Competitive equilibrium and the double auction

TL;DR

This paper challenges the assumption that double auctions always produce competitive equilibria, showing through experiments that actual prices often deviate from theoretical predictions, especially under certain conditions.

Contribution

The paper provides empirical evidence that double auctions do not necessarily generate competitive equilibrium prices, highlighting the importance of the Marshallian path in price formation.

Findings

Prices can deviate significantly from equilibrium predictions in stationary value distributions.

Trader exit without replacement reduces the effectiveness of equilibrium convergence.

Monotone shifts in value distribution may not affect prices as theory suggests.

Abstract

In this paper, we revisit the common claim that double auctions necessarily generate competitive equilibria. We begin by observing that competitive equilibrium has some counterintuitive implications: specifically, it predicts that monotone shifts in the value distribution can leave prices unchanged. Using experiments, we then test whether these implications are borne out by the data. We find that in double auctions with stationary value distributions, the resulting prices can be far from competitive equilibria. We also show that the effectiveness of our counterexamples is blunted when traders can leave without replacement as time progresses. Taken together, these findings suggest that the `Marshallian path' is crucial for generating equilibrium prices in double auctions.