1 Electricity System Reliability as a Club Good Malcolm Ainspan Israel Public Utilities Authority – Electricity malcolm@pua.gov.il Advanced Workshop in Regulation and Competition Annual Eastern Conference May 17, 2007 Abstract: The objective of this paper is to build upon previous research on classifying electricity services as public goods, private goods, and club goods. Electricity reliability services are generally assumed to be public goods, even though the infrastructure providing these services is comprised largely of congestible facilities (e.g., transmission networks). This paper will address only the reliability services that are club goods, and will focus on club good issues common to most reliability organizations, such as heterogeneous membership, multigenerational network facilities, usage uncertainty, and “exit” potential. The paper will develop optimality conditions for club membership and provision, and will propose second-best alternatives that address the problem of first-best optima that are not implementable “politically”, but are nevertheless improvements over the current ratemaking practices of reliability organizations in the US and Europe. I. Introduction Paying for reliability-related elements of electricity supply has become one of the most controversial and politically-charged issues associated with electricity market reform. While electric energy may fit the standard microeconomic model for private goods that may be competitively supplied, reliability services such as capacity and transmission services (including ancillary services) do not. In public finance terms, these services are not simultaneously excludable and non-rival. That is, an individual market participant can exclude others from enjoying reliability services only at a significant cost, but that participant’s consumption does not diminish others’ consumption of those services. 1 The question of "who pays" for incentives to provide efficient levels of non-private services is generally answered as "load". That is, load is obligated to pay for the non- 1 Cornes, R. and Sandler, T. The Theory of Externalities, Public Goods, and Club Goods, 2nd Edition, (Cambridge University Press 1996). I follow the definitions in this book throughout this paper. 2 private aspects of reliability services, and these payments are often analogous to a combination of per-capita and consumption taxes used to finance public goods2. The provider of these services and the party responsible for determining the socially optimal level of those services are assumed to be the generators and the system operator, respectively. Therefore, in club-good terms, the system operator is the welfare-maximizing social planner appointed ex ante by the club, and selects prices and quantities based on competitive bidding where possible. It is interesting, however, that markets outside the US often consider the club to consist of generators as well, and that load can compete with generation to offer reliability services to the system operator – a comparatively recent development in US markets. Viewing electricity reliability as pure public goods whose benefits are non-rival and non-excludable is almost axiomatic3. Consequently, pricing these public goods generally takes the form of a consumption tax, based on the individual's estimated consumption during the period(s) for which the supplied quantity of the service is highest. For example, the costs of reserves provided by regional transmission organizations ("RTOs") are often allocated to "load" (usually load aggregators, including electricity distribution companies) according to their contributions to some measure of the RTO's peak load, thereby linking payment to "cost causation".4 This paper will argue that reliability services are best described not as "public goods", but as "club goods", whose benefits are excludable but partially non-rival. Control area operators such as regional transmission organizations (RTOs) can physically exclude consumers outside the control area from benefiting from its services, and can reduce reliability levels for some consumers within its control area.5 Such benefits include generation and transmission capacity and several ancillary services. Transmission capacity is a particularly interesting example, since it provides benefits 2 The Open Access Transmission Tariffs of PJM, ISO New England, and the New York ISO and their respective FERC filings contain several examples of this pricing mechanism and the underlying cost allocation methodology. 3 See, for example, Joskow, P. and Tirole, J. (2004). "Reliability and Competitive Electricity Markets", MIT Working Paper. Also see Lesser, J. and Israilevich, G. (2005) "The Capacity Market Enigma: Why Haven't Reliability Markets Developed?” Public Utilities Fortnightly, December 2005. 4 The FERC allows the RTO to choose whether to use an annual peak or an average of monthly peaks, depending on the RTO's load seasonality. In other countries, including Israel, allocation is on the basis of energy consumption for both transmission pricing and retail ratemaking. 5 Assuming that the RTO is "selected" and authorized by the "club" of market participants to maximize their aggregate net benefits. Throughout this paper, the club will have already made this selection. 3 that are partially non-rival where one consumer may consume and detract from others' consumption when the transmission system is congested.6 In fact, since club goods cover the whole spectrum between the pure-private and pure-public good paradigms, it is possible to classify many seemingly "ungroupable" reliability services as club goods. For example, some reliability elements of electricity are much closer to the public good paradigm (e.g., system dispatch), while others may be nearly private goods (e.g., consumer-level reliability enhancement from un-interruptible power supply "UPS" systems). This paper will develop pricing concepts applicable to the entire club-good spectrum, and suggest where each reliability service may be located on that spectrum. The possible application of club good models to electricity reliability services is not completely new, having been addressed in policy discussions regarding merchant transmission.7 However, these discussions have focused on conditions under which a merchant transmission "club" contributes to economic and engineering efficiency; i.e., whether or not the club should be formed.8 This paper will address whether reliability services within the existing electricity markets have been "mis-classified" as pure public goods and therefore "mis-priced". II. Club good basics According to Cornes and Sandler (1996), a club is a "voluntary group deriving mutual benefit from sharing one or more of the following: production costs, the members' characteristics, or a good characterized by excludable benefits"9. This mutuality of benefits is responsible for club goods' being inefficiently supplied through competitive markets. In public finance terminology, club goods are market failures, similar to public goods. Like public goods, club goods can be optimally supplied through intra-club bargaining if all club members have complete information and transaction costs are negligible. When these conditions do not exist, however, 6 Transmission congestion occurs when transmission capacity is insufficient to deliver otherwise economic energy into a given location. This concept has been extended to reserves and capacity as well, both of which have club-good attributes as well. 7 Smith, V. in Chao, H-P and Huntington, H.ed. Designing Competitive ElectricityMarkets, (Boston, MA: Kluwer Academic Publishers, 1998), Chapter 7. 8 Joskow, P. and Tirole, J. "Merchant Transmission Investment", MIT Working Paper, February 10, 2003. 9 Cornes, R. and Sandler, T. The Theory of Externalities, Public Goods, and Club Goods, 2nd Edition, (Cambridge University Press 1996). 4 outside intervention is necessary, usually by the regulator ultimately responsible for tariff approval.10 Club goods differ from public goods in at least 4 ways, all of which affect their respective pricing mechanisms: 1. Club goods involve a voluntary membership choice, based on an anticipated net benefit from membership. For public goods, the costs associated with refusing membership are prohibitive. 2. Club goods are congestible, so congestion produces finite memberships that are subsets of a population. Public goods' congestion costs are zero. 3. Club goods are excludable, so that users' utilization rates can be monitored and nonmembers/nonpayers can be barred. The cost of the exclusion mechanism must be less than the benefits of allocating the shared good within the club. Public goods do not allow for such an exclusion mechanism. 4. Club goods can achieve optimal size and usage levels through tolls, membership fees, and usage fees. Public goods do not yield optimal results without some form of government intervention, including taxation. Below is a basic formal model of club goods under the assumptions of homogeneous membership and fixed utilization rates. • A representative member i maximizes its utility Ui = Ui (yi,X,s) over its consumption of the private good yi, the club good X, and the club membership size s, subject to a resource constraint: Fi (yi,X,s) = 0. • Both U and F are strictly increasing in yi and X and strictly decreasing in s, and are continuous and twice-differentiable in yi and X.11 The first-order conditions are: MRSi Xy = MRTi Xy (the "provision" condition) MRSi sy = MRTi sy (the "membership" condition) 10 An outcome well understood in New England in recent years with regard to designing capacity markets. 11 Note, however, that s is an integer; there is no such thing as a partial member. Therefore, the problem must be formulated as a mixed integer programming model, which does not readily provide globally optimal solutions. 5 Since all members are identical, and club costs are equally shared among members, an additional member will reduce per-member expenditures for all members. The provision condition indicates that the marginal rate of substitution between the club good and the private good must be equated to the marginal rate of transformation between these 2 goods (i.e., marginal benefit = marginal cost). The membership condition equates the (negative) marginal benefits of crowding to the marginal costs (also negative) of cost sharing. 3. Model extension and application to electricity markets This basic club model, while a useful starting point, does not entirely reflect the reality of most clubs, including a "reliability club" for electricity, for the following reasons. (1) Most clubs are quite heterogeneous, with members using, and paying for various amounts of each service. Electricity reliability is a derived demand linked to the demands of club members. (2) Most clubs are characterized by significant information asymmetries. There is asymmetric information in the reliability club – no club member knows who else is in the club at a given time - requiring monitoring resources (e.g., within the RTO/market administrator) to assess whether there is anyone free-riding off the club's resources and to exclude them12. (3) The optimal number of clubs for a heterogeneous population remains a research work-in-progress. An optimal number of clubs will exist when the provision level maximizing a consumer's surplus is in the range of decreasing average cost, thereby assuring that the addition of a member to a club will benefit everyone. However, this sufficient condition is not a necessary for an optimum – which is why the issue of the optimal number of clubs remains 12 The issue of whether the agency costs paid to the RTO produce the optimal monitoring level is important, but not addressed in this paper. 6 unsettled.13 The basic model assumes only one club, with no option to re- assess whether that number is optimal. (4) Most clubs are dynamic; club members' decisions today affect the utility of potential club members in the future. Reliability clubs are a prime example of such a dynamic club. In contrast, the basic model is completely static. (5) Most clubs must include uncertainty and club members' risk attitudes in determining capacity and provision levels. Club members may not be able to utilize some services due to unexpected capacity constraints or facilities breakdowns. The basic model has no uncertainty and risk attitudes are irrelevant. (6) Many clubs, including reliability clubs offer more than one product or service. The reliability club is truly a multi-product club, offering a variety of services. These services may be provided more cheaply – and be more desirable - as a package than individually. The basic model assumes one good provided by the club. (7) Electricity flows and transfer capabilities follow the laws of physics, and may enter other clubs rather than remaining in the club. The basic club model assumes that club service packages may change only over time in response to changes in club membership composition, not in response to immediate changes in system conditions.14 To my knowledge, there is no uniform method of calculating compensation among clubs affected by these inter-club flows.15 13 The fact that there may be other necessary conditions in order to derive an optimal club size may help to explain why the issue of optimal RTO size has been difficult to solve. Declining average costs for dispatch and capacity due to increased club size has not created massive RTOs except possibly for PJM. See Andrew Hartshorn and Scott Harvey, “Assessing the Short-Run Benefits from a Combined Northeast Market,” October 2001 LECG working paper. 14 See Hogan, William J. (1991). “Transmission Capacity Rights for the Congestion Highway: A Contract Network Proposal”, submitted to FERC in Docket No. PL-91-1-000. 15 See Hogan, William J. (1992). “Contract Networks for Electric Power Transmission “, Journal of Regulatory Economics, 4:211-242. Also Daxhelet, O. and Smeers, Y. “Inter-TSO Compensation Mechanism”, November 7, 2005 working paper. 7 (8) In electricity markets, club good producers are not necessarily club members, although the supplier to club members must be in the club. The basic model assumes that the supplier “automatically” has the supply available and merely needs to choose the optimal provision level. Therefore, the basic model does not provide much insight when it simply cannot obtain that provision level, due to strategic behavior by producers. Generation capacity markets are examples of such club goods, and the optimal mechanism to compensate capacity producers (generation and demand resources; intra-club and extra-club) is a work-in-progress. At this point, we note that, for reliability clubs, many of the decisions that would ordinarily be made by club members are made by the RTO/market administrator ("RTO/MA") on behalf of club members, due to asymmetric information, and significant transactions costs. For example, overall provision levels for reliability services such as reserves requirements and transmission capacity are determined by the RTO/MA, rather than negotiated by market participants. Nevertheless, it is possible to imagine the RTO/MA as a "benevolent dictator" acting on the club's behalf. We will now discuss how the RTO/MA and/or the electricity regulator address these deviations from the basic club model, and whether there is room for improvement. (1) Heterogeneity: The RTO/MA currently addresses heterogeneity by assessing total charges on the basis of each consumer's contribution to a measure of system peak load. For example, if the system's annual peak demand is 10,000 MW and the consumer's demand is 100 MW, then the consumer will pay for 1% (100/10,000) of the annual costs of providing that service. However, in reality, consumers may have differing values for the quantity of the service being provided, especially if the service is congestible. Therefore, efficient pricing for a club with heterogeneous members requires all consumers to pay the same usage toll, but a high-usage consumer with a high-priority need for that service must incur higher toll payments overall. While this mechanism appears very similar to willingness-to-pay mechanisms for any public good - including priority service pricing of reliability 8 services that are truly public - the congestibility of some reliability services makes such a mechanism more crucial for these services than for pure public goods. (2) Asymmetric information: The RTO/MA addresses asymmetric information among club members by simply being the enforcement authority accepted by the club members through their voting procedures. The need for an RTO/MA to monitor activity by members and non-member infiltrators depends largely on the cost of the RTO/MA relative to the cost of infiltration without RTO/MA enforcement. The need for the RTO/MA remedies to the asymmetric information problem is high if: (a) the expected utility of infiltrating/free-riding exceeds the utility of an honest infiltrator; and (b) the expected fine from infiltrating/free riding is less than the user fee. Since the RTO/MA can affect both (a) and (b), however, it is possible for it to "make itself less relevant" with the proper choice of incentive-compatible parameters. (3) Optimal number of clubs: The major additional consideration when making the optimal number of clubs endogenous is that the optimal number of clubs maximizes the total utility of all members and non-members of each club, rather than the average utility of members already within a club. For reliability clubs, the optimal number has historically been set exogenously along electrical control area or political/jurisdictional lines. In the US, the FERC is ultimately responsible for determining the number of clubs to be allowed. In recent years, the FERC has become increasingly involved in determining the optimal number of RTO clubs, most notably in merger discussions between ISO New England and the New York ISO, the integration of ECAR into the PJM system, and the configuration of the Midwest ISO.16 The analytical work supporting these club configurations has closely followed the theory producing the optimal number of clubs endogenously. However, this work has not entirely addressed the provision, membership and usage conditions associated with each configuration, resulting in tariffs that likely provide inefficient signals regarding provision, membership and usage for some reliability services. 16 This issue is being debated at the distribution level in Israel in the context of future licensing and privatization. 9 (4) Dynamic issues. Introducing dynamics into reliability clubs adds several levels of complexity that are not usually part of any ratemaking. Issues such as intertemporal welfare maximization, time-varying crowding and depreciation/maintenance costs, and time-varying club composition must be reflected in determining optimal membership, usage, and provision conditions in each period. Of these issues, depreciation/maintenance costs may be the most difficult to address, since they are truly public goods within the club; they are not identifiable with an individual club member or its usage. For dynamic clubs, the following characterize the optimality conditions: • Optimal provision requires that the sum of marginal benefits from crowding relief and depreciation reduction must equal the marginal costs of increased provision, accounting for all club members and their membership durations. • Optimal usage requires the usage fee to be set such that the marginal benefits of usage equal the marginal costs of crowding and depreciation on current members (even though marginal depreciation costs themselves affect both current and future members). • Optimal membership requires the net membership benefits for an entrant to be greater than or equal the associated costs of using the club services during his membership duration. Given the complexity of quantifying the optimal provision, usage, and membership conditions over time, it should not be surprising that reliability club pricing to "load" is largely cost-based and single-period, rather than net benefit- based and intertemporal. Moreover, given the "political-cycle" affects encouraging some club members (e.g., utilities subject to significant political pressures) to focus exclusively on their own short-run net benefits, it is unlikely that such efficient pricing mechanisms would be implemented even if this complexity were resolved. 10 (5) Multi-product clubs. The electricity reliability club provides a variety of services under the reliability rubric, mainly due to economies of scope and the physics of electricity. However, it is possible for "cross-crowding" among the products offered, where one member's utilization of one service detracts from the quality of services utilized by other members. In this case, the efficient usage level requires that the marginal benefit from usage equal the marginal crowding and cross-crowding costs imposed by that usage, and the efficient membership level requires that the cost savings from having another member must equal the crowding and cross-crowding costs imposed on the membership. Moreover, since there is a tradeoff between the benefits of offering more services and the associated cross-crowding costs of these services, there must also be an optimal provision level, in addition to the optimal usage and membership levels – although this optimal provision level is usually delegated to the reliability club's system operator as the "benevolent dictator". Currently, such pricing for a "reliability club" exists only in a crude fashion, if at all. For example, if a system operator determines that a generator is required to provide locational reserves, it may not be usable for reactive power, thereby detracting from others' utility from reliability. Club members ("load") simply pays for these services without regard to these "cross-crowding" effects – again using standard tariffed per- peak MW rates for any amount of each service. (6) Capacity constraint/utilization uncertainty. Some clubs, including electricity reliability clubs, have capacity limits for member access to the club good when usage levels reach that limit. Examples of these clubs include highways and telecommunications network facilities. Optimal pricing for such clubs requires a two- part tariff that consisting of (a) a lump-sum membership charge that is independent of a member's usage and the prevailing state of congestion; and (b) a variable user charge collected only if the member is admitted and whose level depends on the number of current users (which is random) and the facility size. Current tariffs are far from this ideal. These tariffs consist almost exclusively of user charges. If a transmission facility is congested and usage must be denied, the member simply pays standard transmission user charges; congestion charges are part the 11 energy charges or simply allocated across all members -- neither of which necessarily reflects the fact that transmission system capacity is configured to meet peak capacity, not energy. That is, there is no (significant) fixed access fee for membership, and transmission usage charges have little connection to real-time conditions on the transmission bottleneck (being based instead on test-year conditions and possibly some ex post side-payment compensation). (7) Efficiency of creating “sub-clubs”. In some cases, where location largely dictates club membership, there may be efficiencies in creating internal clubs to reflect their preferences. In some cases, the coordination costs associated with obtaining even a binding majority vote with these “outlier” members exceeds the cost of creating internal clubs for these members and allowing the “non-outliers” to decide on overall club provision levels. To date, there has been some work in this area with regard to the European Union “club”, where the question of whether to create an inclusive organization (membership cost subadditivity) without sub-clubs or to develop mechanisms to improve club-good provision is a major one.17 Nevertheless, unlike the EU policymakers, some researchers indicate that the correct “sub-club” configuration and provision should be left to the club members to decide among themselves, not through a central planner. In our context, it is unclear whether the RTOs’ focus on a centralized club (where even pricing for club members differentiated by location is determined centrally) is the best approach. It may be best for the RTO to simply define the club-wide reliability requirements and the decision mechanism, while having club members decide compensation and sub-club configuration by themselves.18 An example of a reliability club: Germany Until 1998, when Germany began the process of regulatory reform leading to conformity to European Union standards, it provided an example of a restructured electricity market relying heavily on club arrangements among club members, rather than "re-regulation". Collective agreements among electric companies, the German Business Association and the association of large consumers jointly defined the terms 17 Berglof, E., Burkart, M., Friebel, G. and Paltseva, E. (2006).”Club-in-the-Club: Reform under Uncertainty”, working paper. 18 Scotchmer, S. (2002). “Local Public Goods and Clubs”, Handbook of Public Economics, A. Auerbach, and M. Feldstein, eds. Vol. 4, Ch. 22, pp. 1997-2042. 12 of access to the electricity grid, known as Association Agreements. While apparently functioning well for a few years, asymmetric bargaining power among club members led to excessive access (i.e., membership) charges, and calls for government regulation to replace the club arrangement. Nevertheless, the fact that asymmetric bargaining power led to the club's demise points to the inefficiency of (some combination of) the club's membership, provision, and usage levels, and to mis- application of the club model to truly public or private services, rather than to the club model itself. Possibly recognizing this fact, the initial electricity reform legislation provided for much more light-handed regulation than exists in most other restructured markets.19 Summary of current reliability tariff design and proposed improvements As discussed above, RTOs address pricing of non-private electricity services through various cost allocation mechanisms and charges that resemble taxes for public good provision. For example, even though capacity and some location-dependent ancillary services such as reserves may be considered club services consumed by members of a location-specific reliability club, it has been only recently that this has been translated into tariffs paid by the club members – i.e., the load serving entities ("LSEs") in that location. However, having defined the location-specific reliability club, the RTO often simply allocates the associated costs on some load-ratio-share basis without regard to the inter-LSE differences in reliability preferences and risk attitudes, or the possibility of cross-crowding resulting from the RTO's providing many related services. Potential club approach: Competitive joint venture as club: Recent work by Boffa and Kiesling (2006) proposes a governance mechanism for electricity distribution companies called a competitive joint venture, which holds promise for efficient pricing of club goods. Under this mechanism, club members purchase entitlements to club goods and sell these to other club members who are permitted to own entitlements. Although the CJV translates most directly to network infrastructure itself (e.g., joint ownership of transmission or distribution facilities), it 19 Glachant, J-M, Dubois, U. and Perez, Y. (2003) "Deregulating with no regulator: Is Germany electricity transmission regime institutionally correct?” University of Paris research paper. 13 is usable for pricing rights to other reliability services as well, and may be superior to the current load-share allocation methods for pricing some reliability services for which club members’ demand types vary across the network (e.g., local reactive power). To date, the RTOs as market administrators have not focused on improving these “markets” in non-private goods, instead (correctly!) concentrating on improving the private-good markets in energy and reserves supply. Nevertheless, as reliability services become relatively more important, the CJV may offer a promising starting point for efficient pricing.20 Conclusion Electricity supply is a combination of public, private, and club services, and requires that taxonomy for efficient provision. As Kiesling points out, however, more of these services are being classified as private or club services, due to the increasing importance of location and congestion (and therefore exclusion) in electricity transactions and the ability to use technology to deliver services to decentralized clubs and individuals with heterogeneous preferences. The public good "tax" model, with its "one-size-fits-all" approach to pricing reliability services, has never been fully efficient and is becoming less so over time. Whether increased sophistication in pricing reliability club services such as location-specific capacity will reduce the political opposition to such initiatives remains an open question. Nevertheless, in developing policies for reliability services, the reliability-as-public-good paradigm should be accepted only after the private and club good models have been carefully examined and rejected. 20 Boffa, F. and Kiesling, L. (2006). “Network Regulation through Ownership Structure: An Application to the Electric Power Industry,” working paper. 14 REFERENCES 1. Berglof, E., Burkart, M., Friebel, G. and Paltseva, E. (2006).”Club-in-the-Club: Reform under Uncertainty”, working paper. 2. Boffa, F. and Kiesling, L. (2006). “Network Regulation through Ownership Structure: An Application to the Electric Power Industry,” working paper. 3. Cornes, R. and Sandler, T. The Theory of Externalities, Public Goods, and Club Goods, 2nd Edition, (Cambridge University Press 1996). 4. Glachant, J-M, Dubois, U. and Perez, Y. (2003) "Deregulating with no regulator: Is Germany electricity transmission regime institutionally correct?" University of Paris research paper. 5. Hartshorn, A. and Harvey, S., “Assessing the Short-Run Benefits from a Combined Northeast Market,” October 2001 LECG working paper. 6. Hogan, William J. (1991). “Transmission Capacity Rights for the Congestion Highway: A Contract Network Proposal”, submitted to FERC in Docket No. PL-91-1-000. 7. Hogan, William J. (1992). “Contract Networks for Electric Power Transmission “, Journal of Regulatory Economics, 4:211-242. Also Daxhelet, O. and Smeers, Y. “Inter-TSO Compensation Mechanism”, November 7, 2005 working paper. 8. Joskow, P. and Tirole, J. (2004). "Reliability and Competitive Electricity Markets", MIT Working Paper. 9. Joskow, P. and Tirole, J. "Merchant Transmission Investment", MIT Working Paper, February 10, 2003. 10. Kiesling, L. and Giberson, M. (2005). "Electric Network Reliability as a Public Good", paper submitted to Carnegie-Mellon University conference Electricity Transmission in Deregulated Markets. 11. Lesser, J. and Israilevich, G. (2005) "The Capacity Market Enigma: Why Haven't Reliability Markets Developed?" Public Utilities Fortnightly, December 2005. 12. Oren, S. (2003), "Ensuring Generation Adequacy in Competitive Electricity Markets", University of California Energy Institute, Working Paper EPE-007. 13. Scotchmer, S. (2002). “Local Public Goods and Clubs”, Handbook of Public Economics, A. Auerbach, and M. Feldstein, eds. Vol. 4, Ch. 22, pp. 1997-2042. 14. Smith, V. in Chao, H-P and Huntington, H. ed. Designing Competitive Electricity Markets, (Boston, MA: Kluwer Academic Publishers, 1998), Chapter 7.