Learning about adverse selection in markets

How does a market learn about the number of informed traders and thus adverse selection risk? Using a model, we show that trade sequences convey information about adverse selection risk. Consequently, buy/sell order imbalances can destabilize markets, triggering extreme price movements, flash crashes, and liquidity evaporation. The increasing prevalence of these effects in markets can be explained by more active learning about adverse selection by competitive, high-frequency market makers. We use our model to estimate the uncertainty in adverse selection risk for US stocks and show that it decreases market liquidity and increases extreme price movements.