Market predictability and non-informational trading

This paper studies the ability of non-informational order imbalances (buy minus sell volume) to predict daily stock returns at the market level. Using a model with three types of participants (an informed trader, liquidity traders, and a finite number of arbitrageurs), we derive predictions relating returns to lagged returns and lagged order imbalances. Empirical tests using New York Stock Exchange non-informational basket/portfolio trading data provide results consistent with adverse selection at the market-level, but no evidence of limited risk-bearing capacity. Finally, we establish that these market-wide non-informational order imbalances also affect individual stock return comovement by examining additions to the S&P500 Index.