Modeling High-Frequency Trading Activity (13w5008)

Arriving in Banff, Alberta Sunday, September 1 and departing Friday September 6, 2013


(Simon Fraser University)

(Olsen Ltd)


Given the proprietary nature of high-frequency trading firms, there is relatively little known in the public domain about how they operate. At the same time, their importance in financial markets has grown very quickly in the past decade. As a result, the public understanding of financial markets has fallen somewhat behind the times. The purpose of this BIRS workshop is to establish a frame of reference to begin improving our understanding of the new dynamics of financial markets, of which high-frequency trading is a central contributor.

We will bring together experts from the fields of economics, finance, mathematics, and statistics to establish what already exists in terms of techniques and results that improve our understanding of high-frequency trading. Just as important, we will begin exploring what we do not know, but would like to know, about these traders. That is, we hope to formalize questions, approaches, and research directions that will begin new research collaborations that will continue after the workshop is over.

A vital component of the workshop will be the participation of high-frequency market participants. Their comments and feedback will assist the researchers understand the objectives, constraints, and strategies which need to be modeled. The market participants will benefit from improved research in the high-frequency domain since it may improve their operations and risk management. In addition to academics and industry participants, the workshop will invite several graduate students who are interested in pursuing research in this area.

The workshop will consist of presentations by participants in several important areas which are described below. In addition to formal presentations, we will organize round-table discussions and provide ample time for open discussions and unstructured collaboration. Each of the five days will have a central topic related to high-frequency trading. These topics are:

1. Understanding Market Participants and the Trading Environment:
The first step in modeling high-frequency trading is to clarify the objectives of the participants and the constraints under which they operate. Some of the activities high-frequency traders engage in are market making, arbitrage, directional speculation, and exploiting structural features of the markets. There are also regulatory and structural features of the market that can impact the validity of a formal model. The trading environment has become very complex in the past decade with, for example, multiple trading venues trading a single security and each venue having substantially different features.

2. Risks and High-Frequency Trading:
What are the primary risks faced by high-frequency traders? Are there new systemic risks that high-frequency traders create? What risk management techniques are available to traders and market regulators? How important is speed, and what are the costs of being slower than your competition.

3. High-Frequency Statistical Techniques:
High-frequency market data have several features that make them difficult to analyze. Transactions and quotes are irregularly spaced in time requiring joint modeling of price and duration. Microstructure effects must also be modeled or else these effects impede estimation at high-frequencies. In addition to this, quote data has high dimensionality, and price data may be coming from multiple trading venues.

4. Limit Order Books:
Many important trading venues are organized as an electronic limit order book. Order books are characterized by a multitude of strategic interactions and are difficult to model, though there is a lot of work being done in this area. Competition between limit order books is also relevant to high-frequency trading since trading venues often pay high-volume liquidity providers fees which can substantially impact the profitability and behavior of traders. Modeling order flow and high-frequency dynamics of an order book is vital for many high-frequency trading strategies.

5. Market Impact, Market Models, and Trading Algorithms:
At high-frequencies, even moderately sized trades can have permanent and temporary market impacts. As a result, simulation and back-testing of high-frequency strategies must account for this impact. Second and higher-order effects may also need to be considered since the market impact may induce new orders or cause other to be canceled. Models of market impact are central to decisions on how to trade optimally.

For each of these topics, the objective of the workshop is to (1) present the state-of-the-art results and techniques that are available, (2) stimulate discussion on what would be valuable additions to the literature, and (3) establish interdisciplinary collaboration between researchers and industry participants.