We propose a state space modelling approach for decomposing high frequency trading volume into liquidity-driven and information-driven components. Based on a set of high frequency S&P 500 stocks data, we show that informed trading increases pricing efficiency by reducing volatility, illiquidity and toxicity/adverse selection during periods of non-aggressive trading. We observe that our estimated informed trading component of volume is a statistically significant predictor for one-second stock returns; however, it is not a significant predictor for one-minute stock returns. This disparity is explained by high frequency trading activity, which leads to the elimination of pricing inefficiencies at high frequencies.
JEL Classification: G12; G14; G15
Keywords: trading volume; expected component; unexpected component; volatility; liquidity;
market toxicity, unobserved components time series models; high-frequency data.