An important question in the newly competitive electric industry is whether individual generation companies can exert "market power," that is, push prices above competitive levels without losing customers. Regardless of their actual bids on the spot market, all generators operating on a regional system at a given time receive the "market-clearing" price, defined as the price bid by the last generator to be scheduled. In theory, the market-clearing price should be the marginal cost, that is, the cost of generating the last and therefore most expensive unit of electricity required. But under some circumstances the last generator scheduled can bid a price higher than his actual cost. In addition, generators can occasionally cut back the amount of generating capacity they bid into the market. As in any commodity market, a reduction in supply brings an increase in price (for the same level of demand). What are the conditions under which generators can exert such market power? And how often do those conditions prevail on the electricity spot market?
To tackle those questions, Ms. Visudhiphan and Dr. Ilic are taking a novel approach to modeling the spot market. The traditional approach involves tracking systemwide (aggregated) supply and demand information when supply and demand are in balance (equilibrium conditions). A picture of the market's overall behavior emerges, but understanding what individual participants are doing and how the market behaves over time is difficult. The MIT researchers are instead simulating market behavior by starting with individual generators as they decide how much supply to offer at what price. (The effect of transmission constraints on market power is being addressed by researchers in the Center for Energy and Environmental Policy Research, led by Professor Paul Joskow, Elizabeth and James Killian Professor in the Department of Economics.)
The first step in simulating the decisionmaking process is to define the characteristics of a single generation company. What types of generating units does it own? How big are they, and how much does it cost to run them? How risk averse is the firm? Based on such information, the researchers simulate the decisionmaking process of the firm as it develops its bids. They then perform similar analyses of additional generators with other characteristics. On the demand side, the researchers assume a forecasted load that is fairly unresponsive to price (a reflection of conditions on most current spot markets). Based on the individual supply bids and assumed demand, the researchers simulate the bidding process used on spot markets, thereby calculating the changing price of electricity over time.
In an initial series of case studies, Ms. Visudhiphan and Dr. Ilic demonstrated how the assumed "bidding strategies" of individual generators influence estimated market prices. In their analysis, each generator decides first how much capacity to sell and then the price it will charge if it is scheduled to operate. If the opportunity arises, the generator may withhold capacity to push up market prices. Three bidding strategies are assumed. In the first, all generators simply increase their bids each time by a prespecified amount. In the second, each generator changes his new bid based on how successful he was in the previous bid. (If the bid was successful, he submits the same price. If the bid was unsuccessful, he submits a lower new price.) In the third, each generator chooses as his new bid the highest market price seen over the past 15 bidding samples. The different bidding strategies generate market prices that display differing trends over time. The first strategy leads to a persistent increase in price, while the second and third strategies result in prices that are fairly stable. Clearly, the ability to predict bidding behavior can provide insights into the dynamics of market prices.
Using their new approach, the researchers examined which types of firms would be able to exert market power and under what conditions. They analyzed the behavior of 13 hypothetical generators with differing capabilities, constraints, and objectives. They concluded that the key measure of a generator's potential for exerting market power is the generator's capacity relative to excess capacity at a given time.
An example will demonstrate. Assume that New England has a total capacity of 15,000 MW, and tomorrow's demand is expected to be 14,000 MW. Excess capacity for tomorrow is therefore 1,000 MW. Under those conditions, a generator that owns 2,000 MW of capacity is critical. If that generator chooses not to operate, demand cannot be met. If he offers 1,000 MW at an extremely high price, the system operator will have to take the offer and that high price will prevail. Significantly, the generator's 2,000 MW of capacity is just as critical if New England's total capacity were 30,000 MW and tomorrow's demand were 28,000 MW. Thus, the important measure is how a generator's capacity compares to excess capacity; and when excess capacity is low, more generators will be in a position to push up prices.
Once again, the performance of the electric industry cannot be measured using standard techniques. The US Department of Justice usually judges the market-power potential of a company based on its "market share," that is, the amount of a commodity that the company controls relative to the total amount of that commodity traded. However, according to the researchers' analyses, an electricity-generating company with a relatively small market share may be able to exert market power. Demand for electricity is much less responsive to prices than is demand for other commodities, so generators can more easily push up prices without losing customers. As a result, opportunities for exerting market power are more prevalent in the electric industry than standard analyses suggest. For example, analyses based on the standard Department of Justice index suggest that generators in New England have little opportunity to exert market power. But an analysis using the new MIT index shows that the potential for firms to exert market power is substantial. It varies dramatically from day to day and season to season and is particularly high during the summer.
Thus far, the simulations involve generators developing their bidding strategies based only on past market prices and forecasted demand. How would their strategies change if they had information on future plant outages, maintenance schedules, or the bidding behavior of other generators? How would bidding change if generators had the option of selling long-term contracts? And what would happen if demand actually responded to price? The ability of customers to reduce their demand when prices rise profoundly changes the spot market. Indeed, generators' ability to manipulate electricity prices might simply disappear.
This research was supported by the MIT Energy Laboratory's Consortium on New Concepts and Software for Competitive Power Systems: Operations and Management and by the MIT Department of Electrical Engineering and Computer Science. Further information can be found in references.
Last updated: 26-Jul-2001
Copyright © Massachusetts Institute of Technology 2001. Material in this
bulletin may be reproduced if credited to e-lab.