An Examination of the Strategic Behavior of High-Frequency Traders (HFTs)

This project was part of the team project "Complex Markets: Regulation and Incentives in Secondary Market Design".

Topic and Objectives

In electronic markets, trading firms invest heavily in the speed of market access by employing cutting-edge hardware, relying on real-time market data feeds, and subscribing to exchanges' colocation facilities to place their servers in the exchanges' datacenters. Relying on this technology, institutional investors focus on trading large positions at the lowest possible execution costs, and High-Frequency Trading (HFT) firms attempt to capture the bid-ask spread while supplying liquidity or profit from short-term momentum. While the academic literature has explored the role of trading speed in the context of HFT, its role for institutional investors and broker-dealers remains unexplored. We attempt to fill this gap. We investigated how the execution costs for large orders traded in a proprietary or agency capacity are affected using exchanges' colocation facilities. We also examined the effect of HFT on execution costs and whether this effect differs by exchange members' trading capacity and access to colocation.

Key Findings

• Colocated market participants achieve substantially lower execution costs than non-colocated ones when executing large orders on their own accounts.
• While execution costs of such orders generally increase (decrease) with the amount traded against liquidity consuming (supplying) orders by high-frequency trading firms, the use of colocation helps traders substantially offset these effects.
• However, these results do not hold for observationally similar order flow originating from buy-side customers, who do not benefit from their brokers' use of colocation, potentially raising concerns about whether brokers provide the best execution.
• The buy-side's failure to achieve the same execution performance might be due to sub-optimal use of execution algorithms or to the sell-side offering inferior algorithms to their customers.