier access to fuel sources, and water for cooling. It is possible under these conditions that there will be overbuilding of transmission upgrades, because developers do not bear the costs of any negative externalities they impose on other grid users.5 In addition, this approach may lead to concentration of generating 5
These externalities will exist in as much as adding generating capacity at a point increases transmission congestion, thereby limiting incumbent generating plants’ outputs.
ARTICLE IN PRESS 1406
W.D. Walls et al. / Energy 32 (2007) 1403–1413
units at some distance from the load it serves, because land costs are lower and environmental issues, such as air quality, less prevalent on such sites. On the other hand, when developers bear the full cost of upgrades, they look for sites with lower interconnection costs, which—given the nature of the flow of electricity and of congestion in the existing transmission grid—may encourage development closer to the load it will serve. However, it may cause underbuilding of upgrades because the developers are not compensated for any positive externalities accruing to electricity consumers.6 Cooling water is essential for many of the most commonly built power plants. For the most part, this requires locating near a source of fresh water, although some designs allow the use of waste water.7 The volume of water drawn by power plants in the United States is quite large, ranking second only to agriculture. The watercooling processes used in most newer power plants lose less water to steam in the atmosphere than do older technologies—most of the water is recaptured and returned to its source. However, returning warmer water to a fresh source can have negative environmental implications, and these issues have led to controversy and delays or denials of permits in some cases. Finally, developers and investment bankers also prefer, other things equal, stable regulatory jurisdictions and clear market rules for trading electricity. Very few states have established and maintained clear paths to retail restructuring and this creates regulatory uncertainty. Specifically, only 17 of the 50 states and the District of Columbia have enacted and implemented legislation allowing consumers to choose their retail electricity provider. Even among the states with retail choice programs, the states have simultaneously reduced and frozen retail rates at levels that have discouraged retail competition. A lack of retail competition also feeds back into the development of new capacity by limiting the ability of developers to enter into long-term supplier contracts with large consumers or multiple retail sellers. The absence of multiple buyers in a given utility’s control area puts non-utility sellers at a strategic disadvantage in negotiating terms of electricity trades. Based on the discussion above, we make the following conjectures:
We expect to find multiple proposed projects for every project ultimately built and that these projects fall out of
6 From the perspective of efficiency, the ideal is to assess upgrade costs on developers in the amount equal to the negative externalities imposed on other transmission users, and assess costs on consumers in the amount that they benefit from the new capacity. In practice, there is a great deal of uncertainty about the value of either externality, but it is fairly clear that neither extreme—assessing all costs to consumers or all costs to developers—is optimal except under extreme conditions. 7 It is also possible for some units to use a largely air-cooled process, although this is much more costly than water cooling and has the disadvantage that it works least well during hot periods when, in many parts of the country, electricity demand is frequently at its peak.
consideration at different phases of the development process as information becomes available and firms narrow their options. We expect to find a distribution of power plant types that is consistent with access to fuel sources. This means that the fuel mix that has evolved historically will be expected to continue to vary by region according to the relative availability and cost of acquiring different fuels. With regard to interconnection costs, states that assess greater proportions of interconnection costs on developers will attract, ceteris paribus, less investment overall. The investment pattern will be toward more generators close to the load it will serve. States that ‘‘socialize’’ the cost of interconnections can expect greater investment and more geographic concentration of generators in areas with lower land or licensing costs—typically areas further from the load it will serve. Further, we expect to find proportionally more development by non-utility companies in states that have implemented retail restructuring, thereby signaling a commitment to developing competitive markets for electricity. In addition to differences in the amounts of development by utilities and non-utilities, we expect to find differences in the types of projects undertaken by these groups in various jurisdictions. For example, in states in which regulatory uncertainty is deemed to be more extreme, firms will tend to substitute away from capital intensive investments, such as combined cycle baseload generating plants, toward less capital intensive investments, such as combustion turbine peaker units. Finally, we expect that development will be more prevalent, again controlling for price expectations, in jurisdictions with more stable market rules and regulatory regimes, than in jurisdictions with greater regulatory uncertainty.
3. Empirical analysis 3.1. Data collection The data used in this paper are compiled primarily from monthly reports of the NewGen database published by RDI, a division of Platts.8 RDI gathers data on new generation projects from trade publications and state and federal data sources and reports the status of each of the projects they identify as of the reporting month. The status reports identify projects as being in one of six categories— proposed/early development, advanced development, under construction, operating, tabled, or canceled.9 For the 8 A more detailed description of the data collection exercise and of the assembled data set is contained in Ludwigson et al. [8]. 9 A seventh category—retirement—applies to projects that are being retired. We are only dealing with gross additions to generation in this paper because we are focusing on how projects transition from one status to the next, and retirements do not transition through the status categories in the same way as new generation projects.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
purposes of this paper, we define a project as a unique power generating plant that could be completed independently of any other units. In making the individual power generating plant the unit followed through time, we diverge from the definition of project adopted by RDI. RDI often groups multiple generating plants under the same project identifier, as long as the development of these plants is planned to be completed simultaneously. However, we found instances in which generating plants that were grouped together subsequently diverged in the timing of their development. Therefore, we separated multiple-unit projects into their individual generating units.
1407
Table 1 Taxonomy of project phases Project phase
Description
Early development Advanced development
Initially proposed project Formal requirements of construction have been met including receipt of permits, securing of turbines, and/or secured financing or output sales. These soft development costs range from $1 to $3 million Ground broken, construction in progress Generator is online and producing power Project is delayed for indefinite period Project is canceled
Under construction Operating Tabled Canceled
3.2. Description of new investment Projects progress through the stages of early development, advanced development, construction, and finally operation.10 The various stages of development are set out and defined in Table 1. Projects may also be tabled or canceled at any point in the process. Because we are following projects through their lives, we do not include in our analysis projects that in the initial month—January 2000—were already operating or permanently canceled. This amounted to deletion of 14 canceled projects and 16 operating projects carried over from previous months. The adjustments and deletions described above resulted in 1907 unique projects in North America that we follow over the 30-month interval. Of all new projects, 77% are owned by non-utility companies and 23% by utilities, with considerable variation across jurisdictions ranging from 90% of new projects being non-utility in California to 79% of new projects being utility-affiliated in Quebec.11 About 71% of all new projects are of the combustion turbine or combined cycle types and this accounts for 78% of the entire generating capacity of all new projects. Approximately 80% of these combined cycle and combustion turbine projects are owned by non-utility companies. Non-utility development is 10
In the NewGen database, the advanced development status is assigned to projects that meet two or more of the following criteria:
A power purchase agreement (PPA) for a large portion of the output has been signed with a marketer that is not an affiliate of the developer.
Financing has closed or notification of an expected closing in three months has been received.
Turbines for the project have been secured. The siting permit (often called the Certificate of Public Need and
Necessity) and the air permit have been obtained, or the acquisition of these licenses is imminent. Strong local support is indicated or there is no visible local opposition. The project involves re-powering with no emissions increase (such as nuclear re-power projects or air inlet cooling) or is a project with no emissions byproducts (such as wind or solar power). Projects in advanced development status may still not materialize due to the developer’s decision to abandon the project, or other unexpected developments.
11 Detailed descriptive statistics and cross-tabulations of the data are contained in Ludwigson et al. [8].
responsible for the bulk of renewable fuel generation (portfolio standards). Non-utility companies account for 86% of the projects involving geothermal, solar, waste, or wind, and 52% of hydroelectric projects. Non-utilities account for about two thirds of the coal plants under development in the sample. Natural gas is the predominant fuel source in new power plant development. The combined cycle and combustion turbine categories, accounting for 78% of generating capacity under development, use natural gas as fuel source almost exclusively. Nonetheless, coal plants make up about 10% of all plants under development and (although not shown in the tables in this paper) these are predominantly in regions that already generate large proportions of their power using coal. Hence a regional fuel mix that follows traditional patterns, albeit in different proportions, is still projected for the future. In addition, most of the coal plants under development are in states that have not restructured retail markets. Overall, as shown in Table 2, the majority of development projects—including investments by both utilities and non-utilities—have been in states that restructured. This includes California, which has recently suspended retail choice, but still has a centralized wholesale market run now by the California Independent System Operator. When we include states that delayed restructuring—states that passed some sort of restructuring legislation, but then delayed its implementation—61% of all projects under development have been in states that took some actions that signaled restructuring plans, compared to states that have been inactive entirely. In part this may be explained by the fact that the states taking restructuring actions generally had higher retail rates to begin with. For this reason, the value of additional units was greater in these states than in the inactive states. However, this is not the whole story. The ability of private generators to make money depends on the restructuring status, because a state that allows retail competition will have more buyers of power than a state that still relies on a monopoly utility structure at the retail level. In addition, state actions to restructure signal intent on the part of state legislators and regulators to develop competitive electricity markets,
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1408
Table 2 Projects by plant type, restructuring status and owner entity
Table 3 Status of new projects at end of sample period
Plant type
Status
Non-utility Active
Delayed
Not active
Owner entity type and EIA restructuring status CC/cogen 18 5 18 CT/cogen 21 3 16 Coal 16 16 36 Coal cogen 3 1 1 Comb cycle 222 50 136 Combust turb 225 34 194 Geothermal 1 1 Hydro 3 3 7 Intern combust 2 3 4 Nuclear 7 Other boiler 4 1 6 Solar 4 1 Waste 12 4 Wind 46 13 24
Suspended
3 4 1 36 115 5 1 6 5 8 4 5
Utility Active CC/cogen CT/cogen Coal Coal cogen Comb cycle Combust turb Geothermal Hydro Intern combust Nuclear Other boiler Solar Waste Wind
Delayed
2 1 6
1
16 61
5 10
1
1
11 5 6
1 1
2
1
Not active
Suspended
3 4 23 1 47 120
6 11
2 10 9 5
1 1 1 2
1 11
making these states more desirable for non-utility investors. Table 2 illustrates this point in that the bulk of development undertaken by utilities as opposed to nonutilities is in states that took no restructuring steps. Specifically, utilities accounted for 35% of total projects under development in states that did not pursue restructuring, but only accounted for 14% of projects in states that were either actively pursuing restructuring or had delayed their restructuring implementation. 3.3. Project survival Table 3 shows a tabulation of the status of the projects at the end of the sample period. Of the 1907 projects proposed, 413 were canceled and 219 tabled, while 525 of the projects were already being operated and another 267 were under construction. We cannot determine to what extent these relative magnitudes reflect a steady state, but we have reason to believe that subsequent to this time period, there were a greater proportion of projects that transitioned to delayed or canceled status. This was caused in part by a general contraction in the electricity industry
Owner entity type Non-utility
Early develop Advan develop Under constr Operating Tabled Canceled Total
Total Utility
274 268 211 338 196 180
68 73 56 187 23 33
342 341 267 525 219 213
1467
440
1907
following Enron’s demise and coincident with a general downturn in the economy. It has also been attributed in part to a natural boom-bust cycle. In particular, we know there were a large number of cancellations in California subsequent to the sample period and we will discuss this in greater detail below. We have calculated life tables for all new generation projects. These tables show the empirical hazard function—the probability that a project will die in the following time interval—where the project ‘‘died’’ if it were canceled and it continues to ‘‘live’’ otherwise. Table 4 shows the empirical hazard functions and cumulative failure rates for all projects disaggregated by utility and non-utility generation. The cumulative hazard rates corresponding to Table 4 are plotted in Fig. 1.12 It is clear from inspection of the figure that the cumulative failure rate for non-utility projects rises more rapidly than for utility projects. During most intervals the hazard rate is higher for the non-utility projects as well. The results of likelihood ratio and logrank tests lead us to conclude that the survivor functions are statistically different for these two groups of projects. This observation is consistent with our conjectures and with the theory of real options in general. Utilities, particularly in states that have not restructured their retail markets face far less uncertainty with respect to the return on their investment than do non-utility plants in general. As a result, we would expect utilities to purchase fewer options to build in the form of multiple applications for a single plant investment budget than their non-utility counterparts. The observed lower and flatter hazard rate for utilities is consistent with this explanation. We also calculated life tables for new generation projects in the major regional transmission organizations.13 These 12 See Cox [9] and Kalbfleisch and Prentice [10] for the details of calculating the life tables and hazard rates. An alternative to the simple life tables is the Kaplan–Meier [11] estimator of the survivor function; in the present analysis, the results are not at all sensitive to the estimator employed and the Kaplan–Meier estimates mirror those presented in the more easily understood life tables. 13 Ideally the analysis would be performed at the state level, but there are not enough observations in most individual states to perform the statistical analysis. Grouping the projects into the major transmission organizations that form their market areas provides enough data to
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1409
Table 4 Life tables for new projects: utility and non-utility generation Interval
Non-utility 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Beg. total
Cum. failure
Std. error
Hazard
Std. error
generation 2 1467 4 1466 6 1464 8 1463 10 1462 12 1457 14 1451 16 1448 18 1438 20 1427 22 1417 24 1396 26 1364 28 1338 30 1315
0.0007 0.0020 0.0027 0.0034 0.0068 0.0109 0.0130 0.0198 0.0273 0.0341 0.0484 0.0702 0.0879 0.1036 0.1227
0.0007 0.0012 0.0014 0.0015 0.0021 0.0027 0.0030 0.0036 0.0043 0.0047 0.0056 0.0067 0.0074 0.0080 0.0086
0.0003 0.0007 0.0003 0.0003 0.0017 0.0021 0.0010 0.0035 0.0038 0.0035 0.0075 0.0116 0.0096 0.0087 0.0108
0.0003 0.0005 0.0003 0.0003 0.0008 0.0008 0.0006 0.0011 0.0012 0.0011 0.0016 0.0020 0.0019 0.0018 0.0020
0.0023 0.0045 0.0114 0.0159 0.0182 0.0273 0.0455 0.0545 0.0705 0.0750
0.0023 0.0032 0.0051 0.0060 0.0064 0.0078 0.0099 0.0108 0.0122 0.0126
0.0011 0.0011 0.0034 0.0023 0.0012 0.0047 0.0094 0.0048 0.0085 0.0025
0.0011 0.0011 0.0020 0.0016 0.0012 0.0023 0.0033 0.0024 0.0032 0.0017
Utility generation 4 6 10 12 14 16 16 18 18 20 20 22 22 24 24 26 26 28 28 30
440 439 438 435 433 432 428 420 416 409
Likelihood-ratio test statistic of homogeneity: w2 ð1Þ ¼ 8:0153809, P ¼ 0:00463817. Logrank test of homogeneity: w2 ð1Þ ¼ 7:65, P ¼ 0:0057.
results are presented in Table 5 and graphically in Fig. 2. It is interesting to note the marked difference between California and the other major regional transmission organizations. The sample period encompasses the period of California’s electricity crisis and subsequent state actions that increased the level of regulatory uncertainty dramatically. We discuss this in more detail below. 3.4. Modeling the hazard rate Failure data of the type analyzed in this paper can be analyzed using a variety of statistical techniques. The life tables and empirical hazard rates set out above are the result of simple arithmetic; they make no statistical assumptions, which is good, but they also do not allow us to model directly the effect of covariates on the survival process because the hazard rate is estimated separately for each group of projects. In this section we apply a semi-parametric method to model the hazard rate. This method is based on a survival (footnote continued) perform the analysis; states in each group are similar in their state-level restructuring programs.
Fig. 1. Cumulative hazard rates for new projects by owner entity.
R1 process with a survival function SðtÞ ¼ t f ðxÞ dx where f ðxÞ is the density function of survival time. The instantaneous risk of failure given that a project has survived to time t, called the hazard function, is hðgÞ ¼ f ðtÞ=SðtÞ, and this can be modeled as a function of covariates using Cox’s [9] proportional hazards model.14 The hazard function is hðt; x; bÞ ¼ h0 ðtÞ expðb0 xÞ where h0 ðtÞ represents the unspecified baseline hazard function corresponding to the case where x is the zero vector. The coefficients b represent how the independent variables x multiplicatively affect the hazard. The parameters in this model are estimated by maximum likelihood and hypotheses are tested using standard Wald tests. The regressors in the hazard rate regression included dummy variables for restructuring status, non-utility projects, and the proportion of the project’s output that would be ‘‘merchant power’’, meaning that it would have to be sold in the spot power market and was not pre-sold under long-term contract.15 Table 6 displays the parameter estimates of a Cox hazard regression. Convergence of the model’s likelihood was achieved after six iterations and overall the model is statistically significant. Before discussing the individual regression coefficients it is useful to recall that they represent the multiplicative effect of each independent variable on the hazard rate, so coefficients
14 The advantage of using the proportional hazards model is that the baseline distribution of survival times need not be specified parametrically and the hazard can be modeled directly. As a robustness check, we also estimated parametric survival models that quantify life length—specifically the Weibull and lognormal—and found no difference in the qualitative results. 15 Projects often pre-sell a portion of their future output through longterm contracts, with the remaining generation capacity to be sold in the spot market. The data on the proportion of merchant power and whether or not the project is a utility or a non-utility project is part of the RDI power plant investment database. The data on restructuring status was obtained directly from the Energy Information Administration [12]; the remaining variables are we obtained from the NewGen database as discussed in the data description section above.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1410 Table 5 Life tables for new projects by RTO group Interval
rtogroup CA ISO 11 12 14 15 15 16 17 18 18 19 19 20 20 21 21 22 22 23 23 24 25 26 26 27 27 28 28 29 29 30
Beg. total
Cum. failure
Std. error
Hazard
Std. error
215 210 208 207 205 204 203 202 193 188 186 172 168 167 161
0.0233 0.0326 0.0372 0.0465 0.0512 0.0558 0.0605 0.1023 0.1256 0.1349 0.2000 0.2186 0.2233 0.2512 0.2791
0.0103 0.0121 0.0129 0.0144 0.0150 0.0157 0.0163 0.0207 0.0226 0.0233 0.0273 0.0282 0.0284 0.0296 0.0306
0.0235 0.0096 0.0048 0.0097 0.0049 0.0049 0.0049 0.0456 0.0262 0.0107 0.0782 0.0235 0.0060 0.0366 0.0380
0.0105 0.0068 0.0048 0.0069 0.0049 0.0049 0.0049 0.0152 0.0117 0.0076 0.0209 0.0118 0.0060 0.0149 0.0155
rtogroup ERCOT 8 9 129 22 23 128 23 24 126 25 26 121 28 29 118
0.0078 0.0233 0.0620 0.0853 0.0930
0.0077 0.0133 0.0212 0.0246 0.0256
0.0078 0.0157 0.0405 0.0251 0.0085
0.0078 0.0111 0.0181 0.0145 0.0085
rtogroup MISO 8 9 9 10 15 16 17 18 20 21 21 22 22 23 23 24 25 26 26 27 27 28 28 29 29 30
332 331 329 328 325 324 323 318 315 312 308 307 306
0.0030 0.0090 0.0120 0.0211 0.0241 0.0271 0.0422 0.0512 0.0602 0.0723 0.0753 0.0783 0.0813
0.0030 0.0052 0.0060 0.0079 0.0084 0.0089 0.0110 0.0121 0.0131 0.0142 0.0145 0.0147 0.0150
0.0030 0.0061 0.0030 0.0092 0.0031 0.0031 0.0156 0.0095 0.0096 0.0129 0.0033 0.0033 0.0033
0.0030 0.0043 0.0030 0.0053 0.0031 0.0031 0.0070 0.0055 0.0055 0.0065 0.0033 0.0033 0.0033
rtogroup PJM 2 3 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 25 26 26 27 27 28 28 29
339 338 337 336 332 331 330 329 328 327 323 321 318 312 310 306 305
0.0029 0.0059 0.0088 0.0206 0.0236 0.0265 0.0295 0.0324 0.0354 0.0472 0.0531 0.0619 0.0796 0.0855 0.0973 0.1003 0.1121
0.0029 0.0042 0.0051 0.0077 0.0082 0.0087 0.0092 0.0096 0.0100 0.0115 0.0122 0.0131 0.0147 0.0152 0.0161 0.0163 0.0171
0.0030 0.0030 0.0030 0.0120 0.0030 0.0030 0.0030 0.0030 0.0031 0.0123 0.0062 0.0094 0.0190 0.0064 0.0130 0.0033 0.0132
0.0030 0.0030 0.0030 0.0060 0.0030 0.0030 0.0030 0.0030 0.0031 0.0062 0.0044 0.0054 0.0078 0.0045 0.0065 0.0033 0.0066
Likelihood-ratio test statistic of homogeneity: w2 ð3Þ ¼ 39:555176, P ¼ 1:324e 08. Logrank test of homogeneity: w2 ð3Þ ¼ 50:81, Pr 4w2 ¼ 0:0000.
Fig. 2. Cumulative hazard rates for new projects by RTO grouping.
Table 6 Hazard regression Haz. ratio
Robust Std. err.
Haz. ratio
Robust Std. err.
Non-utility Restructuring Active Delayed Canceled Merchant (%) Combust turb Coal
1.811468
.3784981
1.959942
.4177448
1.083616 1.055826 3.371199 1.026478
.1919256 .3120933 .6099967 .0069154
1.058059 1.195648 2.957074 1.026404 1.988257 .5341334
.1883613 .3539154 .5283339 .0066883 .2936597 .2733636
Log likelihood Wald w2 ð5Þ Prob4w2
1464.5039 88.90 0.0000
1450.8163 103.71 0.0000
less than unity lower the hazard while those greater than unity increase the hazard. The results of the Cox regression show that projects in a jurisdiction that canceled restructuring—namely California—have a statistically higher hazard rate than the baseline hazard; projects in jurisdictions with active or delayed restructuring have hazards that are not statistically different from the baseline hazard of no restructuring. The coefficient on the percentage of merchant power was statistically larger than unity, indicating that projects with a lower percentage of locked-in sales had a significantly higher probability of failure. Non-utility projects also had a statistically higher hazard rate. Overall, the results of the hazard regression are consistent with the sequential abandonment of projects over time. In an auxiliary regression we included dummy variables for combustion turbine and coal power plant types and found that combustion turbine plants had a statistically larger hazard, holding other factors constant. Again, this is consistent with the real options nature of power plant development under regulatory restructuring.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
3.5. A closer look at California and Texas As discussed above, the regulatory environment in which permitting takes place as well as the stability of the regulations in a state can have a great influence on the pattern of investment in new generating facilities. To illustrate how different regulatory environments can lead to different investment patterns we focus on California and Texas—two states with very different regulatory environments. At the time of our review there were significant differences in regulation of new power plant development across these states. Overall, California’s regulatory process to approve the siting of a new power plant was more complex and often required more stringent environmental mitigation efforts. The overall process of receiving regulatory approvals is more complex in California than in Texas. The process in both states includes application and approval by state and local agencies to ensure compliance with environmental, land use, and other requirements. However, the process in California requires application and approval by the California Energy Commission, a state energy agency that conducts its own analysis of whether the benefits of additional electricity supplies outweigh any potential negative environmental or other effects. In addition, the regulatory process in California can also require new power plants to install more stringent air pollution mitigation equipment. Extensive areas in California have witnessed high levels of air pollution over many years and face more stringent federal oversight. As a result, federal air quality regulations generally require new power plants to undertake efforts to mitigate the pollution that they introduce into these already polluted areas. Exacerbating these federal requirements, the state of California has, in recent years, sought to require more stringent levels of air pollution mitigation than what has been required in Texas and many other states. Negotiating what types of air pollution mitigation efforts may be required can lengthen the time for some projects to be approved. Furthermore, in California the California Energy Commission can designate members of the public or other stakeholders as intervenors, allowing them to become formal participants in the regulatory process with the ability to request data, file motions and take other formal actions. Of the 72 siting cases filed from 1995 to 2001, 39 had one or more intervenors. The regulatory process in Texas also allows some members of the public to request contested evidentiary hearings, during which these members can present evidence. However, in only 15 of 84 of the siting cases filed since the law allowing these proceedings was enacted were these types of hearings held. The emergence of significant local opposition to a new plant is a significant factor in receiving the necessary approvals, delaying regulatory decisions in many cases. As a result, developers seek locations where their project will receive local support.16 16
For example, communities in Texas generally welcome new natural gas-fired power plants because they add to the communities’ tax base and pose few environmental concerns.
1411
It would be difficult to find two more disparate regulatory regimes than those of California and Texas with respect to siting new power plants. This difference is evidenced in the data: although the two states both have similar process for approving applications to build power plants, this process took 6 months longer on average in California (14 months) than in Texas (8 months) in the period from 1995 to 2001 [13]. Perhaps of more significance to developers is that California’s process was much more variable, with 5 of 21 medium to large-scale projects taking 18 months or longer. The uncertainty about the outcome of the permitting process also appears to be much greater in California than in Texas. In discussions with several large energy companies and several large investment banks, two of the authors repeatedly heard that California is a very difficult place to site a power plant because of the strength of opposition at the local level and the extent to which the permitting process in California can be held up by such opposition. In addition, Texas ‘‘socializes’’ the costs of interconnection, while California makes developers pay for all the interconnection costs associated with their new plant. If our conjectures are correct, then the data should show predictably different investment patterns in the two states. A look at the data indicate that this is the case. Table 7 shows the composition of power plant types in California and Texas. In California 58.6% of proposed projects are combustion turbines as compared to 17.05% in Texas. In contrast, 38.76% of projects in Texas are combined cycle plants as compared to 19.53% in California. This different pattern of investment is not inconsistent with our conjecture that greater uncertainty will cause investors to choose power plant types with lower capital to generating capacity ratios.17 Table 8 displays the composition of projects across development phases in California and Texas, and shows an 51 50 eventual observed success rate of 215 in California and 129 in Texas. This is a dramatic difference, but may not be reflective of a steady state. During the sample period, California underwent, along with the rest of the western states, an electricity crisis, largely caused by insufficient supply and a severe drought. During that crisis, prices were extremely high and the state responded by streamlining and expediting their permitting process for certain new generating plants. In response, many companies proposed a number of power plants in California and began the permitting process. Subsequent actions by the state to (1) suspend retail competition, (2) renegotiate long-term contracts signed during the crisis and (3) to demand
17 There are other likely factors in the explanation of this pattern, most notably, the greater prevalence of hydroelectric generation in California and nearby states. In high water years, there has typically been enough hydroelectricity generated to drive a number of base-load generators out of the market for significant amounts of time. This increases the number of years to expected payoff, which favors investments in combustion turbines (peaker units) to some extent.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413
1412
4. Conclusions
Table 7 Distribution of plant types in Texas and California Plant type
Texas
California
Freq CC/cogen CT/cogen Coal Comb cycle Combust turb Geothermal Hydro Intern combust Nuclear Other boiler Solar Waste Wind
9 10 1 50 22 0 0 0 2 3 0 12 20
Total
(%)
Freq
(%)
6.98 7.75 0.78 38.76 17.05 0 0 0 1.55 2.33 0 9.30 15.50
3 4 1 42 126 5 2 7 1 7 8 4 5
1.40 1.86 0.47 19.53 58.60 2.33 0.93 3.26 0.47 3.26 3.72 1.86 2.33
129
215
Table 8 Project phases in California and Texas Phase
Year/month 2000/01 2000/06 2001/01 2001/06 2002/01 2002/06
California Advan develop 2 Canceled 0 Early develop 19 Operating 0 Tabled 0 Under constr 3
8 0 13 1 1 3
4 5 26 2 3 7
16 10 55 10 7 23
25 29 47 44 32 16
26 60 22 51 40 16
Total
24
26
47
121
193
215
Texas Phase
Year/month 2000/01 2000/06 2001/01 2001/06 2002/01 2002/06
Advan develop 7 Canceled 0 Early develop 15 Operating 0 Tabled 1 Under constr 15
15 0 16 5 2 20
13 1 36 12 2 25
17 1 35 23 2 27
19 8 28 43 8 17
14 12 17 50 9 27
Total
58
89
105
123
129
38
refunds from generators who sold power in the state during the crisis caused many developers to flee.18 It is our understanding that almost all power plants proposed during this time were ultimately abandoned.
18 California’s suspension of retail competition was done in order to ensure that the state utilities could charge prices high enough to recover the costs of power purchased by the state at high prices during the height of the electricity crisis. These prices are considerably higher than current or expected future wholesale prices in the state.
Power plant investment is higher in states that have restructured electricity markets than in states that have taken no restructuring actions. Development is also more prevalent in areas of the country with a robust wholesale market infrastructure. Ownership of new power plants also differs across states, with non-utility companies accounting for the bulk of new power plants in states taking restructuring actions, while utilities still have a strong or dominant role in new development in states that have not restructured at all. States’ decisions to implement retail competition result in more investment in new power plants. These patterns indicate that state regulatory actions are an important determinant of how well restructuring at the national level will ultimately work. The bulk of the potential benefits of restructuring the industry will come from improvements in efficiency of wholesale generation and sale of electricity and this depends critically on the ability of new companies to enter and exit. In addition, reliability of the electricity system depends critically on the ability of companies to make timely investments in power plants and other infrastructure. However, non-utility companies are far less likely to make the investments necessary to achieve these benefits in uncertain regulatory environments. References [1] Brennan TJ. Provincial and federal roles in facilitating electricity competition: legal and economic perspectives. In: Walls WD, editor. Regional transmission organizations: restructuring electricity transmission in Canada. Calgary: The Van Horne Institute; 2003. p. 20–40 (Chapter 2). [2] Feldberg P, Jenkins M. Reciprocity, regional transmission organizations, and standard market design: some implications for Canadian participation in North American wholesale electricity trade. In: Walls WD, editor. Regional transmission organizations: restructuring electricity transmission in Canada. Calgary: The Van Horne Institute; 2003. p. 60–75 (Chapter 3). [3] U.S. General Accounting Office. Lessons learned from restructuring: transition to competitive markets underway, but full benefits will take time and effort to achieve. Technical Report GAO-03-271, GAO, Washington, DC; December 2002. [4] Joskow PL. The difficult transition to competitive electricity markets in the U.S. Working paper, Department of Applied Economics, University of Cambridge, UK; 2003. [5] Brennan TJ, Palmer KL, Salvador AM. Alternating currents: electricity markets and public policy resources for the future. Washington DC; 2002. [6] Borenstein S. The trouble with electricity markets: understanding California’s restructuring disaster. J Econ Perspect 2002;16(1): 191–211. [7] Joskow P, Kahn E. A quantitative analysis of pricing behaviour in California’s wholesale electricity market during summer 2000. Energy J 2002;23(4):1–35. [8] Ludwigson J, Rusco FW, Walls WD. Regulatory uncertainty and the development of new electric generation: competing risks with option value. In: Integrating the energy markets in North America: issues and problems, terms and conditions. Proceedings of the 23rd IAEE North American Conference on CD-ROM. International Association for Energy Economics; 2003. p. 1–10.
ARTICLE IN PRESS W.D. Walls et al. / Energy 32 (2007) 1403–1413 [9] Cox DR. Regression models and life tables. J R Stat Soc 1972;B34:187–220. [10] Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: Wiley; 1980. [11] Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457–81.
1413
[12] U.S. Energy Information Administration. Status of state electric industry restructuring activity. hhttp://www.eia.doe.gov/cneaf/electricity/ chg-str/restructure.pdfi; February 2003. [13] U.S. General Accounting Office. Restructured electricity markets: three states’ experiences in adding generating capacity. Technical Report GAO-02-427, GAO, Washington, DC, May 2002.