Adverse Selection vs. Opportunistic Savings in Dark Aggregators

Abstract

Dark aggregators provide access to liquidity opportunities at multiple venues—yet the prevalence of high-frequency trading operations has raised concerns that these benefits may be lost to adverse selection. This article provides a methodology to separately measure opportunistic savings and adverse selection costs. These measures are far more accurate and less subject to selection bias than implementation shortfall. The authors find that in the case of dark aggregators and broker algorithms, most of the benefits of dark liquidity are counterbalanced by adverse selection costs. They analyze results from random marketneutral basket trade experiments conducted with Pipeline’s Algorithm Switching Engine to measure implementation shortfall without selection bias. The Engine uses a nonlinear model to predict participation rates and reduce adverse selection by switching into algorithms that are expected to perform well in the current market conditions. Predictive switching eliminates two-thirds of adverse selection costs relative to the continuous use of dark aggregators with only a small loss in opportunistic savings, resulting in a 40% reduction in the implementation shortfall. The authors show that a modified aggregator that blends traditional dark pools with displayed-market access to enforce minimum and maximum participation rates was able to provide full access to the benefits of dark liquidity while reducing adverse selection by 37%, contributing to a substantial improvement in trade performance for the desk.

TOPICS: Statistical methods, factor-based models, exchanges/markets/clearinghouses