Market design, bidding rules, and long memory in electricity prices

In uniform price, sealed-bid day-ahead electricity auctions, the market price is set at the intersection between aggregate demand and supply functions built by a market operator. Each day, just one agent - the marginal generator - owns the market-clearing plant. Day-ahead auctions are moreover embedded in multi-segment systems, wherein diverse protocols coexist and change over time. Such a complex environment leads to adoption of simple, adaptive bidding rules. Specifically, such a market design lets two different types of routines emerge, depending on whether the agent is a likely marginal or inframarginal generator. However, because of the uniform price mechanism, only the bidding behavior of the former can be reflected into market prices. Depending on the specific way marginal generators process past information to set their bids - 'hyperbolic' or 'exponential' - electricity prices are likely to display long- or short-memory. Experimental evidence on hyperbolic discounting - a quite robust behavioral bias in humans - supports a long-memory view of electricity prices. This insight is broadly confirmed by spectral analysis of daily data from NordPool and CalPX markets, in sharp contrast with most previous empirical studies. This paper underlines the importance of institutional settings in determining market outcomes, and an interesting mapping of bidding rules and models of information processing into the time series properties of market prices.