High Frequency Trading: Impact of Fragmentation and Hidden Liquidity

Krzysztof Herman

Raja Velu

 

The focus of our research is the multi-stage decision process that models the submission, cancellation and execution of a trade. Recent changes that have occurred in the equity markets call for a new approach that would focus on the issues related to the presence of hidden liquidity and market fragmentation. Specifically, we investigate how the proliferation of trading venues and the extensive use of hidden liquidity affect the decision of (i) how to slice a parent order, (ii) whether to submit a limit or a market order and (iii) when and where to submit the order. With the advancement of computing technology, trading has accelerated to millisecond durations. We use Level III market data collected from four major US exchanges (NASDAQ, Bats, EDG-X and A) to study the dynamics of the market via the dynamics of the limit order book. We find evidence of interaction in order flow across venues and of the use of multi-venue trading strategies by fast traders. These findings highlight the difficulty for sophisticated investors to find and access liquidity in today’s post Reg-NMS equity markets and the necessity to monitor and act on several venues simultaneously in order to attain an acceptable quality of execution. We follow the recent work of van Kervel (2015) to identify the fast traders via clustering of orders and build a logistic regression model to study the intra-venue and inter-venue characteristics that influence the probability of cancellations.

JEL Codes: G10; G12; G15;

Level III market data, also called “Message Level Market Data”, represents a considerable improvement over previous Level II or TAQ data. It’s a collection of almost all the messages generated by an exchange and it allows to match each reported event to a specific limit order, thanks to the use of a unique order ID identifying each order submitted into the book.

Keywords: Market microstructure, Fragmentation, High Frequency Trading, Trading algorithms


Newsletter Subscription

How many eyes does a typical person have? (ex: 1)
Name:
Email: