Working Paper: NBER ID: w21996
Authors: Ulrich Doraszelski; Gregory Lewis; Ariel Pakes
Abstract: We document the evolution of the newly created market for frequency response within the UK electricity system over a six-year period. Firms competed in price while facing considerable initial uncertainty about market demand and rival behavior. We show that over time prices stabilized, converging to a rest point that is consistent with equilibrium play, and then adjusted to subsequent changes in the market quite quickly. We draw on models of fictitious play and adaptive learning to analyze how this convergence occurs and show that these models predict behavior better than Nash equilibrium prior to convergence.
Keywords: frequency response; market equilibrium; learning models; bidding behavior; electricity market
JEL Codes: D02; D22; D83; L10; L51
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| strategic uncertainty (D89) | heterogeneous bidding behaviors (D44) |
| heterogeneous bidding behaviors (D44) | frequent adjustments (F32) |
| reduction of uncertainty (D80) | price stabilization (E64) |
| learning mechanisms (C45) | convergence to Nash equilibrium (C72) |
| fictitious play models (C73) | better prediction of bidding behavior than Nash equilibrium (C70) |
| price bubble (E32) | competition drives bids down (D44) |
| experience and information about rivals' reactions (D80) | improvement of equilibrium models (C62) |