SPECIAL
SERIES
Part 3
Energy Risk & Markets
Default Retail Supply:
The BGS Auction Enigma
New Jersey's recent basic generation service auction
shows how ignoring the many sources of risk can
be financially ruinous.
March 2005
By Tom Brady
Bidding at last year's basic generation service (BGS) auction in New Jersey was generally found to be extremely aggressive as many merchant energy providers watched in amazement as the bid prices continued to fall during the course of the auction. Many speculated on the reasons for the low bid prices-was it a thirst for market share no matter what the cost? Was it
a sophisticated application of bidder-specific risk profiles? Or was it simply poor modeling? This article describes the sources of risk inherent in load-following deals (such as temperature and price variations, etc.) and how these risks can impact resulting bids in such auctions, which are a relatively new phenomenon. In fact, several states around the country such as Illinois, Ohio, California and Maryland either experimented with such an auction structure or are proposing similar
structures for the immediate future.
In the case of New Jersey, the state implemented retail choice in 1998, adopting a four-year transition period. During the first three years of the transition, New Jersey's electricity distribution companies (EDCs)1 supplied those customers that did not switch to competitive providers using pre-established rates or rate-making processes. During the fourth year of the transition (2002-2003), the EDCs jointly proposed an auction where suppliers competed to supply BGS.
An essential goal of this auction process is to provide full-requirements BGS to customers at a cost consistent with market conditions. A second important goal is to further the transition in New Jersey by providing an opportunity for energy trading and marketing companies to compete in
the auction by aggregating the supply
of wholesale products and providing price-risk management services.
The state's four EDCs participate in this auction process, whereby portions of their load requirements are divided into individual tranches upon which suppliers may bid during the auction for the right to serve over terms of 1 or 3 years.
The BGS-Full Price (FP) auction, which is known as a clock auction, is designed to solicit offers for the supply of full-requirements tranches for smaller commercial and residential customers for each of the four EDCs. The initial round of an auction begins with the posting of a starting price.2 During the bidding phase(s) of an auction, the
auction manager will announce prices. After the bidding phase for a given round, if the number of tranches bid
is greater than the number of tranches needed for a given product,3 the auction manager will reduce the price of
a tranche by a given decrement (0.5-5.0%) and a new bidding phase begins. At the end of each round, participants may reduce the number of tranches bid. The auction continues until the total number of subscribed tranches falls to the point where it equals the number needed. When the auction ends, those bidders holding tranches
at prices of the final round are the
auction winners.
Suppliers under load contingent arrangements such as those for the BGS ACECO full requirement supply tranches are exposed to numerous sources of risk, including:
n Seasonal, daily and hourly load variations;
Temperature swings;
Daily and hourly price variations inherent in electricity markets; and
Supply disruptions due to congestion, unit outages, etc.
Using models that consider these risks are essential when determining potential bid prices in supply auctions and/or contract negotiations.
Using Monte Carlo simulation methods, multiple analyses are completed to examine the impact of including various risk elements (or layers) on expected bid prices for the 2004 BGS ACECO FP 1-year supply tranche.3 To replicate attempting to obtain meaningful lower-bound bid prices prior to the auction date, each analysis is completed assuming a simulation start date of
Feb. 2, 2004.4
Using both historical or market implied volatilities, two simulation models (Multivariate Normal and seasonal multi-factor) are used to simulate ACECO on and off-peak forward electricity prices. The 1-year supply tranche is then simulated using a mean/variance approach (Risk Layer 1) or an auto-regressive model that is driven with a temperature dependant variance (Risk Layer 2). Subsequent risk layers include hourly load and price shaping, intra-month (delivery) spot price volatility and multi-state regime switching of spot electricity prices. Under each risk layer, the tranche is valued based on simulated loads and forward curves and spot prices.
Table 1 displays estimated break-even bid prices for the ACECO 1-year supply tranche using the multi-variate normal model to simulate forward prices. Table 2 shows the results obtained if the seasonal multi-factor model is used to simulate ACECO
on -and off-peak electricity prices.
Also displayed on each table is the
New Jersey Board of Public Utilities approved 2004 auction price for this
1-year tranche.
Risk Layer 1 simulates expected load based on random draws from a historical load distribution defined by a mean and variance, as shown in Table 1. Under this base scenario, estimated bid prices of 5.752¢/kWh and .749¢/kWh are obtained using historical and market implied price volatility assumptions, respectively. These compare to the 2004 BGS Auction results for this 1-year tranche of 5.473¢ /kWh implying bid winners could end up losing significant amounts on this deal.
With the inclusion of a more sophisticated load model (temperature dependant - Risk Layer 2), along with additional risk layers, the discrepancy between the actual 2004 and model estimated bid prices increases. The incremental inclusion of a 3-regime spot price model (Risk Layer 6) results in estimated break-even bid prices of 6.300¢/kWh and 6.296¢/kWh using the two different volatility estimates implying bid winners could end up losing significant amounts on this deal.
Using a seasonal multi-factor model to simulate forward prices results in slightly lower breakeven bid prices than those using the multivariate normal model; however all are significantly higher than the 2004 final Auction
price (Table 2).
Estimated values are shown in Table 5 using the 2004 Final Auction price over the ACECO 1-year tranche term. As seen in this table, using the multivariate normal method to simulate
forward prices shows increasing levels of losses as risk layers are added.
Further, under the Multivariate
Normal approach, using either historical or market implied volatilities, the fully layered model (adding the
3-regime spot price model) results in losses over $3.7M. (Similar ranges
of results are shown with using the
Seasonal Multi-Factor model to
simulate forward prices).
These results demonstrate the financial impact of ignoring or incorrectly modeling the many sources of risk inherent in supplying load services for this particular 1-year tranche.
Tom Brady is a Senior Manager at Risk Capital Inc. where he develops software solutions for energy merchants. He has
a Ph.D. in mineral economics from
Colorado School of Mines. Contact him
at tbrady@riskcapital.com.
Endnotes
- These companies include: Atlantic City Electricity Co. (ACECO), Jersey Central Power & Light, Public Service Electric & Gas (PSEG) and Rockland Electric.
- As stated on the BGS Auction website, during an initial round, for any statewide auction, prices will be set no lower than
a minimum of 6.6¢/kWh and no higher than
a maximum of 9.3¢/kWh.
- Per BGS Auction Rules (www.bgs_auction.com), current tranches for ACECO are approximately 96 MW (4.55 percent of their total load). The term of this supply tranche is from June, 1, 2004 - May, 31, 2005.
- Each analysis used various configurations of Risk Capital's AcuRisk simulation software. All parameters and forward curves were based on data prior to this simulation start date. The auction occurred in early February 2004. All simulation parameters and forward price curves are based on data prior to this auction date.
Articles found on this page are available to subscribers only. For more information about obtaining a username and password, please call our Customer Service Department at 1-800-368-5001.
Public
Utilities Fortnightly & Spark
