Information, liquidity, and high frequency trading

In my first essay, we examine the effects of investor disagreement on price discovery using a recurring public information event in the highly liquid crude oil futures market, a market free of short-sale constraints. We show that prices reflect positive news within one-half second of trading, but continue to drift for five minutes when news is negative. Evidence suggests the drift arises from a systematic surge in buying pressure that impedes the price discovery process when news is negative. Our results are consistent with price drift arising from differences in investment horizon, where traders taking long positions condition trades on information beyond the news. In the second essay, using intraday trading activity for S&P 500 stocks, we find evidence of trade clustering in the first second of trading at discrete time nodes (10:00, 10:30, etc.). These unusual volume spikes are pervasive, varying in magnitude across both the cross-section and time-series. Regression results show that news, order imbalance, and stock characteristics do not fully explain this variation. We explore using these spikes to construct a proxy for cross-sectional variation in HFT activity. Empirical tests suggest that this daily measure (which can be computed across a broad range of stocks) captures a dimension of daily HFT activity related to liquidity provision.