Published on Fortnightly (http://www.fortnightly.com)

Home > Printer-friendly > East vs. West: Comparing Electric Markets in California and PJM


 

While neither is perfect, the former has done better than its critics would admit.

In France, analysts admit that markets work in practice, but question if they work in theory.1 Now this Gallic flavor has taken hold in the New World, but with a different twist. Today, as electricity restructuring proceeds in North America, the regulators and analysts there seem enraptured with theories of ideal market structures, but they largely ignore the practical results.

In the real world, commodities with immediate spoilage, such as transportation, sports, and theater—and yes, electricity—require a transparent and robust forward market. Yet in power markets, this reality often has been sidestepped—lost in the debate of the structure of independent system operators (ISOs) and regional transmission organizations (RTOs). The idea is not such a mystery in other industries. It is obvious to any business traveler that a reserved seat on a scheduled airline flight is preferable to relying on the uncertainties of a spot market. Travelers (and airlines) need to plan ahead if supply and demand functions are to be matched.

In fact, when it has been allowed to do so, the wholesale power market has acknowledged the importance of forward markets by dividing itself into three segments.

The first segment is the real-time market for balancing the grid and managing congestion. The second is the prescheduled market, which occurs in advance of day-ahead scheduling. The third segment includes forward and futures markets where deals may be struck months or even years in advance of delivery. Prices from the prescheduled market allow participants to plan daily generation and load. Forward and futures markets allow participants to perform seasonal planning and make rational investment decisions. While these market segments remain separate (having distinctly different prices), they are interrelated. Under a well-designed system, all three will function efficiently.

In this paper, I compare the power prices emerging from two restructured markets, each having a different approach. One market is PJM, the Pennsylvania-New Jersey-Maryland Interconnection. The other is California. The results show similarities on the surface, but reveal distinct differences when one digs deeper. These differences may well turn on the philosophical debate illustrated by these two markets. Which is more valuable in designing a power market with efficient prices: a focus on grid congestion, or on price transparency and certainty?

The first market, PJM, operated as a "tight" power pool for many years but reconstituted itself as an independent system operator in January 1998. Later, in April 1998, it shifted its congestion management to a system of Locational Marginal Pricing (LMP). This market structure has been distinctly short term, with pricing in the forward market left to bilateral negotiations.2 The restructured California market, on the other hand, began operations in April 1998, with two new institutions: an ISO that combined the transmission systems of the state's three largest utilities; and a power exchange operated and managed separately from the ISO. In California, the ISO operates a real-time market and the Power Exchange operates a series of forward markets. In addition, in each of the two markets—California and PJM—buyers and sellers have access to futures contracts with nearby points of delivery. Thus, the New York Mercantile Exchange (NYMEX) has established futures markets based in the Western Hub of PJM, and on the Oregon and Arizona borders with California. Further, the "over-the-counter" market provides options and swaps, but not with the transparency of the other markets.

While the PJM and California markets have very different histories and are geographically separate, they have much in common...both seek to integrate the transmission systems of a number of utilities and create a broad competitive market with diverse buyers and sellers. Moreover, in recent years, natural gas prices have converged across the nation. As a consequence, generation costs for many marginal units are similar in both California and the Mid-Atlantic (the PJM region). That makes a comparison of results (i.e., the levels and variability of prices) in the two markets a useful, if not conclusive, exercise.3 Following is a summary of the performance in the largest trading zone of California and PJM:

These findings should give pause to the many analysts and regulators advocating that other regions adopt the PJM pricing structure without critical review. While neither system adequately has developed forward markets, California was first to integrate such trading into its system and appears to have performed much better than many of its critics have suggested. In short, there is much to be learned from both approaches.

Market Structure: Differences and Similarities

PJM has over 2,000 buses or nodes, and the price of power at each one could vary under the LMP system. In practice, however, the system is much simpler. For example, there are no price differences between nodes unless the grid is congested. PJM has created a large "hub" in order to ensure an adequate number of participants and competitive trading, even in the presence of congestion. Prices for purpose of settlement in PJM's Western hub are calculated as the average of 111 buses, a grouping chosen to be the least likely to experience congestion difficulties.

California chose a zonal, rather than a nodal approach. Twenty-four separate zones were created, but again, that overstates the complexity of actual operations. In practice, there are two dead zones (Humbolt and San Francisco) and two active zones (Southern and Northern California).4 The remaining 20 zones are points of intertie with the remainder of the Western System Coordinating Council (WSCC) grid. The two large active zones are referred to as SP15 (south of the state's principal transmission line, Path 15) and NP15 (north of Path 15). Important interconnections between California and the other states are NW3 and NW1 (the California-Oregon border and Nevada-Oregon border) and AZ3 (Palo Verde, which connects to Arizona).

Although there is trade between PJM and surrounding pools, PJM is generally a self-contained system. In contrast, California is clearly a net importer of power. This circumstance reflects the seasonal surplus of hydroelectric energy available from the Pacific Northwest and ownership by California utilities of nuclear and coal generation facilities outside the state. Both PJM and California are largely thermal systems and depend on a variety of generation resources. Findings presented later suggest that gas-fired generators as a class represent the best indicator of marginal generating costs in both systems, although the correlations between electricity and gas prices are not high.

The PJM market combines bilateral trading with a spot market operated by the pool. In April 1998, PJM implemented a market-based bidding system for determining spot prices. The system is very short-term, analogous to the real-time market in California. PJM chose to implement its real-time spot market before adding a forward market because of the complexities of implementing both at the same time. Initially generators within the pool faced bidding constraints out of concern for the exercise of market power; these restrictions were removed in April 1999.5 In the event of congestion, an engineering model, combined with the bids at each node, is used to determine LMPs. Electricity scheduled by bilateral traders is charged a flat transmission fee when there is no congestion. In the presence of congestion, differences between LMPs calculated by the pool are used to determine the transmission charges.

PJM does not separate market operations from grid operations. Spot-market bids for each node, submitted a day in advance, include offer price curves for generation. If there are no transmission constraints, the LMP is equal to the bid price of the highest increment of generation requested to operate. In the presence of congestion, the system operator will dispatch "one or more of the generating units out of economic merit order to keep transmission flows within limits."6 LMPs reflect the cost of delivering energy to the congested node. The resulting prices for each node are posted on the PJM website in real time in five-minute increments. Until PJM implemented its forward market (initiated in June 2000), buyers did not know in advance the transmission costs or what price they would pay for power. In similar fashion, suppliers did not know what portion of their offer price curve would be dispatched. While it is too early to analyze the effectiveness of PJM's new day-ahead market, early reports from traders suggest that it has helped to stabilize prices and make the market more predictable.

By contrast, California chose to separate grid management from market activities, and established the California Independent System Operator (CAISO) and the California Power Exchange (CalPX). The CAISO balances the grid, determines "real-time" prices, and ensures reliability through its ancillary-services market. The CalPX operates a series of forward markets: the day-of, day-ahead hourly, and a block forward market. California's three largest investor-owned electric utilities initially were required to purchase and schedule power only through the CalPX. (The state PUC lifted that requirement on June 8, when it ruled that the state's wholesale power market had matured sufficiently to allow the three utilities to buy power on other "qualified" exchanges.7) Bilateral trading by "scheduling coordinators" is allowed, but the CalPX handles nearly 90 percent of market volume. Congestion differentials for managing the grid are determined by the ISO, based on adjustment bids that accompany day-ahead schedules.

The centerpiece of the California market structure is the CalPX's day-ahead hourly market. Pricing in the day-ahead hourly market is a two-step process. First, participants submit "portfolio" demand and supply price offer curves.8 The CalPX aggregates these bids, and an unconstrained market clearing price (umcp) is determined for each hour. Once a participant's bid is accepted, initial proposed schedules for each zone are submitted, along with adjustment bids in the event of congestion. The CalPX's schedules, along with schedules of other scheduling coordinators, are submitted to the CAISO, which determines their feasibility and makes adjustments, if necessary. If there is no congestion, the UMCP is the price for all zones. In the presence of congestion, zonal prices are adjusted down or up to account for the ISO's differentials.

The CalPX day-ahead hourly auction is held once each morning, which concentrates liquidity, rather than spreading out bidding through a longer period.9 The first step of the auction also concentrates liquidity because portfolio bids do not differentiate between zones. With the participants and volume thus concentrated, this system is intended to reduce the risk that localized market power at specific nodes will put upward pressure on the aggregate price level. The CalPX also offers a block forward market, for month-long and quarterly advanced purchase and sale of on-peak blocks of power.

NYMEX offers month-long on-peak power futures contracts for both the PJM and California regions. So far, however, these contracts have not proven successful. As Figure 1 illustrates, the volume of trading in the electricity contracts has remained significantly below comparable levels in the natural gas contract since the power contracts were launched.

Many factors explain why trading in electricity futures is so thin. Consider two of the more important reasons. First, NYMEX entered the market very early, just as deregulation got underway; without well-developed trading in "physicals" it is unrealistic to expect a high volume in "futures."10 Second, so far there has been high basis risk due to the volatility in transmission charges related to congestion; this factor limits the market for a futures product. All the same, it is a good bet that the run-up in prices in the Western market in early summer will change perspectives on the NYMEX contract (and the CalPX block forward market). A Southern California buyer could have locked in a price between $33.15 and $36.80 per megawatt-hour for May 2000, by purchasing the NYMEX futures contracts in March for delivery in May.

Survey Findings: Price Differentials and Correlations

Figures 2 and 3 illustrate pricing in PJM and the CalPX. Each graphic compares four separate series: (1) exchange or pool prices in the largest zone, (2) bilateral prices (determined by survey) for the same region, and estimated marginal costs of operating a (3) gas turbine and a (4) diesel oil generator in that region of the grid.11 Prices represent the average of a 16-hour on-peak block for each day of operation.

A closer look at the data (see Table 1) reveals the differences in price trends between the two systems.

Average Prices. Over the 26-month period covering this comparison, the results are almost identical...average prices for SP15 resulting from the day-ahead CalPX auction and congestion adjustment from the California ISO were $32.54 per megawatt-hour. The average for the Western Hub in PJM was $31.14.

Volatility. The removal of bidding restrictions in April 1999 also has been accompanied by increased volatility in PJM's prices. From April 1998 through March 1999, the standard deviation of on-peak average prices in PJM was $21.94, compared to $15.48 in Southern California. In the 12 months beginning April 1, 1999, the standard deviation of prices in PJM rose sharply to $60.17, while the standard deviation of California's prices fell to $10.33. Overall, across the entire 26-month period, the standard deviation of the California market was $18.54, as compared to $44.53 for PJM.

Bilateral Contracts. Prices in PJM and California are comparable to prices in bilateral contracts, as determined from market surveys. Prices in the PJM Pool have been slightly higher than bilateral prices in the Western hub of PJM, with an average difference of $0.22 per megawatt-hour. CalPX prices for Southern California have averaged $1.73 per megawatt-hour greater than in bilateral deals. In fact, much of that difference has occurred in the most recent two months. During the period from April 1, 1999 through March 31, 2000, bilateral prices in PJM were more strongly correlated with Pool prices than they have been in California (see Table 2).

Fuel Price Correlation. The relationship between oil and gas prices and electricity prices is interesting for both markets. Over the last 14 months, oil prices in California have been higher than in the Mid-Atlantic states. The marginal cost of operating a typical oil generator exceeded average electricity prices in both markets. For the period April 1999 through May 2000, California prices were $11.84 per megawatt-hour below diesel generation costs, while the difference in PJM was $7.91 per megawatt-hour. Although there is more oil-fired generation capacity in the Mid-Atlantic Area Council than in the Western Systems Coordinating Council, prices in both markets suggest that these generators would have been used infrequently. Indeed, simple empirical analysis (and Figures 2 and 3) fails to show a relationship between oil prices and electricity prices.

Spark Spread. Another way to view the data is to look at the average difference between electricity prices and the marginal cost of producing power with a gas generator, often referred to as the "spark spread." This figure is an important indicator of the performance of the two markets, as gas turbines are growing in relative importance and can be installed in a fairly short period of time. As Table 1 indicates, on-peak power prices in PJM have averaged $14.78 per megawatt-hour above typical marginal costs for gas-fired generation from April 1999 through March 2000, compared to a difference of $10.50 per megawatt-hour in Southern California. Figure 4 illustrates the comparison between the marginal cost of generating with a typical gas turbine and on-peak power prices in each of the markets. Volatility in the PJM system was centered principally in the summer air conditioning season, and was much greater during the summer of 1999 than 1998. California's prices also have exhibited volatility, but until April and May 2000, the peaks had not been nearly as high as those seen in PJM.

Congestion Management. Congestion differentials in the two systems reveal similarities and differences. Even though PJM's generation capacity is in broad balance with its load, the pool actively exports and imports electricity. As a consequence, PJM's important interconnections to other grids can be congested in either direction, as illustrated in Figure 5. On the other hand, California is a net importer of electricity. Thus, congestion there normally occurs in one direction, from the surplus-producing area to California. Figure 6 illustrates the congestion charges involved in moving power from Palo Verde (Arizona) to Southern California, and from the Nevada-Oregon Border to Southern California.

Nevertheless, there remains substantial uncertainty with respect to the frequency and level of congestion in both systems. That is because the principal causes of congestion...extreme weather and unexpected generation and transmission outages...are not predictable. In the California system, for example, the path from the Pacific Northwest (NOB to SP15 in Figure 6) is normally congested from May to July (the period of high water flows and hydropower imports), but the dollar measure of congestion seems almost random. Further analysis that considers the effects of weather and outages might clarify the pattern. The overview results presented here simply suggest that congestion impacts in the two systems are similar.

Commodity Price Certainty: The Unsolved Riddle

It has often been pointed out that, as a commodity, electricity is highly complex. It has three characteristics that combine to make it particularly difficult to manage in a market structure—instant decay, the necessity to balance the grid in real-time, and the externalities of transmitting through a common grid.

Electricity only has value for an instant. In order to define electricity in a form in which it can be bought and sold, it must be scheduled as a rate of flow over a block of time. Thus, a variety of "products" are offered in the market—day-ahead hourly blocks, on-peak and off-peak blocks, weeklong contracts during on-peak or off-peak periods, month-long contracts, balance-of-the-month contracts, and various combinations. This solution may be complicated, but it is not unique—the economy has many commodities that are tied to delivery schedules.

The Engineering Problem. Balancing the grid presents a difficult, but not impossible task for a market to resolve. In practice, there is a very small tolerance between the balance of load and generation. In nearly every jurisdiction the electricity market resolves the problem in two steps. First, generators and buyers contract for forward deliveries and schedule the power—typically one day ahead, but procedures vary. Plans often go awry, however, and on the day of delivery, actual dispatch and control passes to a central authority and the "real-time" market. Usually the ISO or system operator must buy and/or sell a small amount of power to bring the grid into balance. The process can be almost completely administrative. It involves adjusting generators based on their marginal costs, or in a market-based system through the advance purchase of ancillary services and real-time bidding. The various pools and ISOs have a variety of means to solve this problem. Some work better than others. However, because grid management ordinarily involves a very small percentage of total generation or load, the process usually need not impose a great impact on the market as a whole.

By contrast, however, transmission pricing is by far the most difficult task for a market to solve. The problem boils down to a lack of information for market participants.

The Information Problem. In most commodity markets, participants can assume that transportation costs are fixed (at least for the short term). That is, suppliers in a variety of locations know exactly how much it will cost to ship their product to various consumers. As a consequence, they can make bids based on a predictable set of production and transportation costs. Nevertheless, that is not the case for an electrical grid in the presence of congestion. The cost of delivering power from one location to another can change radically with weather and outages. Furthermore, transmission costs for any given market participant depend on the behavior of all other market participants.

Economic modelers often assume there are no information or contracting costs to trading. An auction is held without having to pay the auctioneer. In reality, of course, there are transaction costs associated with the operation of every market. Sometimes these costs are formal, as in an exchange or brokerage fee, while other times they are measured by the personnel time and effort required to find a counter-party and agree on the price and other terms of trade. Most markets operate through bilateral trading, so search time and information costs are important. Even though these per-unit costs may be high, if the number of buyers and sellers is limited and the commodity varies in quality and location, bilateral trading is likely to have the lowest transactions cost. Usually a formal exchange is cost-effective only if the volume is high. Exchanges require product standardization, a large number of buyers and sellers, and incentives for rapid turnover and trading.

When transportation costs are variable (depending on market outcomes) instead of a fixed amount, the transactions cost of bilateral trading can be high. Assume for a moment that buyers and sellers sort out prices and other contractual terms for delivering and receiving power on various nodes along the grid. The negotiations do not, however, take account of the impact of their collective decision on congestion, since transmission costs will be unknown until the power is actually scheduled. The initial market-based solution may prove infeasible or prohibitively costly. Furthermore, this solution would provoke another round of market negotiations and possibly another infeasible solution. Both PJM and California rely on a central authority combined with market bids to determine price differentials by zone or node in the presence of congestion.

Forwards and Futures: The Crucial Ingredient

The system of combined grid management and spot market operations adopted in PJM frequently has won praise as the most efficient market structure. Yet this emphasis on centralized control may not offer the most promising solution to the riddle of achieving price transparency and certainty for such a time-sensitive good as electricity.

Last year, for example, William Hogan endorsed PJM's combined model when he commented on the FERC's initial RTO proposal that ultimately led to Order 2000. The FERC's proposal, he said, "contains the elements of an efficient market design built around a system operator that coordinates the spot market. That is the only design we know of that is both internally consistent and actually works."12

Meanwhile, the California system has been criticized: "The bottom line is that California's transmission pricing makes it needlessly difficult for its ISO to relieve congestion and makes consumer prices needlessly high."13 Commissioner William Massey was even more pointed in comments wrote in January on a proposal by the CAISO for pricing congestion. He stressed that the FERC's order "sends a strong signal to the California ISO to move toward locational marginal pricing as it considers a comprehensive replacement congestion management approach."14

These comments, however, are based mainly on theory and concern about the engineering problem of congestion management. Their focus on resolving the congestion quagmire is not wrong, but it is incomplete. It has overshadowed the more important need to provide robust and transparent markets for prescheduled energy and forward rights.

Electricity prices ought to do more than manage congestion. Most importantly, transparent forward markets are crucial to planning thermal generation and load as well as for making investment decisions. Most thermal generators need advance warning as to whether it is cost-effective to offer power or shut down. Forward prices are the signals that determine such behavior. From an economic perspective, the PJM system attacks the problem after the fact. It puts the cart before the horse. Volatile pricing may be necessary to balance the real-time market, but there is no economic rationale for base-load generators and consumers with inflexible load profiles to receive price signals to which they are unable to respond.

Commodity markets generally make a distinction between physical and financial trading. The debate over the electricity industry's market structure often has been characterized as a distinction between financial and physical trading, but the difference is meaningless.15 No one can possess the electrons for more than a split second, and all contracting is effectively financial.16 What matters is not the form of the contract, but the number of markets that are required, the time(s) that they are open, and their efficiency and transparency.

The precedents for the market structures in PJM and California were set in the past by the evolution of their respective systems. PJM evolved from a system of central control and "economic" dispatch. The California market, on the other hand, adopted many of the elements of a power pool, but also was broadly based on the bilateral trading experiment of the WSCC, which was the most mature market-based pricing system for electricity in the United States. In bilateral WSCC trading, the most important market is (and was) the prescheduled market. This market is conducted during the morning of each business day, and determines forward prices from one to three days in advance. The purpose of the prescheduled trading is to give participants adequate time to plan forthcoming operations. The drawback of this system is that it cannot account for real-time problems and constraints dispatchers face during grid balancing operations.

Nevertheless, regardless of how the grid is operated, price information must be available early enough to give market participants time to plan generation and load. The PJM system has depended on bilateral trading and contracts for differences (CFDs) settled against LMPs. These contracts would allow participants to plan for firm prices at specific nodes well in advance of actual market operations. This solution, however, compounds the transaction cost problems described earlier. To ensure a firm price at a specific delivery point, market participants would have to enter into two contractual arrangements instead of one. Unlike futures contracts, CFDs are not standardized. That just increases transaction costs. And, because CFDs are not traded on an exchange, their pricing and terms are unlikely to be transparent. Most importantly, there is little evidence that CFDs are widely available at reasonable cost. California's market structure with respect to forward trading is more robust than PJM's, though the CalPX has not yet developed adequate trading volume in its block forward market.

Prices are signals. They must arrive in a timely fashion. The signal sent to Admiral Kimmel that the Japanese might attack Pearl Harbor was correct, but was either misplaced or it arrived too late.

In electricity markets, with their high transaction and information costs, there is a fundamental tradeoff that must take place. Should pricing be based on an imperfect signal that arrives in adequate time for buyers and sellers to adjust, or should the information be delayed until there is perfect knowledge of transmission costs but quite possibly inadequate time to react? The dilemma can be resolved by understanding that, at a minimum, the situation demands separate markets and settlement systems for forward trading and real-time operations.

1 From Ezra Beeman, resident of Paris.

2 On June 1, 2000, PJM plans to implement a day-ahead prescheduled market with a two-settlement system of accounting. In a very recent article, John Hanger acknowledged PJM's concern with high prices and volatility, , May 1, 2000, pp. 34-35.

3 Many factors other than market structure explain regional differences in prices, and the following statistics should be reviewed in that light. For example, the summer of 1999 was unusually hot and humid on the Atlantic Coast and relatively cool in California. In addition to the prices analyzed in this paper, both systems levy transmission charges and other fees, which include additional revenue to generators for a variety of uplift payments, ancillary services, etc.

4 California's SP15 zone will be redefined into several zones in the near future.

5 Hogan, William W., "Getting the Prices Right in PJM, Analysis Summary, April 1998 through March 1999, The First Anniversary of Full Locational Pricing," p. 7, www.harvard.edu/people/whogan.

6 PJM website, www.pjm.com, PJM Manuals.

7 See "California's PX: Monopoly No More," , July 1, 2000, p. 16. Nevertheless, at press time, it appeared that the California legislature was considering a measure that would postpone the effect of the PUC order until July 1, 2001.

8 In the initial round of bidding, participants do not have to identify specific locations for load or generation. Instead, bids are aggregated into a statewide "portfolio."

9 The principal alternative to a daily auction would be the continuous bidding model adopted by the Automated Power Exchange (APX) and used in many stock and commodity exchanges. The optimum design depends on expected volume, the length of time between contract execution and delivery, and the maturity of the market. Nordpool experimented with both designs and determined that for the day-ahead market a single auction was more reliable. CalPX uses a continuous bidding model for its longer-term block forward market.

10 With physicals, the trade involves a change of title and possession. Financials, on the other hand, are trade in future or forward rights to possession of the commodity.

11 The heat rate for a representative gas turbine is assumed to be 9,140 Btu per kilowatt-hour and diesel generation is assumed to be 10,000 Btu per kilowatt-hour. O&M expenses are estimated at $2 per megawatt-hour.

12 Hogan, William W., "The RTO NOPR: No Mandate, But a Plan That Works," , July 1, 1999, p. 18.

13 Kirsch, Laurence D. , "ISO Economics: How California Flubbed it on Transmission Pricing," , Oct. 15, 1998 p. 24.

14 FERC, CAISO , "Order accepting for filing in part and rejecting in part proposed tariff amendment and directing reevaluation of approach to addressing intrazonal congestion," Jan. 7, 2000.

15 Ruff, Larry E., "Competitive Electricity Markets: Why They Are Working and How To Improve Them," May 1999, www.harvard.edu/hepg.

16 Typically power sales contracts are written such that in the event of default, the obligation is for "liquidated damages."

Articles found on this page are available to Internet subscribers only. For more information about obtaining a username and password, please call our Customer Service Department at 1-800-368-5001.

Source URL: http://www.fortnightly.com/fortnightly/2000/07-0/east-vs-west-comparing-electric-markets-california-and-pjm