Testing the Effects of Holding Forward Contracts On the Behavior of Suppliers in an Electricity Auction

Abstract

The objective of this paper is to test how spot prices are affected by forward contracts using experimental economics. One set of tests used students to represent suppliers in an electricity auction with 1) no forward contracts, 2) permanent forward contracts, and 3) renewable forward contracts. In the latter test, the price of a new forward contract is affected by conditions in the spot market. An identical set of tests was also conducted using computer agents to represent all of the suppliers. The objective was to demonstrate that computer agents can be used effectively to test electricity auctions and do additional sensitivity tests to supplement the results obtained using humans. The results show that holding a forward contract is an effective way to mitigate high prices if the same contract is held for all trading periods and the price of this contract is independent of the spot prices (i.e. is fixed). However, there is more speculation and the spot prices are higher when a forward contract is renewed periodically and spot prices influence the forward price. With a renewable forward contract, a price spike increases a supplier's current earnings in the spot market and expected future earnings from a new forward contract. The overall conclusion is that long-term bilateral contracts will reduce speculative behavior, but this effect will be dissipated in an active forward market with a large amount of secondary trading. The results using computer agents were encouraging. In the first set of tests, two students competed in each market with four computer agents. In almost all cases, the average earnings of the computer agents were higher than the average earnings of the students. In the tests with all computer agents, two computer agents replaced the students, and these agents (latent speculators) were more likely to speculate than the other computer agents. The average spot prices and average earnings with all computer agents corresponded closely to the highest values obtained by the students.