Market power after two years.
Robert McCullough is the managing partner of an energy consulting firm—McCullough Research—based in Portland, Ore. The firm specializes in business and public policy issues throughout the United States and Canada. McCullough Research also represents industries in Oregon, Washington, California, Idaho, and Utah. Before starting McCullough Research, Mr. McCullough was an officer at Portland General Corporation where he had responsibilities in finance, power marketing, and rate setting. He was educated at Reed College, Portland State University, and Cornell University in economics and finance. He is a member of the American Economics Association, the American Financial Association, and the Econometric Association. Mr. McCullough is also an Adjunct Professor at Portland State University.
Two years ago, the California market erupted in a year-long series of emergencies, price spikes, and financial crises. For a short while, a well-fueled public relations campaign had much of the world convinced that the state had run out of electric generating capacity as a result of its own unrealistic environmentalism.1 Now that the storm has seemingly passed, the more dispassionate view that this was market failure, rather than resource shortage, is gradually gaining the upper hand.
From the beginning, the electric industry was poorly prepared to handle a major market failure. The Western Systems Coordinating Council (WSCC)-the body tasked with the electric reliability of the West Coast of Canada, the United States, and Northern Mexico-never took an effective role in the crisis. Indeed, most of the debaters never even noticed that the West Coast had a reliability council that had been studying electric reliability issues since 1967. The WSCC, itself unfamiliar with a role that would bring it in conflict with its member systems, has never directly commented on the origin of California's problems.
The crisis in California ended with a whimper, not a bang. Although predictions for the summer of 2001 were catastrophic, the last California emergency took place soon after the implementation of a regional price cap. Simply stated, the crisis turned out to be a problem in institutions and not resources.
California's restructuring was characterized by six words-"bad design, bad incentives, bad results."
AB1890, the law that launched California on this path, was complex and difficult to understand. Its unanimous passage was evidence that every interest group had gotten its every desire. When every party to a negotiation leaves the table happy, there is a strong implication that they have been promised far more than can be delivered.
The basic design involved turning all power decisions over to an hourly market. This decision was so audacious and so misinformed that years after the design failed, regional utilities and industries are still having to explain to FERC that the hourly market has little to do with the industry. Further, reliability, the historical strength of the North American supply system, was only considered as an afterthought.
The design flaws were so extensive that the fundamental relationship of the state's Independent System Operator (ISO) to the WSCC had never even been considered.2 Reporting relationships were fragmentary and staffing and training was minimal. The ISO's motto, "Better Reliability Through Markets," was emblazoned above the attractive receptionists who manned the welcome desk at its Folsome, Calif. headquarters, in spite of the fact that the ISO's reliability principles were far less stringent than the system they had replaced.3
The crisis started with the announcement of a Stage 1 and Stage 2 emergency on May 22, 2000,4 and ended on July 3, 2001, with the final emergency declarations. The summer of 2001 actually saw declining prices and increased thermal generation. Every warning that price controls would reduce generation and contribute to the crisis turned out to be wrong.
Politically, the response to the onset of the crisis was like a scene from a frontier bar in an old western. Once the first punch was thrown, every interest group leaped into the fray with its own two-fisted agenda. Generators launched preemptive attacks on air pollution agencies, the California governor accused marketers and generators of price fixing, Secretary Richardson moved to seize scarce Pacific Northwest reservoirs, and municipals like L.A. and federal agencies like the Bonneville Power Administration were accused of profiteering. Within minutes, the bar was a roiling mass of special interest punching, kicking, and screaming. Policy responses were especially hopeless. The ISO spent months tinkering with price controls that always contained fatal loopholes.5 FERC dithered in appalling indecision for seven months, only to gun down one of the victims of the crisis-the California Power Exchange (PX)-on Dec. 15. Governor Davis's contribution was to negotiate deals with the marketers and generators that effectively fixed the unfair prices for years to come, while simultaneously assailing them for price fixing. Only after the composition of FERC was changed, were substantive steps taken-the adoption of a must-offer rule and WSCC-wide price caps.
Had the West Coast Run Out of Electric Capacity?
While pundits from San Diego to Maine opined daily on this issue during the crisis, the truth is that under the California ISO's rules, no one was certain exactly where the region stood. The WSCC had published, as they had done for the preceding 33 years, a summer load/resource appreciation indicating that while California supplies for the summer might be tight, that there was no immediate cause for alarm if 1,642 megawatts were available for import during June.6 In May, for example, they projected a reserve margin of 29.2 percent for California.
When the California ISO announced its first emergency on May 22, 2000, the industry was completely taken off guard. I can remember the exact moment of the first emergency. I was at a conference in Quebec when calls began arriving from utility and industrial clients as well as other industry experts. Every call started with the same question: "How can we be having an emergency in May, when loads are low and resources are high?"7
Under the complex structure of the California system, an emergency did not require a true shortage. The definition of an emergency is when the capacity offered the previous day in the computerized markets of the Power Exchange and ISO was less than 107 percent of forecasted demand. At the time, the ISO had no mechanism in place to determine if it was actually facing an emergency, or if the phone had just stopped ringing.8,9
McCullough Research's response to the crisis was typical of the electric industry. We asked a group of our clients-industries and utilities-to fund an investigation of the problem. Our initial estimate of the completion of the study was July 1. Little did we know that the task of accumulating data would require the intervention of every state regulatory body in the WSCC and would take months to complete.
To this day, it is unclear to whom the staff actually planned to sell access to its database. We were surprised that after 20 years in the industry, and a client list that included a dozen WSCC members, that we were told that we were not an appropriate user of the data. Seattle City Light then attempted to sign up and was told that it could sign up, but it could not show the data to its lawyers and consultants.10 After nearly a month of wrangling-and an article in the Wall Street Journal-a suitably complex arrangement was finally arrived at.11
The data from the WSCC (supplied to it from the ISO) did not support the hypothesis that California plants were out of service. Instead, the data showed that the plants tended to be operating during the ISO system emergencies, but were not being fully dispatched-even during the hours when actual emergency operating conditions were in place.
We were very surprised to learn that overall thermal operations in the California ISO's control area were running at levels far below the levels of comparable plants elsewhere in the WSCC. Comparing the dispatch rates with price data, our preliminary conclusion was that the California PX and ISO had suffered a one-time supply curve shift of 8,000 megawatts leftwards towards the origin. In simpler terms, the crisis looked like 8,000 megawatts had simply been removed from service. Eighteen months later, this is still our conclusion.
Enough time has passed that we now know the WSCC was not facing a capacity shortage at the time. On an annual basis, the WSCC publishes a 10-year forecast of resource sufficiency. This forecast is usually named the "10-Year Coordinated Plan Summary." One important part of the report describes the ratio between resources and loads for the previous year. Figure 1 shows this data from the WSCC reports from 1980 to 2001.12
In describing this chart to the House Energy and Commerce Committee, I used the metaphor of creating a happy household by ensuring that the ratio of snacks to teenagers always stayed high. In the utility industry, this ratio is called the reserve margin. A reserve margin of 15 percent means that the area has 15 percent more resources than requirements. This level-15 percent-is generally regarded as an acceptable margin, since one power plant in six would have to fail for an interruption in service to take place.
As the chart shows, the WSCC has fallen near this level frequently in the past decade. From 1991 through 1998, reserve margins routinely fell below 20 percent during the summer. In each case, actual interruptions of service were unnecessary, since we always had enough resources to meet load.
The situation in 2000 was far better than the situation we faced from 1991 through 1998. In 2000 we were able to get through the summer with a reserve margin above 20 percent.13
Pundits have identified the real problem in 2000 and 2001 as the serious drought that afflicted the Pacific Northwest during this period. As it turns out, this argument is wrong theoretically (reserve margins are always calculated assuming drought conditions) and factually (the serious drought started in 2001, not 2000).
It is important to take a moment and explain the nature of planning when a major resource like the Columbia River is to be used to meet the requirements of customers. Unlike thermal resources, hydroelectric systems are energy limited. A useful metaphor is to think of them like a car that has the ability to drive up to 100 miles an hour, but needs to visit the gas station every several hundred miles. Capacity-the ability to operate the machine at its maximum capability-is 100 miles an hour. As anyone who has lived past age 16 can testify, this capability can be demonstrated right up to the moment that an empty tank makes a visit to the gas station a necessity. The rated capacity of the hydroelectric system is always measured using drought conditions. Thus, we measure the speed of the car in the worst case-an empty tank-and report the capacity at zero miles per hour. It would not be prudent to announce an ability to meet load that could not be delivered during a drought year. In 1974, the WSCC recognized this fact by issuing instructions that the capacity of hydroelectric projects should always be calculated assuming drought conditions.14
Thus, the reserve numbers reported above have always assumed drought conditions. Even if the flows on the Columbia River were only at 92 percent of normal, this would not have affected its ability to meet peak loads.
As it happens, Columbia River flows during 2000 did not represent a drought. Flows in 2001 did, however. The emergencies within 2000 took place during a period of roughly average water. Put succinctly, there was a drought, but it started after the first summer of the California crisis.
Figure 2 shows the January through July flows on the Columbia since 1980. While 2000 was less than average, it was not a drought in any real sense. The most recent drought before 2001 was the drought of 1992 through 1994. Unlike 2000, the drought of 1992-1994 led to an announcement of service interruptions to the direct service industrial customers of BPA. The drought of 1987 through 1990 was also a far more serious operational problem for Pacific Northwest hydroelectric utilities.
The very straightforward conclusion that comes from the reserve margin chart when combined with the Columbia River flows is that 2000 was both a better year in terms of resources-22.9 percent reserve margin compared with 15.4 percent in June 1994-and Columbia River inflows-92 percent of normal compared to 71 percent in 1994. If these facts explained the emergencies in 2000, how did the lights stay on in 1994? The answer is that the organization of the industry rewarded meeting load in 1994. In California's complex structures, this incentive had been changed in 2000 and 2001.
When faced with this data, proponents of the resource shortage theory usually fall back on a secondary explanation that low plant operations were caused by local environmental rules. In fact, the environmental authorities granted exceptions, changed market rules, and accelerated permits. The comments of two of the most important districts-L.A. and San Diego-on Feb. 6, 2001, used very blunt language to describe the value of the generators' claims.15 A careful review of the environmental argument indicates that local air quality districts aggressively reacted to the crisis-far more aggressively than the generators.
By the planning methods we have used for the past century, the WSCC was not facing a capacity shortage in 2000 or 2001. While such methods have their shortcomings, a consistent review of the data indicates that if the emergencies of 2000 had been caused by a true shortage, our situation in 1994 would have been considerably worse. The major difference between the relatively stable conditions we experienced in 1994 and the emergencies in 2000 was in large part the difference in the operations of traditional utilities and the structure of the California market. In 1994, the generating plants belonged to the utilities. In 2000, the generating plants were dispatched according to the complex incentives hidden in the rules of AB1890.
Why Were Prices So High?
A simpler explanation is available, though. Starting in 2000, the WSCC had established a database showing the hourly plant operations of many of the plants on the West Coast. The California ISO provided plant data to the WSCC which, in turn, provided it to any interested WSCC member. While secrecy of operating data is a cornerstone of the California market design, the practice of secrecy at the ISO was unusual. The ISO provided this secret data in contravention of its FERC-filed tariff throughout the summer and fall of 2000.16 Any market participant equipped with this data would be able to easily adjust their operations to accentuate the California ISO's problems during an hour when demand was high. Curiously, Portland General Electric, Enron's subsidiary, did not contribute data to the database. Enron had access to the data of others, but did not welcome access to its own plant operations.
The California ISO has provided numerous charts that show that as its system approached peak, supplies offered to the California PX would begin to drop off. The resulting deficit would become an operating problem at the ISO. Once emergency conditions were declared, prices would skyrocket and supplies would reappear.
Documenting this was not easy. During the first part of the crisis, the generators' representative was the chairman of the ISO board. ISO market surveillance was rudimentary and timid. Generators' lobbying at the WSCC made access of the operating data to non-market participants slow and controversial.
Ironically, the hourly data is public outside of California-even today-as part of the EPA's emissions database. Unfortunately for consumers and policymakers in California, access to this data is usually delayed from three to five months.
Figure 3 shows the monthly operations of the units owned by Duke, Dynegy, Mirant, Reliant, and AES over this period. While plant operations in the rest of WSCC reached 100 percent, plant operations for the groups who have primarily profited from the crisis averaged 50.3 percent from May 2000 to June 2001. Interestingly, plant operations were actually slightly higher for the three months that followed price controls, even though market prices were significantly lower.17
We have been unable to explain the hourly operations of these five generators even after enormous effort. Frequently, plants went undispatched during system peaks and even during ISO-declared emergencies. Whistleblowers from the plant operations staff have indicated that their directions from management were inexplicable. Operations at plants outside of California have shown none of these problems. In fact, outside of the plants in Figure 4, operations have been as close to 100 percent of capacity as the owners could reach.
Many analysts break the California crisis into two periods. The first-economic withholding-represents the period when generators either did not bid resources into the PX and ISO, or made bids at unrealistic prices. A second period-physical withholding-takes place from November through June. While it is possible that the decision to take 50 percent of California's thermal units down simultaneously for planned outages was simply coincidental, an alternative explanation is also possible. After the ISO stopped providing operating data through the WSCC, generators may have simply switched to communicating their operating levels through planned and forced outage announcements. Regardless of the explanation, operations in the second part of the crisis roughly mirrored operations during the first portion.
From November until the onset of price controls, the five generators reported massive plant outages. The ISO did not reliably solicit or record plant outage data until 2001, so it is difficult to compare the outages in November 2000 to May 2001 with previous years for the same plants. The North American Electric Reliability Council (NERC) accumulates detailed historical data on the performance of similar plants-by size, technology, and fuel. Its data shows vastly lower outage rates on similar equipment.18
FERC conducted a preliminary investigation into the high rate of outages in February 2001.19 Although honestly undertaken, this report shows the futility of trying to regulate reliability standards after the fact. FERC's staff simply didn't have the background or access to do a detailed evaluation on a plant by plant basis. Surprisingly, FERC's report made no attempt to compare these units with similar units in the GADS data.20
Implementation of Price Caps-Correcting "Bad Results"
While predictions of widespread blackouts were common through the spring of 2001, FERC's decision to implement a WSCC-wide price cap appears to have had a significant impact on plant outages, short-term prices, and long-term prices in the late spring. As always, shifts in long term prices are the most interesting, since they are not affected by weather or other operating problems.
The onset of price caps in June led to the larger of the West Coast's two long-term price reductions in 2001.
The success of the price caps can be seen immediately. The presence of a counterweight to California's fragile power markets almost immediately returned long-term prices to the levels we have seen for the past 20 years. As FERC's recent report notes " the average price (both simple and weighted) at which the Western utilities sold power in the daily spot market was significantly below the price cap of $92/MWh."21 This is quite an understatement-by the end of June, prices had fallen to $43/MWh at Palo Verde. (See Figure 5)
While price caps are unlikely to work in a competitive market, the California market was hardly competitive. The incentives under AB1890 rewarded shortages. Once the ISO entered an emergency, it offered prices five to 30 times higher than normal levels for emergency supplies. Once FERC eliminated the ISO's ability to pay such distorted prices, generators in California were rewarded by producing more electricity, rather than less. All of the data indicates that once the incentives were repaired, plant operations improved and prices fell.
The shift in generator behavior is even more significant when each of the plants is modeled on an hour-by-hour basis from Jan. 1, 1997 through Sept. 30, 2001. Table 1 shows the forecasted operations of the plants based on market prices for energy, natural gas, and NOx RECLAIM credits.22
Actuals are significantly lower than forecasted levels from May 2000 through June 2001-the duration of the California crisis. After FERC's intervention in the market, the deviation between actual operations and forecasted operations fell from 6,030 megawatts to 699 megawatts.
Actual generation for the plants analyzed in Table 1 increased after the WSCC wide price cap was implemented, even though overall prices decreased markedly after the onset of price controls.
Overall, the standard model of economic dispatch of these plants fits very well before the crisis and after the crisis. During the crisis, the plants generated 6,030 megawatts less than a market model would have predicted.
Enron's Role in the Market
Clearly, enormous concentration in California markets was required for this to take place. FERC does not accumulate the data necessary to show the degree of concentration on a systematic basis. FERC does require energy marketers to file quarterly reports. Enforcement of this provision is weak. Some marketers fail to file their reports. Others file their reports in illegible or illogical formats. Still others, like Enron, do not specify any detail on the hubs where they bought and sold electricity.
Figure 6 (page 32) shows Enron's share of the major California hubs over time. The data used to generate this chart was taken from sales and purchases of major Enron trading partners, who do show where Enron's transactions take place.
Figure 6 matches our detailed research on Enron's trading activities.23 Enron's marketshare-for both sales and purchases-increased dramatically in 2000. By the fourth quarter of 2000, the evidence from FERC's quarterly marketing reports indicated that Enron's sales represent nearly 30 percent of the market. As Enron entered 2001, the growth of its wholesale operations appears to have stalled. Overall statistics indicate that Enron's physical sales declined after the fourth quarter of 2000.
In almost any other commodity market, a 30 percent marketshare is clearly sufficient to exercise price leadership. Pacific Gas and Electric's share of California wholesale markets before April 1, 1998 was similar, and its ability to use its scale to affect prices had long been observed.
Enron's sales directly to the California ISO were not large. Enron's sales at the hubs were vastly greater than its sales to the ISO. This simply reflects the fact that the market leader need not show up in every transaction. Price leadership sets the prices for all participants. Each transaction would reflect the price leader's price even though the price leader only had 30 percent of the market.
Do we know whether Enron exercised its market power in an attempt to increase prices during the market crisis that occurred between May 2000 and June 2001? No.
Publicly available data simply isn't that detailed. And while the California ISO continues to restrict availability of such data through its aggressive use of confidentiality agreements, the public debate will not become much clearer. The irony of the situation is that the ISO-the victim-has restricted market information to the market participants since they must have access to participate in the FERC refund cases and ongoing litigation. But the ISO has taken the same data out of the hands of the public, the press, and policy makers.
If arrogance was a clue to the exercise of market power, Enron's behavior during this period was legendary. During one transaction we were involved in, a junior Enron trader simply hung up on a senior executive of a Fortune 500 company because he could not move fast enough. This is market power with a vengeance.
The clarity of the evidence leads to a striking conclusion. If FERC had intervened in May 2000, the entire crisis might well have been avoided. FERC should have imposed a WSCC-wide price cap in May 2000 along with the "must offer" rule on California generation at the beginning of the crisis. If FERC had taken this step, the bankruptcy of Pacific Gas and Electric and the closure of industries from Arizona to British Columbia could have been avoided, and thousands of jobs could have been preserved.
Even better, allowing open access on the Pacific Northwest model-access for large customers ready and able to enter the bilateral market without the unnecessary California Power Exchange and ISO superstructure-might have conferred the benefits of open access without the costs of a distorted market.
The lesson for policy makers is that simple market rules apply only to simple markets. The California market failure was hardly simple and California's markets never reflected true competitive conditions. The complex AB-1890 structure required and continues to require extensive FERC regulatory intervention to operate in a fashion even remotely similar to a competitive market.
1 The phrase, "Ten years of rapid load growth without new resources," was a hallmark of an excellent public affairs campaign waged by marketers and generators in the California crisis. Interestingly, both parts of the phrase were strikingly untrue. The West Coast had a better load resource balance in 2000 than in previous years and peak loads actually were lower in the ISO's control area than they had been since 1997.
2 One irony of the California market failure was the phrase "Independent System Operator." The ISO board was was chaired by the representative of very generators that its internal market analysis division was monitoring, leading to a passive market surveillance that never seriously addressed the market power problems until FERC and the Governor of California replaced the board in toto. In retrospect it is clear that the ISO was hardly independent.
3 The California ISO purchases its reserves on a daily basis. In practice, this was like trying to buy insurance after your house had started on fire. Under traditional reliability planning, hydro-electric generation above minimum isn't even considered as a possible source of capacity reserves. The ISO attempted to use this frequently absent resource as the reserve of last resort during California's frequent emergencies from May 2000 to June 2001.
4 The ISO issues emergency notices when its forecasted hourly reserves fall below set levels -7% for Stage 1, 5% for Stage 2, and 1.5% for Stage 3. In practice, this mechanism has never worked. Emergency declarations have tended to reflect the need for additional operational rights for the ISO rather than hard and fast standards.
5 The ISO typically set price caps that they could exceed if there was a declared emergency or if the energy was imported from out of state. Prescheduled "exports" to Oregon that were exactly matched by emergency "imports" to California ballooned over this period, for example.
6 Assessment of the Summer 2000 Operating Period, Western Systems Coordinating Council, Spring 2000, page 3.
7 West Coast power planning revolves around two fundamental facts. First, loads are high in California in the summer because of cooling needs and high in the Pacific Northwest during the winter for heating. Second, the WSCC's largest
8 Alvin Alexanderson, the General Counsel of Portland General Electric, relates a story from his college career that describes a similar market failure. The local pizzaria in Ann Arbor had a policy of selling pizzas that had been ordered but not picked up at a discount. The students, quickly understanding the incentives this system created, would always order pizzas, but fail to pick them up. Since they often were sitting in the pizzarea at the time, they could then help the owner by offering to eat the abandoned meals. Unlike the ISO, the pizzaria owner was quick to see the folly of his ways and changed the rules.
9 The ISO did not start keeping a careful log of plant outages until 2001. At the time of the first emergencies, the ISO had no idea whether the resources were being exported out of state or were out of service. This led to many outlandish statements by generator representatives that price controls would simply make them sell their energy "out of state" in complete contravention of the laws of physics - Pacific Northwest loads are always quite low during the summer since air conditioning is not the major source of demand like it is in California.
10 This was a surprise to Seattle City Light, since their superintendent, Gary Zarker, sat on the board of the WSCC and had never approved any such restrictions. We deduce, without evidence, that other members of WSCC were lobbying energetically to keep this data out of the hands of possible critics.
11 Wall Street Journal, July 27, 2000.
12 Since the 2002 report isn't available, we have used the forecasted levels from last years report for 2001.
13 The ISO's Department of Market Analysis had invented a new way of studying the problem at the time by assuming that all resources that had not been offered to the California ISO had, in fact, failed or been exported. This simply adjusted the reserve margin down to the level available after market power was taken into account.
14 Criteria for Uniform Reporting Of Generator Ratings, Western Systems Coordinating Council, June 20, 1974.
15 February 6, 2001 letters by Barry Wallerstein (SCAQMD) and Richard Smith (San Diego APCD). Mr. Wallerstein's letter includes the phrase "[t]hese statements by AES are completely false and call into question AES' motivation in this matter."
16 California ISO Information Availability Policy, originally dated October 22, 1998, modified November 1, 2001.
17 This chart was based on data provided by the EIA. The EIA has faced substantial pressure to reduce the amount of such data available to public, as has FERC, the WSCC, and the North American Electric Reliability Council.
18 NERC's Generation Availability Data System (GADS) can be used to review the history for any type of plant. It is available on NERC's web site.
19 Report on Plant Outages in the State of California, FERC, February 1, 2001.
20 The 1982 through 1999 GADS report provides an availability rate (WAF) of 82.14% for units of 600-799 megawatts, 81.48% for units of 300-399 megawatts, and 84.30% for units in the 100-199 megawatt range. Availability factors have improved over time, even though the average age of the units has increased - contrary to the impression FERC staff received during their preliminary audit.
21 The Economic Impacts on Western Utilities and Ratepayers of Price Caps on Spot Market Sales, January 31, 2002, page 4.
22 The model uses heat rates derived from EPA hourly generation and fuel use data, MWh/Nox data from the same source, and market natural gas and electric prices. RECLAIM prices are the monthly average for coastal and inland markets.
23 Deconstructing Enron's Collapse, McCullough Research, January 10, 2002.