Does Algorithmic Trading Attenuate Asset Price Bubbles? An Experiment

Abstract

We experimentally examine the impact of algorithmic trading strategies on asset price mispricing and the relative payoff of human vis-a-vis algorithmic traders. We imple- ment a 2x2 treatment design varying the algorithmic strategies into a market-making (liquidity-making) and a sniper (liquidity-taking) algorithm, and the algorithm’s speed into an instantaneous and a 5-second speed-bump algorithm. Our benchmark treat- ment is the one with only human traders. We show that the market-making algorithm results in roughly 21% lower mispricing than the benchmark. The sniper algorithm reduces mispricing significantly less than the market-making algorithm. Imposing a 5-second speed bump on the algorithm has no impact on mispricing. Except in the sniper treatment, the algorithm generally does not outperform human traders