CRM Toulouse

CENTRE DE RECHERCHE EN MANAGEMENT
CENTER FOR RESEARCH IN MANAGEMENT

ECOLE DE MANAGEMENT  - IAE
Université Toulouse 1 Capitole - UMR 5303 CNRS

CENTRE DE RECHERCHE EN MANAGEMENT
CENTER FOR RESEARCH IN MANAGEMENT

ECOLE DE MANAGEMENT  - IAE
Université Toulouse 1 Capitole - UMR 5303 CNRS

Algorithmic Trading



Contents

Presentation 
About the ANR project "Algorithmic Trading"
Workshops 
Research articles
Team



Presentation

Algorithmic Trading and the quality of financial markets


Figure 1 These pictures respectively represent a high frequency trader and a financial market nowadays: HFT’s strategies are algorithms that are programmed on super-computers. Orders are matched on the « market », that simply is a collection of servers in a highly secured room.



Understanding the impact of ‘algorithmic trading’ on financial markets: Does it improve market quality? Should it be regulated? How?

Since the mid-2000s, technological innovations and changes in regulation have fostered the development of ‘algorithmic trading’, that implement automatic trading strategies conditional on market conditions, via powerful computers. First, trading became electronic in most stock exchanges in the late 1990s, enabling financial institutions to access historical market data and to automate the execution of their buy and sell orders. Second, the regulations RegNMS in the U.S. and MiFID in Europe (both implemented in 2007) have fragmented the markets, requiring the development of tools to manage liquidity supply (for dealers) or liquidity demand (for brokers) across trading venues (see Fig. 1b). While algorithmic trading was spreading across the equity markets, little was known about its potential societal benefits and risks. The objective of this research project was to analyze the impact of algorithmic and of high-frequency-trading on financial markets. Does it improve market quality? Should it be regulated? If so, what would be the policies to implement? The answer to these questions would have policy implications for the regulators. 

Combining economic modelling to understand the potential effects of regulatory policies, and empirical studies to test the theoretical predictions.
The project relies on two main approaches, namely 1) theoretical models, 2) empirical analysis of intra-daily data. 
Economic modelling enables to understand and analyses the various tradeoffs at play with the introduction of algorithmic trading. In particular, it sheds lights on indirect effects on market quality that may be counter-intuitive. Second, this further offers a framework that is suitable for policy implications1. Third, theoretical modelling enables to draw some empirical predictions that would guide empirical research. 
This approach is complemented by empirical studies based on high-frequency intra-daily data. First, descriptive statistics are useful to quantity a completely new phenomenon2. Second, we use econometric tools (e.g. panel regressions, probit regressions) to test the hypothesis derived from the predictions of theoretical models. Third, the empirical results may suggest new directions for future (theoretical) research, which would help deepening further our understanding of algorithmic trading. Given the nature of the data (e.g., time-stamped at the micro-second level), this empirical analysis relies on new developments in high-frequency econometrics.


A project that has adjusted to changes in its environment
‘High frequency trading’ (hereafter HFT), combining algorithmic trading with speed, soon became a key issue for regulators3. On the one hand, its exponential growth has open debates (that recently spread to public opinion) such as: Is there a risk for market stability? Are financial markets still a level-playing-field? On the other hand, many large brokers and asset managers started complaining about their difficulties to execute large trades, and routed most of their order flow to non-lit and/or unregulated venues, such as dark pools or Over-The-Counter (OTC) markets. This further raised concerns on market fragmentation for regulators.
In this context, we have enlarged our set of research questions to tackle some of these new issues. Besides, we have often been invited by various regulators to participate to debates and discussions on HFT4


A project that established our expertise on the issue of HFT
The project has enabled the publication of 4 articles, and of 5 working papers. Our team has taken part to more than a hundred talks on these articles. Our expertise on the market microstructure of financial markets (and more particularly on high frequency trading) is recognized at the international level, as shown by the number of invitations that we received (keynote lectures or invited sessions of international conferences, research seminars, discussions, and by regulators), and others prizes and distinctions.




1 For instance, we show that fast trading increases adverse selection on the market, generating a negative externality and an arms’ race in investments in speed. However, we also find that opening “slow-only” platforms is not an optimal regulatory reaction to the development of high-frequency trading.
2 We address for instance the following questions: What is for instance the market share of algorithmic trading, or of HFT, in financial markets? Are firms investing in speed heterogeneous? Can we identify regularities in the trading behavior of HFT firms?
3 This is even more true as it now accounts for more than 70% of stock trading volume (http://www.usatoday.com/story/money/business/2013/09/09/high-frequency-trading-proposal/2787127/). Competition currently occurs at the micro-second level.
4 E.g. Autorité des Marchés Financiers, Banque de France, International Organisation of Securities Commissions, European Commission, see below.