Algorithmic trading

0 Comments • Wednesday April 30, 2008 • by Camille Caclin

Long before the subprime crisis and the huge losses of a French bank, behavioral finance has challenged one of the foundations of the current economic model: Homo economicus. This utopian representation of individual behavior is so persistent that the mathematicians were replaced by an algorithm. However, whether direct or hidden behind a mathematical tool, human intervention is not she fallible?

Homo economicus is a perfect being. It is rational and optimally treats the information and decisions regarding the management of its portfolios. There is, however, that professional and amateur traders react to the contrary, says Patrick Roger ^1, behavioral finance researcher at the Laboratory of Research in Economics and Management (WIDE). For over a decade, scientists at this laboratory has conducted studies on behavioral finance, that is to say the application of psychology to finance. They show the imperfections of the choices made in relation to financial risk. Everyone has flaws, says Patrick Roger. When he makes a decision, he was influenced by his overconfidence, like a lucky player. It may also give more weight to information that is in line for his opinion (conservative) or search only for information to corroborate his belief (confirmation bias).

Homo economicus is it a robot?

To remove the vagaries of human behavior and evacuate all liability, specialists have developed a tool for portfolio management: algorithmic trading. It involves using a series of complex mathematical operations to analyze and manage portfolios of securities to more or less short term. It is coupled to a computer system that sifts through the keywords from the news agency dispatches.

Why?

There is nothing more perishable than financial information. Once the newspapers have announced an event, it is too late. The impact of the news is already in the share price, says Elsa Poupardin, professor of information science and communication, member of LISEC ^2. The dispatches, in turn, are available in real time. Depending on the number of occurrences detected and their context (growth or crisis), the price of a share will potentially increase or decrease. The algorithm examines the data stream to translate it into a binary choice: to buy or sell. Whatever the method used, the quality of the algorithm depends on the data entered. There is much left to the designer's imagination. Garbage in, garbage out [garbage in, garbage out], warns Patrick Roger. A raw information does not make sense, says Elsa Poupardin. For example, the price of wheat does not depend exclusively on its production but also the weather, geopolitics, ecology, etc.. A broad general education in several areas is necessary to contextualize and make the right decision for the purchase or sale of shares. Finally, the man behind a clever formula. His irrational behavior are present in the base of the algorithm and how it is fed with information.

Patrick Roger quips: "From a stock proverb: better to be wrong with the market that reason alone. Many models work only because a significant number of people using them, even if based on a false assumption. "

¹ Dean of the Faculty desSciences Economics and Management, Professor of Finance

² Laboratory interuniversitairedes educational science and communication.

Article available at: www-ulp.u-strasbg.fr/actualites/pdf/ulp-sciences/ulp-sciences-31.pdf or download [download id = "1"]

Share it!