Testing the Performance of Uniform Price and Discriminative Auctions

The high prices that occurred in southern California since the Summer, 2000 led to a substantial amount of regulatory and political intervention. Price caps were lowered and the Federal Energy Regulatory Commission (FERC) proposed that a new type of "soft cap" auction should be adopted. This auction combines a standard uniform price auction with a discriminative auction for offers higher than a specified level ($150/MWh). Nevertheless, there is little available evidence to show that this new auction works well, or guarantees lower average prices. The objective of this paper is to provide some experimental evidence about the relative performance of different types of auctions for electricity markets. The experiments involved engineering and economic graduate students at Cornell University and the University of Illinois, and regulators in the New York State Department of Public Service. These individuals represent generators in a "smart" market, POWERWEB. This market replicates the physical constraints of meeting loads in an electrical grid. The first part of this paper describes how the high price volatility observed in many electricity markets can be replicated. The key features are 1) load is stochastic, 2) incentives are provided to withhold capacity from the market, and 3) the price is determined by a uniform (last accepted offer) price auction. The results with six identical generators in the auction show 1) price spikes are common, and 2) average prices are higher than competitive levels. This confirms the belief that electricity markets with totally inelastic demand need more participants than typical markets to ensure competitive prices. The second part of this paper describes experiments using four different types of auctions. In a "smart" market, the total cost of meeting load is minimized, subject to operating constraints, in all cases. The four auctions are 1) a uniform price auction with price inelastic load, 2) a uniform price auction with price-responsive load, 3) a discriminative auction in which generators are paid their actual offers instead of a uniform clearing price (commonly, and incorrectly, called pay-as-bid), and 4) a "soft-cap" auction combining a uniform price and a discriminative auction. The main result for all three groups of participants shows that both the uniform price auction (1) and the discriminative auction (3) produce average prices fifty percent above competitive levels. However, the prices for the uniform price auction are more volatile with many price spikes. The soft-cap auction (4) has price characteristics similar to the discriminative auction. In contrast, the uniform price auction with price-responsive load (2) has lower average prices (about thirty percent above competitive levels). The lower price volatility associated with the discriminative auction and the soft-cap auction is caused by the flatter offer curves in these auctions. However, this flat shape is likely to undermine the effectiveness of demand conservation as a way to reduce average prices. The uniform price auction with price-responsive load is the best among the four auctions tested, because it produces the lowest average price and has relatively low price volatility.