Lights Out: Conditional Execution Failure \\in Dark Pools Under Market Stress

Abstract

This thesis examines whether the mechanical heterogeneity of dark pool execution, the specific protocols through which individual trades are matched, shapes how dark liquidity behaves during market stress. Using second timestamped trade prints from a single off-exchange venue matched to millisecond NBBO data from the NYSE TAQ database, I document two distinct channels through which lit market deterioration affects dark execution. Price volatility fragments institutional order flow into smaller prints without reducing total activity, consistent with traders splitting parent orders to manage execution risk. Meanwhile, quote instability disrupts the conditional matching protocols on which complex dark orders depend, reducing their observed share and worsening adverse selection for the orders that do complete. At the trade level, complex dark orders exhibit significantly negative signed price impact at all forward horizons, and this adverse selection cost roughly doubles during stress. Using FINRA ATS weekly volume reports, I show that agency electronic pools lose nearly five percentage points of market share during high VIX weeks while agency block pools gain approximately the same amount. Because the two shifts cancel, aggregate concentration measures appear stable, masking a redistribution that is most prominent for stocks reliant on complex execution and that unwinds within two weeks as volatility subsides.