Clustering of Trade Prices by High-Frequency and Non-High-Frequency Trading Firms

We examine clustering of transaction prices in a sample that contains high-frequency trading firms' transactions. We separate our sample into four categories: transactions with a high-frequency trading firm on both sides of the transaction, on only one side of the transaction (either liquidity provider or liquidity demander), or on neither side of the transaction. We find that transaction price clustering is less frequent when a high-frequency trading firm is on both sides of the transaction than when a high-frequency trading firm is on only one side of the transaction or if a high-frequency trading firm is not involved in the transaction. Further, we find that transactions where the liquidity providing order, which more likely dictates the price, is submitted by a high-frequency trading firm cluster less than when the liquidity providing order is submitted by a non-high-frequency trader. We thus conclude that the tendency of prices to cluster appears to be driven by a distinctly human bias.