End Notes

1. The shift to modular construction is particularly dramatic in the resource plans issued by the Northwest Power Planning Council (NPPC). Their 1991 plan describes the long term marginal resource as a combined cycle generator fueled by gas from a coal gasification plant. The levelized cost was 60 mills/kwh; a typical 420 MW plant would take 7 years to build and cost over $1 billion. Just five years later, the NPPC described the region's long term marginal resource as a gas fired, combined cycle combustion turbine (CCCT). The levelized cost was only 30 mills/kwh; a typical 228 MW unit would take only 4 years to build and cost around $150 million.

2. System dynamics has its roots in control engineering and has been defined as "that branch of management science which deals with controllability" (Coyle 1977, p. 2). The method was pioneered by Forrester (1961) and has been put to good use in the regulated electricity industry (Ford 1997) and the competitive electricity industry (Bunn and Larsen 1992, Lomi and Larsen 1999; Vlahos 1998). System dynamics is especially useful when we need to understand the feedback mechanisms in a system.

3. Demand-side-management programs are falling out of favor as the utilities brace themselves for an era of competition. The current emphasis is on "industry transformation" programs in which utility funds are used to leverage important changes in a national industry such as lighting and motors. Both of these programs are outside the scope of the model.

4. This is the approach adopted by the Aurora model described in Appendix B. California has advanced further than other western states in restructuring; its rules are better defined and more amenable to simulation analysis.

5. More than 80% of California's electricity has been traded through the Px since it opened in March of 1998, and the exchange is cited as a reference on daily prices for other markets (Moore 1999).

6. Once again, we follow the approach used by Aurora. Ignoring demand side bidding is reasonable if one believes that consumers place a high premium on a reliable supply of electricity. (Rose (1995) argues that customers are willing to "pay at least 100 times the average price" to avoid curtailment of electricity. ) On the other hand, certain consumers with huge electricity requirements (like the aluminum smelters) can estimate the maximum price of electricity that would allow them to continue operations in a profitable manner. These industries might turn to demand-side bidding in the Px, or they might negotiate long-term contracts with special interruptible provisions to deal with their unique situation. Simulating these actions is beyond the scope of the current model.

7. Administrative limits are a difficult topic since a discussion of how they would be imposed is an admission that the market mechanism may not work effectively. It is not surprising, therefore, that information on price caps is limited. In testimony on the California pool, Grow concludes that "Apparently, the intent is that the market clearing price should be high in this oversubscribed condition and that some of the demand might drop in response to the high price. Perhaps details will be filled in later" (Grow 1996A, p. 3). The model imposes a circuit breaker mechanism whenever a shortage of energy develops and the Px Price has climbed to a point that is beyond a reasonable value to stimulate generation. The cap is based on the variable cost of the region's most expensive gas-fired unit or the total, levelized cost of a new CCCT. The model takes the larger of these two costs and adds a user-specified reserve margin. Under base case conditions, the administrative cap would be around 43 $/mwh. The 43 $/mwh cap may sound low compared to "price spikes" that have appeared in competitive markets. Price spikes in the California Px, for example, exceed 150 $/mwh during August and September of 1998 (Moore 1999, p. 11). But remember that the 43 $/mwh is an average over the entire year, and this average value may be created by high price caps imposed for a small fraction of the hours in the year. To illustrate, consider Grow's (1996B) calculations for a simple system with three types of generators and market clearing prices ranging from a low of 17.5 $/mwh to a high of 40 $/mwh. The annual average turns out to be 30 $/mwh. To match the administrative limit adopted in this paper, one could assume that a price cap of around 330 $/mwh applies for 5% of the hours in a year.

8. To simulate price feedback on demand, one must combine the Px price with the rules adopted in different states to arrive at the retail price to the customer. In some cases (as in Assembly Bill 1890 in California), the rules may turn out to be quite complicated. The complications are the inevitable outcome of lawmakers' attempts to "provide something for everyone" (Richard and Lavinson 1996). These rules strive to protect utilities from a low Px price by adding a competitive transition charge (CTC) to the consumers' bill. In so doing, the consumer is shielded from the benefit of low rates in the short-run. To protect consumers from high Px rates, on the other hand, the state imposes a variety of targets on retail rates that may be charged in California. The overall effect of these rules, is to shield the retail consumer from major variations in the Px price. For these reasons, it is reasonable to assume that the growth in demand may be specified independently from variations in the Px price.

9. We might see a reduction in nuclear generation if the owners feel they cannot control operating costs. On the other hand, nuclear owners may be stimulated by the competitive market to improve maintenance procedures. In the process, they might discover what Farney (1997) calls "hidden capacity" or what CERA (1997) calls "capacity creep."

10. These adjustments are made quite quickly as the simulation proceeds through time, and the Px Price finds its way almost immediately to the appropriate value to clear the market. This simple heuristic has been tested under a variety of conditions to verify that the model moves extremely quickly to the market clearing price. A sudden reduction in the aMW from hydro, for example, leads to a sudden increase in the Px Price to clear the market. A gradual increase in the demand for electricity, on the other hand, leads to a gradual increase in the Px price to clear the market.

11. Aurora's Px Prices are calculated hourly for several delivery points in the WSCC. The hourly prices for delivery at Mid-Columbia were combined into an annual price forecast of around 20 $/mwh which provides a credibility check on our initial result

12. Grow's purpose is not to advocate Px Prices in the range of 1,000 $/mwh to stimulate investment. He elaborates on his purpose in a separate article (Grow 1996B) which describes a simplified system comprised of peakers, intermediate and base load units.. All generation is satisfied; the average annual price turns out to be just under 30 $/mwh; but the generators net operating revenue is not sufficient to cover their fixed costs. He then calculates the value of a demand-side bid which, if allowed to clear the market for 5% of the hours, will make all generators whole. He concludes that "the solution is easy, because the required bid equals the full revenue requirement of a peaking unit that runs only 5% of the year." In this case, a demand side bid of 168 $/mwh which sets the price for 5% of the hours "will make all generators whole."

13. The proportionality constant, alpha, is a highly uncertain parameter which is set at 2 for a base case simulation. If investors' profitability is at 20%, for example, initiations are set at 40% of the initiations that would be needed to keep pace with demand. If profitability increases to 100%, initiations are set at 200% of the initiations needed to keep pace with demand. Alpha represents a simple guess as to investors' responsiveness to perceived profitability. It would be desirable to estimate this parameter from data on permit applications, but this estimate would be clouded by the apparent tendency for investors to stock up on pre-approved permits well in advance of profitability. Since we can only guess at the size of this parameter, it's important to remember that its value is highly uncertain and should be the subject of sensitivity analysis.

14. The price cap is based on the cost of the most expensive generating unit plus an administrative reserve. Since the price of natural gas is increasing over time, the price cap increases as well.

15. The destabilizing effect of faster growth is explained in text book examples of logistic models with a lag in the negative feedback (Hastings 1997, Ford 199).

16. The model does not calculate carrying costs during the construction interval. The construction cost and the length of the construction interval are both specified by the user.

17. The sensitivity tests focus on parameters that are uncertain today. But our uncertainty about the length of the construction interval today doesn't mean that investors won't have good information when it comes time to invest.

18. Dickens (1998) reviews the capacity payments adopted in different markets around the world. He reports that several markets have evolved without separate pricing for capacity. These include the markets in Norway, Victoria, Australia, and New Zealand as well as the California Px. "In those markets without a separate capacity price, the market designers appear to have believed that energy prices alone will provide sufficient incentive for new investment in response to market demand." Some of these markets have been running for several years, but Dickens cautions that there is not enough experience from these markets to conclude whether the energy prices will be sufficient to stimulate the needed investment.

19. The most relevant argument is by Michaels (1997) who believes it is unwise to rely solely on energy markets. He acknowledges that an energy market can lead to efficient allocation of energy from existing plants, but his concern is for new plants. He argues that an efficient pattern of new construction would occur only as an "accidental" outcome of an energy market. Energy markets will "display unstable prices and chaotic investment patterns even if, in the long run, investors in generation earn competitive returns on average." Michaels concludes that "timely investment in economically warranted capacity is as necessary for long-run efficiency as marginal-cost allocation of existing capacity is for short-run efficiency, yet very little attention is being given to capacity planning. If an efficient energy market gives rise to an inefficient capacity market, utilities may become the investors of last resort."

20. Michaels is not alone in calling for capacity payments (Conkling 1995, McCullough, Rose 1995 & 1997).

21. The 5 mills/kWh is based on an annual fixed cost of 40 $/kw per year. Rose increases this cost by 18% to cover a reserve requirement and spreads the cost over the 8760 hours in a year.

22. Rose has argued that energy markets will carry a "hidden capacity price" that will appear in the form of spikes in the energy price during peak periods in a year with tight supply. He might argue that this hidden price would disappear if capacity payments were made explicit -- investors would respond with adequate investment in new CCCTs, and there would be NO impact on wholesale rates. This paper adopts the more cautious view that wholesale rates will increase in the short term.

23. The Trust Transfer Account involves charges on "rate reduction bonds" issued by California's distribution companies to "make up" for the AB1890 requirement to reduce rates by 10%. The public purpose charges pay for aid to low income customers, efficiency and renewable programs, and research and development. The remaining small charges pay for nuclear decommissioning and a state regulatory fee.

24. Hirst (1996, p. v) estimates stranded costs for the entire U.S. from $100 to $200 billion. He believes stranded costs would be "especially severe in California, New York, Ohio and Pennsylvania" (Baxter and Hirst 1995, p. 18).

25. The analysis in this paper could be expanded in several directions:

	First, greater diversity in assumptions (fuel costs, O&M costs) for both existing generators and for potential generators would be a useful addition. (With greater diversity, one might see greater stability.)
	Another important improvement is to allow existing capacity to be retired over time. The retirements would probably be concentrated during low points in the price cycle, and they could stabilize or destabilize the system, depending on the magnitude and timing of their response.
	The model could also be improved by explicit treatment of the region's transmission system and the construction of new transmission capability.
	And finally, the model could be expanded to treat capacity issues in a more explicit manner. With this improvement, the model could be used to further explore the challenges of adopting the capacity payments recommended in this paper.