|analytical Q||Suggest a Link||Contact||Search||Energy|
by Anne Ku (April 1997)
A power pool is a situation where output from different power plants are "pooled" together, scheduled according to increasing marginal cost, technical and contractual characteristics (so-called must-runs), and dispatched according to this "merit order" to meet demand. The so-called power pools that exist today in the US are different from the poolco concept advocated by Bill Hogan of Harvard University.
Poolco refers to a mandatory bid-based power pool as opposed to a cost-based pool or a flexible pool. A mandatory pool is one in which all plant must be bid, scheduled, and dispatched centrally. No physical dispatch is allowed outside the pool. A flexible pool arrangement is one in which trading outside the pool is allowed.
With respect to pool terminology, tight and loose power pools (as termed in the US) relate to the centralisation and mandatory nature of power pooling arrangements prior to full open-access, as required by FERC Order 888 and are not yet bid-based. Bid-based pools usually refer to prices as bids, therefore do not have to reflect costs. There are cost-based bids, as in Colombia, which differs from a cost-based (non-bid) pool. Loosely termed power pools in the US, such as New York Power Pool (NYPP) or New England Power Pool (NEPOOL) are cost-based pools not bid-based. They, however, are evolving into bid-based pools through the filings made to the Federal Energy Regulatory Commission (FERC). The bids can be supply-side only, e.g. England and Wales; or supply and demand-side, e.g. Norway.
The pool serves two main functions: price determination and physical "trading." Like a stock exchange or an antique auction, the prices are determined by the bids followed by the actual exchange of stock or antiques. A central dispatch is believed to give more technical and economic efficiency than bilateral arrangements, as the power grid requires balance of supply and demand at all times.
Tomiak (1997) summarizes main characteristics of a few of the markets in Table 1.
Features of Power Markets
In addition to the above summary table, there are several features important in describing power markets, explained below and used in table 2.
Market Power and Gaming
We focus on what happens in a bid-based pool, as opposed to a market that is facilitated by bilateral contracts only or one which is based on a monopolistic central-dispatch of all plants in the system in which all costs and characteristics are known. In a supply-side only bid-based pool, prices do not always follow demand, since the presence of market power could induce gaming behaviour.
Market power is the ability of a particular seller or of a group of sellers to influence the prices to their advantage. The US Department of Justice defines market power as "the ability to profitably maintain prices above competitive levels for a significant period of time." Market power arises when markets are not perfectly competitive. Economic theory says that the equilibrium price in a competitive market is the marginal cost of the highest marginal cost unit necessary to satisfy aggregate demand. [This is precisely the mechanism by which a centralised power pool works: units are scheduled in order of their individual bid prices until demand is met. The last unit, which happens to be the highest bid, sets the price for that time period.] Thus, the extent to which market prices are sufficiently above this equilibrium price would suggest a market that is not fully competitive. Gaming is the act of exercising market power to reap large profits. Gaming differs from arbitrage in that the former arises from persistent opportunities rather than a temporary advantage created by inefficiencies in the market.
In the electricity markets, the presence and exercise of market power arises from three factors: ownership structure, plant portfolio, and market design (i.e. existence of a pool and intricacies in its rules).
The ownership structure is defined as the way the industry is broken up, i.e. the degree of vertical and horizontal integration. One measure of market power is market concentration as defined by the Hirschman-Herfindahl Index (HHI) which is the sum of squares of market shares of output of individual participants. The greater the concentration, i.e. fewer the number of participants and more unequal sized, the greater is the potential for market power. In other words, market concentration is inversely related to competitiveness. The duopoly situation in the UK has been studied at length with the main conclusion that the two largest fossil fuel generation companies have market power as a result of way in which the industry was first split up, and also have exercised that market power. However, Wolfram (1995) found that their exercise of market power could theoretically push prices even higher if not for the threat of regulation, entry deterrence, and financial contracts between suppliers and customers. Given a pools vulnerabilities to gaming as discussed later, what is the optimal market structure for maximal competition? In their frequently cited paper, Green and Newbery (1992) conclude that five equally sized firms would be sufficiently competitive, i.e. allow the benefits of competition to be reaped. In the extreme case of equally sized firms, Rudkevitch et al (1997) found that prices are still bid above marginal costs, suggesting that greater profits could be obtained with players of unequal sizes. This latter point has not been formally proven.
Plant portfolio refers to diversity of the capacity mix, whose value comes from owning plants which are strategically located ("constrained on"), with flexible characteristics (can adjust quickly), and low cost (to capture the gap between final price and the bid). A generator who owns a variety of plant has the power to influence prices in different markets. Diversity as a strategic advantage is discussed in the section on gaming mechanisms.
Finally, pool rules could enable "gaming" to occur. A mandatory (as opposed to voluntary) pool is one in which no physical bilateral trading is allowed. Such a pool acts as the sole buyer of electricity for all sellers and sole seller of electricity for all buyers, as it sets a single price during each period (hourly or half-hourly). All power must be prescheduled through the pool, which then determines the price. This is the context in which we analyse market power and gaming opportunities.
A centralised power pool is particularly vulnerable to gaming for the following seven reasons, which are summarised in Table 3.
First, as electricity cannot be readily stored, supply and demand have to be equal at all times. This means that only those units already running or those flexible enough to run instantaneously would be able to meet fluctuations in demand. These ready units therefore capture the "balancing" or "regulating" market.
Second, demand is relatively inelastic so that it does not automatically adjust downwards when prices are high, nor does it increase when prices are sufficiently low. This implies that generators are able to push prices up. [Note: price responsiveness varies for different classes of customers and different types of contracts. The majority of the load in Colombia is residential, which tends to be relatively inelastic as a whole.]
Third, wholesale demand is relatively predictable. Figures 3 and 4 on the next page show aggregate load (demand) in February 1996 for England and Wales and Colombia respectively. The weekend effect (deviation from weekday) and general daily shapes are quite predictable. In both countries, the forecast demand is given to the bidders beforehand.
Four, plants do not have continuously varying variable costs. Generators could learn the shape of the supply curve by tracking, e.g. varying bids in the early stages of a pool operation to "discover" the shape.
The predictability of points three and four combined with knowledge of other generators bidding patterns (after awhile, daily bidding tends to an infinitely repeated game) and maintenance schedules (the disclosure of which is sometimes required by regulation), it is not difficult for participants to strategically bid to "game" the system. One method is by withholding the amount of capacity which is available, thereby shifting the supply curve to the left and pushing prices up. This method does not work in the flat portion of the supply curve. For curves that are relatively steep or "gapping", this method works very well.
Fifth, a power pool is not one market but several, as defined in time and space, i.e. temporally and geographically. The market during peak periods is very different from off-peak. Similarly, regional demand, location of plant, and transmission congestion can create "load pockets" which are effectively sub-markets. It is possible to game by using a plant (that sets the price) in one market to influence prices in another. This is called "leveraging," and will be explained later.
Sixth, constraints caused by transmission and ancillary services call for some plant to run simply because of their location and technical characteristics. Over time, it becomes evident which plant will be "constrained-on" and therefore their owners would seek a "premium" for them, by bidding much above their marginal cost. Henney (1997) discusses the effects of transmission constraints on power prices as well as how plants can also activate constraints. In the UK, there are 10 major constraints under normal conditions and 40 further constraints which are mostly local. While some relate to a single plant, others are not simultaneously active. Many are nested or overlapping. Bidding patterns can also activate a constraint.
Seventh, if a pool is not designed to be perfectly competitive from the start, regulatory intervention becomes necessary to mitigate market power. At present, there does not exist a perfectly competitive power pool. The Nord Pool is the closest, albeit only 20% of power is through a pool mechanism, the rest is bilaterally physically traded. Patchwork regulation in the UK has created a very complicated pool with the latest set of pool rules running about 500 pages!
This section summarises specific gaming mechanisms as observed in the UK pool and analysed by the regulator and various academics. The extent to which gaming can occur depends on the market structure, plant ownership, (threat of) regulatory intervention, and pool rules. These analyses show that gaming opportunities are not solely due to market concentration, which so far has been the only recognised and standard measure of market power in the industry.
According to Rosen and Kroll (1996), a generator is able to exercise its market power in one market to influence its revenue in another market, thereby maximising overall profit for the firm, if it owns at least two types of plant: one which would definitely get dispatched because of its low bid, and the other which is at the margin thus capable of setting the marginal price. Multi-plant owners can bid their cheaper stations at marginal cost, or less (to make sure they run) and then raise the prices of their other stations, creating additional profits for those which are already running. The more units owned that are bid below the marginal price, the greater incentive the generator has to bid above marginal cost. If units are well-distributed across the supply cost curve, then the owner could exercise this high-price bidding strategy at almost any demand level. "Leveraging" refers to this ability to earn extra profits on plants already dispatched by exercising market power on the margin plant, i.e. the plant that sets the price for that hour or half-hour. Since baseload, mid-merit, and peakload plants represent generation in different markets due to the temporal delineation of power markets, leveraging allows the exercise of market power in one market to influence the price of power in another market. In other words, the generator could use one plant to "leverage off" or "take advantage of" another plant.
The game of pool is repeated everyday. After awhile in a market with limited number of big players, competitors are able to predict each others pricing and production behaviour based on past activity and the underlying fundamentals involved in the industry. Leveraging then benefits all generators simultaneously, i.e. a generator leverages off another generators margin-setting plant to reap higher profits. This type of leveraging is sometimes referred to as tacit collusion.
The amount of extra income to be leveraged depends upon the steepness of the supply function. The steeper it is, the more likely the low-bid plants will gain from the prices set by higher-bid plants.
Capacity Availability Withholding Bidding Strategy
After studying prices, quantities, and bids over a four year period, Wolak and Patrick (1996) found that major generators in the UK have been able to use existing market structure and pool rules to achieve prices (Pool Selling Price) significantly above marginal cost and average total cost. Together they have set the system marginal price in 84% of the half-hourly periods from 1st November 1995 to 31st March 1996. This so-called capacity availability withholding bidding strategy is applied under the following circumstances.
One, the vast majority of excess revenues due to the exercise of market power arises from extremely large within-day price swings. These swings occur for extremely short durations lasting no more than two or three half-hour periods. The days in which these very high priced load periods occur tend to follow one another in the same week. This qualitative feature of price behaviour is constant across the four years of the studied sample.
Two, contrary to conventional expectations, producers do not exercise market power by explicitly bidding prices for each unit (generating set) substantially in excess of its marginal cost. This would be too obvious to the regulator and other observers. Instead they manipulate the maximum amount of its generating capacity made available to the pool and the prices bid in for each generating set made available. During periods in the day when forecast demand is known to be very low, the two duopolists will bid very aggressively in terms of available capacity relative to demand and prices so that the marginal price will be very low to deter entrants. During periods in the day of extremely high system demand, they will not bid as aggressively because whatever they bid, a large fraction of their capacities will be called upon to serve forecast demand. They will therefore take advantage of their superior market position to raise the pool price.
Several features of the environment contribute to the ability of the duopolists to maintain prices in excess of average cost. Transmission constraints of interlinks (to Scotland and France) and the actual amount of capacity that other producers own enable a producer to establish the upper bound that it can bid into the pool during each load period. Since there is also a finite upper bound on maximum residual demand these two firms face, the only source of demand uncertainty left for the purposes of price-determination is the amount of capacity that will be supplied by remaining players. The combination of upper bound of capacity and demand makes it easier for the duopolists to estimate the residual demand so that when expected aggregate demand conditions favour it, they can use their availability declarations to obtain very high values of pool price.
The day ahead demand forecast estimated by the pool is completely inelastic with respect to the value of the system marginal price. Because there is no demand side bidding and the confidential bids are supplied all at once by the bidders, prices are set once and not reset throughout the day. Thus, the generators are able to adjust their bid functions to intersect demand at the highest level through a combination of availability and bid declarations.
The firms are also able to engage in leveraging tactics because they each own a diverse mix of generating capacity: baseload, mid-merit, and peaking load plant. This diverse mix implies a steadily increasing (at an increasing rate) marginal cost function (up to total capacity of the generator) within any load period.
The nonlinear relationship between forecast demand (expected peak demand plus reserve margin) which sets the system marginal price and the loss of load probability (LOLP) which determines the amount of capacity payment indicates that by strategically withholding capacity to obtain a small margin, large values of LOLP and hence large values of capacity payment are possible. As Colombias pool does not make use of this daily loss of load probability to assign capacity payment, it is less likely that generators will be able to manipulate daily availability to gain on capacity payment.
The very high positive relationship between capacity payment and uplift suggests that when generators gain on the capacity payment, they also gain indirectly on the uplift. In other words, they are able to declare the bid price and capacity availability such that very high values of capacity payment and uplift coincide with periods of high values of system marginal price. By declaring baseload capacity unavailable, the generators can control the steep part of the bid function.
Bid Low or High?
To participate in the pool, generators have to decide which plants to declare available, how much capacity to make available paired with the prices they are willing to accept. A key decision is when to bid low and when to bid high, relative to the plants running costs and relative to other competitors.
Lucas and Taylor (1995) found that plants with lower running costs are bid below their costs, while the more expensive ones are bid more than their costs. This reflects a trade-off of capturing revenue from low prices over longer periods versus high revenues over shorter periods. Lower running costs tend to correspond to those "inflexible" plants that are contractually or technically constrained. A technically inflexible plant cannot start or shut down as quickly as a flexible one. Plants which must be run to boost system voltage and frequency in areas which import power command a premium for their flexibility. Therefore owners of these plants can afford to bid high and still expect the plants to be run. A contractually constrained plant is one which must run for so many hours in a year. The new Combined Cycle Gas Turbines, for example, are technically flexible but contractually constrained, and as a result, are bid at zero price to ensure they are scheduled in the dispatch, effectively as baseload plant.
The bid and forecast demand determine how likely the plant will be selected: the lower the bid, the higher the probability of being selected. Bidding close to marginal cost is optimal for the owner of a single plant. A higher bid would reduce its running time, with little impact on the price it receives, which is generally set by other, more expensive plant. In 1992, the UK regulator found that generators tend to submit low bids during the week when demand is high and they could expect to run for long periods as part of the unconstrained schedule. They submit higher bids at weekends to maximise their revenues while transmission constraints are binding, since lower levels of demand means that the plants would not be able to run as part of the unconstrained schedule.
In a duopoly situation, Von der Fehr and Harbord (1993) found that the two generators would bid above marginal cost, unless demand is so low that it could be met by either generator on its own. This is supported by Rudkevich et al (1997) who show that supply-side participants would always bid above their marginal costs as this is the Nash equilibrium. In general, profit-maximising producers have the incentive to either increase their bid prices (for plants which are constrained-on or at the margin) or to withhold some capacity, where possible, to raise the market clearing prices of electricity paid to each dispatched unit. The ease and frequency with which this can be done indicates the amount of market power that can be exerted in a pool.
In the UK, the generators are rewarded as long as they declare capacity available. In other words, even if the plant is not scheduled to run or not called to run after it has been scheduled, it will receive some compensation for being available. Therefore prices are bid according to the strategic advantage of each plant (must-run, constrained-on, or ability to ramp up and down, etc) rather than at marginal cost.
Powell (1993) reports that generators profit from selling contracts at more than the expected spot price, while the buyers benefit from the reduction in the generators market power that results from their contract cover. This study has not explored the relationship between pool activity and separate financial contractual arrangements (mainly for risk management purposes) which could have a significant effect on bidding strategies.
Table 1 Main Characteristics of Restructured Wholesale Power Markets
Table 2 Overall Comparison of Competitive Power Pools @April 1997
Table 3 Vulnerability of a Power Pool to Gaming
Types of Analysis to Determine Extent of Market Power and Opportunities for Gaming