May 5, 2016

Dr. Frank Rusco
Director of Natural Resources and Environmental Issues
U.S. Government Accountability Office
441 G Street NW
Washington, DC 20548
(202) 512-3841

cc: Senator Lisa Murkowski
Senator Maria Cantwell

Re: Response to U.S. Senators' Capacity Market Questions

Dear Dr. Rusco:

On November 19, Senators Lisa Murkowski and Maria Cantwell from the U.S. Senate's Committee on Energy and Natural Resources issued a letter to the U.S. Government Accountability Office (GAO) requesting that the GAO examine the efficacy of U.S. electricity capacity markets.1 The letter asked how capacity markets affect reliability, costs, and the generation mix compared to traditionally-regulated systems. We offer below our responses to each of Senators' questions as an input to your assessment. Our responses are informed by many years of consulting experience analyzing the very issues raised by the Senators, conducted on behalf of clients from all sectors of the electricity industry.2 We provide references to additional information and industry studies where possible.

The following responses to the Senators' questions use the same numbering convention, with the original questions repeated for reference.

1. We are concerned about the relationship of the increments of new capacity cleared in an auction and the increments of new capacity actually installed. Two recent surveys suggest that only a small fraction of new capacity has been built in organized markets except under bilateral power purchase agreements or direct ownership by LSEs [Load Serving Entities].3 Additionally, it is our understanding that except for one sub-region within PJM, capacity has never cleared above the "cost of new entry" in PJM or MISO. These observations prompt us to ask a central overarching question:

1a. Since their establishment, how effectively have capacity markets influenced the construction, maintenance, or retirement of generation in order to ensure resource adequacy and reliability in a cost-effective manner?

After more than a decade of experience, the U.S. capacity markets have demonstrated that they generally fulfill the design objective of meeting "resource adequacy" requirements cost effectively.4 They do so by establishing the quantity of capacity needed, and procuring that capacity through a competitive auction that is open to all types of resources. This auction-based, competitive format has proven effective at leveraging competitive forces to attract the lowest-cost combination of available resources, including demand response resources and the refurbishment and upgrades of existing resources. Capacity markets have created a level playing field that enables competition among new and existing generators, incumbents and new entrants, internal supply and imports, traditional and new types of technology, generation and demand-side resources, and centralized and distributed resources. These competitive forces have consistently achieved required reserve margins at prices below the system operators' estimates of the long-run costs of new generating plants.5

The success of capacity markets to date does not mean that they cannot be improved. In fact, each of the markets has encountered challenges that needed to be addressed over time, and has areas for improvement.6 We anticipate the future will continue to pose new challenges as market forces evolve. As long as these solutions comport with fundamental economic principles and rely on sound analyses, we anticipate that capacity markets will continue to perform well.

The PJM experience provides a good example of capacity market performance in achieving reliability objectives cost-effectively. That capacity market was instituted in 2007 at a time when PJM anticipated impending shortfalls in capacity, especially in import-constrained areas.7 By implementing the capacity market, PJM was able to procure enough capacity to meet and exceed the requirement by attracting a substantial influx of new, low-cost resources. These resources included increases in net imports, uprates to existing generation, and demand response resources. Few analysts had anticipated so many low-cost resources. Their entry is a testament to the creativity of competitive markets.

Securing a large quantity of low-cost resources postponed the need for new generation investments for almost a decade. More recently, new generating capacity has been needed due to load growth, retirements, and limited additional capacity available from existing resources. Capacity prices have risen sufficiently to attract those investments. Even so, capacity prices remain substantially below the system operator's estimates of the long-run cost for new generating plants. For example, PJM's recent auction for the 2017/18 delivery year attracted nearly 6,000 MW of new generation commitments at prices that were 35–41% of PJM's estimated net cost of new entry (Net CONE).8 This further demonstrates the competitive market's success in maintaining resource adequacy in a cost-effective manner.

We now address the specific concerns noted in the Senators' question:

  • Increments of New Generation Cleared versus Built. The Senators state that they are concerned that less generating capacity will actually get built than has cleared in PJM and ISO New England's forward capacity markets. It is true that that some of the capacity commitment cleared in the forward auctions will likely be bought out in incremental auctions and thus not get built; some of it may also come online with a one or two year delay.9 While one of the APPA reports referenced by the Senators suggests that a lower quantity of realized capacity additions would demonstrate that FERC and the system operators have to revisit the resource adequacy procurement mechanisms, we are less concerned.10

    First, we note that there generally will be a difference between the quantities cleared and built. The magnitude of that difference is likely to be modest or consistent with a decline in load forecasts for the delivery year. The APPA reports do not attempt to quantify this magnitude or explain the reason for any difference, perhaps because the timing of the reports would have made such a comparison impossible. The latest APPA report was issued in 2014, but it was not until the 2015 delivery year that significant new generation was committed to come online (a point acknowledged in the report).11,12 It is still too early for a complete comparison of the quantities cleared in forward capacity markets versus the generation actually built. But as a partial comparison, approximately 18,000 MW of new (not refurbished or life-extended) traditional thermal capacity has cleared PJM's capacity auctions starting with the delivery year 2015/16.13 That compares to 13,500 MW that have either come online or are currently under construction.14 In other words, the majority of the plants committed in prior capacity auctions are online or are being built; and others have additional time before they will need to begin construction to fulfill their future commitments.

    Second, if there is some discrepancy between original commitments and actual construction, it is most likely related to the fact that PJM's three-year load forecasts have been overstated compared to the subsequently-revised forecasts for the delivery year. As a result, PJM has procured more capacity in the three-year forward auction than what was actually needed.15,16 As load forecasts for the delivery year have been revised downward, some of the new generating units that cleared in the auction can buy out of their commitments or postpone their online dates. This is a more cost-effective outcome than requiring the originally-committed plant to proceed with construction if the plant is no longer needed based on the updated load forecast. However, persistently over-forecasting loads imposes additional costs and is therefore undesirable. Recognizing this concern, PJM has been working to address the issue through enhancements to its load forecasting methodology.
  • Prices Below the Net Cost of New Entry (Net CONE). As the Senators noted, prices have been below the administrative estimates of Net CONE in most of the capacity markets for most auction years in ISO New England, New York, and PJM.17 We do not see this as a concern for these three markets. Rather, we view this as evidence of beneficial competitive market performance. It may be disappointing to generation owners hoping for more financial support from these markets; but, from a customer's point of view, quality of service has been high and less expensive than if prices were clearing at the administrative estimate of Net CONE. Each of these markets has maintained low prices while meeting or exceeding reliability requirements, thus over-performing in both dimensions. We anticipate that in future years, average prices will rise to levels that are sustainable in the long run, but market forces will determine whether that long-run average price is above, below, or exactly at the administrative Net CONE estimate.18

    In the Midcontinent ISO (MISO), however, we take a different view. Prices in MISO's capacity auction have been consistently near zero and are not likely to rise sufficiently to attract new generation investments when needed. In most of MISO, capacity needs are satisfied through state resource planning efforts by regulated, vertically-integrated utilities such that there is no need for additional capacity to be attracted through MISO's capacity auction. That is, the capacity market in MISO is not really the prime driver of entry or expansion decisions. Rather, it is more of a balancing market for temporary variance in the timing or performance of assets being developed for other reasons, under state requirements. However, a modest portion of MISO LSEs do need to rely on market-based capacity, and so may fall short of their requirements if prices cannot rise sufficiently to attract entry once the current capacity surplus is depleted. MISO has identified this concern, and we recommended a series of reforms to address the issues.19 MISO has recently issued its own proposal to stakeholders.20
  • Generation Being Built Under Contract with Load Serving Entities (LSEs). It is not correct that new generating capacity has been built only under bilateral agreements with LSEs or under direct ownership by LSEs, although this was likely the case up until the 2011 and 2013 periods examined by the APPA studies. Until those years, competition from lower-cost resources had postponed the need for new generation, which meant that no private entity would make an investment without a long-term contract. Thus, in those years with excess supply, only regulated entities with cost recovery were building generation or signing contracts to build new generating plants.

    More recently, we have seen substantial investments in new merchant generation resources. Of the 13,500 MW of traditional thermal capacity recently built and now under construction, approximately 11,000 MW are new merchant generation.21 In ISO-NE, 4,050 MW of new generation has cleared in the last five auctions.22 As mentioned previously, most of these resources have now begun construction.

To read this letter in full, please click here.

Footnotes

1 Murkowski and Cantwell (2015).

2 This letter is not sponsored by any client company. It reflects the views of the letter's authors and not necessarily the views of other members of The Brattle Group or any of its clients.

3 American Public Power Association (2012); American Public Power Association (2014).

4 The purpose of capacity markets is to procure sufficient capacity to meet resource adequacy requirements, and to do so cost effectively by allowing all qualified MW competing to meet the need. These markets were never intended to directly address other policy objectives, such as fuel diversity or environmental quality. Such other objectives have to be addressed through other means, which can be implemented to complement capacity markets.

5 For a review of the experience with the first decade of capacity market operations, see Spees, Newell, and Pfeifenberger (2013).

6 For examples of recommendations to improve the existing capacity market designs, see Pfeifenberger, et al. (2014); Spees, Newell, and Lueken (2015); and Pfeifenberger, Spees, and Newell (2012).

7 See a more detailed discussion of this history in Pfeifenberger, et al. (2011) and PJM Base Residual Auction Results.

8 See Newell, Oates, and Pfeifenberger (2015).

9 New York ISO will not have any similar cases given its near-term capacity market design.

10 American Public Power Association (2014), p. 5.

11 Neither of the two pieces of evidence cited in the APPA report demonstrates any difference between the quantity cleared and the quantity built. The first APPA point was that a non-public internal projection from a third-party consulting firm (ICF) assumed that not all of the cleared capacity would get built. This assumption may prove accurate or inaccurate in retrospect. Even if accurate, this does not indicate whether this would be a problematic outcome. The second APPA point was that more generation projects were cancelled than built between 2008 and 2012. This does not acknowledge that it is common in all regions that many more projects will be proposed than completed. As in other industries, only the most competitive projects will tend to move forward. The report fails to note that no new generation was actually needed over that period as discussed above.

12 As noted in the second report "the 7,700 MW of planned merchant generation that cleared the last two auctions in PJM appears to mark a dramatic change in the pattern reported in this study." American Public Power Association (2014), pp. 4–5.

13 See PJM Base Residual Auction Results.

14 These resources are primarily natural gas-fired combined-cycles in the range of 300 to 900 MW in size, as well as 120 MW of combustion turbines. We identified units as being non-merchant if they were owned by a public power entity, were listed as "regulated," or were owned by the traditionally-regulated Dominion utility. We believe that this has screened out resources that are supported by regulated cost recovery, but we have not undertaken a more thorough review of each project. Data source: ABB, Inc., Energy Velocity Suite.

15 As an illustration of the magnitude of over-forecasting, see Newell, Oates, and Pfeifenberger (2015), p. 9. Drivers of the downward revisions in load forecasts include recognition of the long-term effects of the economic recession, policy-driven energy efficiency investments, and lower energy intensity associated with new economic growth.

16 A much more problematic situation would be if a substantial quantity of new generating capacity were committed and actually needed for the delivery year but failed to come online without procuring replacement capacity. We have not yet observed such outcomes, except under an ISO-NE provision that explicitly allows for a delayed online date in certain circumstances. PJM and ISO-NE have included measures in their market designs to protect against such outcomes, through qualification requirements, credit requirements, milestone tracking during construction, and penalties for non-delivery. If such undesirable outcomes were to arise, these market design elements would need to be refined.

17 The statement that prices have exceeded Net CONE only once is not correct. ISO-NE prices exceeded Net CONE in the 2016/17 auction for new and existing resources in NEMA and in the 2017/18 auction for new resources in the whole ISO and existing resources in NEMA. PJM prices have exceeded Net CONE in the ATSI Zone in the 2015/2016 auction and above zonal Net CONE for numerous MAAC regions in the 2013/14 auction.

18 It is somewhat unclear why prices in these markets have remained consistently below administratively estimated Net CONE, despite new capacity being built, and the reasons likely differ by market. Capacity offers reflect how much the entrant needs to be paid in the capacity market to be willing to enter, given its costs and its anticipated net revenues from energy and ancillary services markets as well as future capacity prices. Offers may differ from administrative estimates of Net CONE for a variety of reasons. The relatively low offers could reflect lower capital costs, lower financing costs, higher anticipated net energy revenues, technological innovation, different technology types, or greater optimism about future capacity prices than assumed by the system operator in these estimates. It may be that the relatively lower net costs could be a transitional effect as low-cost opportunities are developed first.

19 The concerns are driven by a combination of a vertical demand curve, a non-forward market, and a low price cap. These issues do not affect the ability to meet capacity needs in most of the MISO footprint where regulated utilities build new generation under traditional resource planning processes. These are likely to raise resource adequacy concerns for the approximately 9% of the loads in MISO that will rely on market-based investments to meet capacity needs. We have recommended reforms to address these concerns in a recent report see Spees, Newell, and Lueken (2015).

20 See MISO (2016).

21 We identified units as being non-merchant if they were owned by a public-power entity, were listed as "regulated," or were owned by the traditionally-regulated Dominion utility. We believe that this screened out resources that are supported by regulated cost recovery but have not undertaken a more thorough review of each project. Data source: ABB, Inc., Energy Velocity Suite. As another comparison point, for the past three PJM auctions spanning delivery years 2016/17 to 2018/19, PJM's capacity auctions cleared 13,600 MW of new generation and uprates in total, of which 11,230 MW was merchant and 2,370 MW was LSE built or contracted. Prior to those years, a substantial quantity of new generation did clear in prior PJM auctions, but the large majority of those resources were likely LSE self-supply (although PJM did not report precise statistics on the portion designated as merchant until the 2016/17 auction). See PJM Base Residual Auction Results.

22 ISO-NE Forward Capacity Auction Results, see ISO-NE (2016).

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.